There’s just not enough games against too strong of opposition to swing the balance too heavily in their favor. I’ll leave it to others to argue whether this be good or whether this be bad.

Does anyone think it’s unfair? Winning 100 games in a season and then beat by an 82-80 underdog in 5-games series…

tyc,
no, but I really like that sentence.

I think it is “unfair” in that there are three divisions.  I’m comfortable with just “leagues” and the best record plays a WS (like pre-1969).

I’m sad to say I saw Ozzie Smith say this (that he preferred just the two teams with the best records play for the WS) on, gulp, Quite Frankly.

I also think that the times have changed enough that attendance would be sustained with that situation (that’s not to say we could have gotten to where we are without it though).

The amount of disposable income, and the country’s attitude toward entertainment dollars have moved so far from where it was in the 1960s and before, would allow baseball to thrive with just the league setups - I think.

Then we could have the Cardinals and White Sox going at it right now for all the marbles.  OTOH, yes, if you have more games to show on tv, then you make more money.

I can see the argument for an extra round given expansion. 

I’d like to have seen the addition of leagues, rather than divisions, when expansion got going (though I guess that is tough given the slow pace of expansion).  So, I see 4 8 team leagues.  Each team plays the other 7 teams in its league 13 times.  That team then plays the remaining 24 teams 3 times. 

Then the four winners play a two layer playoff.  Both rounds best of nine. 

Really, whatever system you have, there should never, ever, ever be a best of five playoff series. 

Chris, why were you sad to see Ozzie say what he said?

Really, whatever system you have, there should never, ever, ever be a best of five playoff series.

One of the interesting things I took from Voros’s tables above is how little difference it theoretically makes whether you play a five- or seven-game series.

I think Chris was sad to say that he watched “Quite Frankly.”

One of the interesting things I took from Voros’s tables above is how little difference it theoretically makes whether you play a five- or seven-game series.

What table?  I don’t actually see any article or link that this thread is attached to.  Which I have found odd.

Voros had a couple of win-probability tables at the top of his article.

Had? 

And, as I’ve said over and over, I don’t find those win-probability tables all that convincing for short series’ or, especially, for single games.  I think they work well for large samples, but you start to see “quantum” effects when you bring the analysis down to small samples.

What table? I don’t actually see any article or link that this thread is attached to. Which I have found odd.

Voros is on the Home page.

Thanks, Repoz.  Still, I don’t see anything at the top of the thread.  No info on who posted it, no info on how many posts.  Nothin’.  It just seems odd.

I guess my question is:  have home page articles always appeared on the sidebar?  I didn’t think so, though I should have thought to look at the home page.

Bunyon,
yes,it was because I was watching QF.

No, front page articles just started in teh sidebar.

There’s something ironic about beating to death the ‘cursory’ nature of the analysis and then ‘announcing to the world’ that your conclusion is right…

No, front page articles just started in teh sidebar.

That really is an excellent addition.  Thanks, Jim!

There was some discussion of Voros’s article late in yesterday’s Dugout.

Had?

Yes, ‘cause I’m done wit’ ‘em, dere yers now.  8-)

There was some discussion of Voros’s article late in yesterday’s Dugout.

That discussion continues.  Here is a post I made over there this morning:

I’m not sure anyone other than Backlasher has argued that luck has had no impact on the A’s (or Braves). And I think his argument is one of semantics in that he says events happen and you have to be prepared to deal with them. Which I think is a salient point, whether we refer to it as events happen or luck. In a short series, you have to be prepared for the bounces to go against you and react much more quickly than in the regular season.

Take Cox, one of his greatest attributes as a manager is his willingness to give players a lot of chances to impress. He rarely makes a snap decision with new players and remains calm in the midst of long streaks of mediocrity. Which I think is all great in the regular season but the kiss of death in the playoffs. I’ve seen him stick with his aces when they were getting shelled. I’ve seen him stick with a really cold bat throughout a losing series. Hell, yesterday he started Brian Jordan. I simply think he has a philosophy, it is well suited - damn near perfect, actually - for the regular season but he doesn’t adjust to the new realities of the postseason. That is, when events happen against him or he experiences bad luck, he doesn’t react quickly or well. The same is true, I think, for the A’s.

Sure, the A’s could have won any of those series if they catch a few breaks. But a good postseason manager won’t rely on catching breaks to win, they’ll take action to try to force events to happen in their favor. This is where those of us who don’t like the constant reminder of the luck factor have a problem. It is fine to acknowledge that luck is a factor. But, having acknowledged that, it is the duty of a manager to plan for it and react to it, not just to throw their hands up.

There’s something ironic about beating to death the ‘cursory’ nature of the analysis and then ‘announcing to the world’ that your conclusion is right…

Yes, welcome to the days of pre-classic primer.  I thought I had stepped into one of those Steve Treder time machines.

A team’s starting pitcher will influence the odds much more so than any other single variable.  This is just another variant of the Blossom article that had the A’s as 5:1 favorites over the Orioles and 3:2 favorites over the Yankees.

There are so many problems, I don’t know where to begin:

(1)  Teams are in flux:  This isn’t a static team.  These teams built their records with players coming and going, and in some cases, with only a subset of the players, or better replacements, playing at the relevant time.  Some may be in periods of better performance, or worse performance due to injury, fatigue or rest.

(2)  Still a small sample;  Even if these were robots with similar output, 162 games is a pretty small sample, based on the metric.  It would be like looking at 2 seasons worth of Greg Maddux data and concluding that pitchers have no ability to prevent hits on balls in play.

(3)  Matchups:  As bunyon mentioned, one of the micro effects is that you are really dealing with batter/pitcher/park/weather matchups more than anything else.  Some players may be very good at hitting HRs in 12-0 blowouts against a KC AAAA pitcher that was called up due to injury.

If the point is that on “Any Given Sunday” any team can win, I think everyone knows that.  I wouldn’t need a cursory or even an extended look at a probability table to know that.  And I doubt that these teams have calculable odds at the rate of that table.

So, I’m not sure of the goal of this piece.  The article is definately conclusatory, and it contains misleading elements.  And it has a very textbook tone to it, like its teaching the unwashed masses a lesson.

whatever system you have, there should never, ever, ever be a best of five playoff series

Back when 5-game series started, it took several years before an underdog won a series.  The first time a team with a “worse” record won was in 1972, when the Reds beat the Pirates—and the Reds were all of a half-game “worse.”  I remember a lot of people thinking, back in 1969-71, that the 5-game series (which was often a sweep in those days) was a pointless coda to the regular season, because the best clubs were just going to beat their league opponents easily in the playoffs anyway.

1973 corrected that impression, but was seen as an anomaly.  The Reds and Yankees and Dodgers clubs of the late 70s were pretty serious champions, and the victories they rang up in 5-game LCS play were largely seen (correctly or not?) as evidence that they were legitimate pennant winners.

I’m not quite sure when this sense began to erode, but my memory is that the LCS came to seem less “legitimate” in 1985 and after, when the LCS expanded to seven games.  That seems counterintuitive, and maybe coincidental.  But the ‘85 Royals were a relatively weak champion and the ‘87 Twins an absolutely weak champion.  The ‘88 Dodgers shouldn’t have beaten anybody, &c.

The three-tier playoffs have made the crapshoot nature of things glaringly evident, with wild-card
teams winning pennants and Series as often as not lately.

In any case, I loved the “old” pre-1985 LCS format.  You didn’t want to start with two on the road because you didn’t want to start with two on the road.  OTOH, you didn’t want to start with two at home because you knew you would have to win the series on the road if at all.  And the games seemed to have so much more at stake than the current LDS or LCS games; that’s an illusion from a fan’s perspective, I know (the players wouldn’t say that), but nowadays, it’s like LDS, so what, stupid crapshoot, and LCS, that’s fine, but the postseason marketing plus interleague play have made it into just a semifinal rather than a true league championship.

Yet individual games and series remain potentially exciting as hell, which goes to show that baseball is a hard product to break or have go stale.

Yet individual games and series remain potentially exciting as hell, which goes to show that baseball is a hard product to break or have go stale.

This is a good point.  A single elimination tournament of the top 16 teams in the league would also likely produce great baseball and a lot of excitement.  It still wouldn’t be a good way to find a champion. 

FWIW, I never liked the 5 game LCS series.  The 5 game series also seems illogical to me.  If there is no difference between 5 and 7 games, why have the LCS and WS be 7?  If there is a difference, why should the best team in the league have to play the 4th best in a 5 game series?

Make them all 5 or all 7.

And I think his argument is one of semantics in that he says events happen and you have to be prepared to deal with them.

I may not articulate it well, but I have multiple different points when it comes to this “luck” that keeps getting thrown out.

(1) Contingency Planning -  You have mostly covered this ground.  The only thing that I would add is that even using the macro tools, it would be an extremely small possibility that you will not get an injury to one of your contributors.  The key to the game is not going and getting the twelve best VORP players.  The key is to have enough contributors to put yourself in position to enter and win the tournament at the end of the season.  Over six months, you have to plan that some parts are going to fail.

(2) Lack of Analysis

This is the most frustrating one.  “Luck” has this amorphous definition.  Apparently, one of its meanings is “everything external to a team’s control, and everything that I haven’t figured out how to model.  I could live with the first half of the definition, even though its still laziness.  It blurs the need and the reward for contingency planning.  It blurs the need for converting opportunities when they present themselves.  But the second is just foolish.  If you haven’t analyzed something properly don’t blame a predictable outcome on luck.  This article illustrates a classic usage.  Rather than examine any specific set of outcomes, or prospective matchups, we just decide that if a .500 team plays a .500 team, its a coin toss on who will win.  Luck is not some holding bucket for things you didn’t analyze, or analyzed improperly.  For instance, pitchers BABIP is not luck; it has a skill component.  Because you improperly determined it was random, or its too hard for you to identify skill does not mean that you have a random outcome.

(3) Excuse Generating

No matter what the interval space may be, no matter what evidence is at your disposal, luck is the catch all excuse for performance failures and analytical failures.  Even if you use these undergraduate statistical techniques to show the improbability of a series of events occurring, it will still be explained as, “well that shows it could happen, so it is just luck.”  Yet other series that are even less remote are chalked up to skill.  With that criteria, I could do the same thing.  If I am allowed to use those rules, I can construct a proof that shows the Royals were the best team in baseball, and the Yankees are the worst team over the last decade.

(4) Misdirection of Analysis

Forecasting is one really small piece of analytics.  That seems to be the only thing most saberists care about.  They even analyze such things as decision science using only forecasting metrics.  If you have a space where you can derive important information, they will claim that you cannot derive anything because of “luck”, occasionally they might say “noise” here too.  This is especially true if you show one of the more hair brained conclusions is wrong.  That is just pure sophism.

(5) Terminology and Creeping Excuses

This is a big problem that just isn’t semantics.  As I previously mentioned, “luck” has an amorphous definition for saberits.  It expands or shrinks to fit a rhetorical need.  “Noise” is also misused.

YOu have events.  There are past events and their are prospective events.  By definition, there is nothing random about a past event.  The space is closed.  Prospective events may be determinate, indeterminate or partially determinate.  Indeterminate and partially determinate events may have random sub-events (unless your are studing irreducable quantums), subevents we treat as random, or known events.  In the grand scheme of things, there are truly few, if any at all, random events.  As the time to the measured event approaches 0, most all factors are known and the event is (at least theoretically determinable).  Most randomized sub-events are merely treated as random because of our imprecise instrumentation, the cost of performing the analysis, or the inability of the analyst to determine the causal property of the event.

  In past events, the information is either complete, partially complete or incomplete.  That is mostly always due to instrumentation.  The method of delivering the information can lose data, or not lose data.  The method of delivering the information can have superflous elements.  The latter is “noise”

  If an actor has a series of low probalistic outcomes, then you might define that series as luck.  If those outcomes are fortunate, you may say the actor is lucky.

  By and large, I normally wouldn’t care about precision.  But in analytical discussions, using methods and arriving at conclusions, it becomes pretty damn important.  Did you determine that a pitcher has no influence over an event due to not having enough information being collected; due to your own failings as an analyst; because you introduced error into the calculation; because you lost information when you retrieve it, etc.

  If you want someone to value a conclusion and make decisions based on it, we need to know how much garbage is in your data, and how good the analyst or machine is at arriving at the conclusion.  You can never get to that area, because all things are thrown into a luck basket, which is then imbibed with these unchallengable attributes like its the Will of God.  Then if you pierce through all of that and really take the long path to unwind what may be faulty sets of assumptions and error introducing methods, and determine a utility for the conclusion or debunk the conclusion, luck is then turned into a sword to reject the better information.

  So if you ask me what the the St. Louis RF will do against his opponent in the 2nd at bat on June 21, 2009, I’ll tell you that is pretty indeterminate.  I’ll tell you that there is a lot of intervening events that I can only treat as random, and I’ll spit back out whatever the current league average is for the RF position in MLB.  I’ll also tell you there isn’t much utility in that outcome.

But if you ask me the same question on June 21, 2009 while said batter is in the ondeck circle, I can give you a better estimate.  And I won’t just use the current leage average for the RF position.

If you wait until the person is at bat and the moment just before bat hits the ball, and you give me a fast enough processor, and all information about the spin and vector of the ball and the swing, I’ll give you a damn good estimate.  If you wait until after the play is over to ask me, I can report the event with 100% accuracy.

<u>I’m not sure anyone other than Backlasher has argued that luck has had no impact on the A’s (or Braves). And I think his argument is one of semantics in that he says events happen and you have to be prepared to deal with them. Which I think is a salient point, whether we refer to it as events happen or luck. In a short series, you have to be prepared for the bounces to go against you and react much more quickly than in the regular season.</u>

For the record, I believe I was the originator of the anti-luck position and continue to espouse it. BL has argued it more forcefully and longer than I, but I’m in virtual agreement w/all his points. I brace when I read at this site that such and such team was “lucky” to have finished first or wherever, b/c their “true” record would have them second or wherever. That’s nonsense on stilts and it comes across as pure rationalization in pseudo-scientific cloth.

If this is where we vote on this, I’ll pique JC again and say that I think unpredictability plays a very important and generally unappreciated role in both nature and human affairs, including baseball. No science can afford to overlook it.

I think unpredictability plays a very important and generally unappreciated role in both nature and human affairs, including baseball.

I don’t know what you are saying Mark.  In what we deal with on a daily basis, there are more partially determinate events than any other.  There are precious few determinate events, and there are a sizable number of indeterminate events, at least based on the present availability of cost of the information.

I doubt anyone disagrees with that contention.

Both today and in the future, these things will remain the same in terms of comparitive sizes.  There are some things that we can prove are intractable without even reaching unsolvable.

Again, I doubt anyone disagrees with that contention.

And I’m not sure how that is relevant to the discussion regarding “luck”.  I’m not sure how that is relevant to a discussion on macro win probabilities or some pseudo-measure like a DIPS.

But, you go here:

No science can afford to overlook it. 

And you lose me completely.  You don’t overlook anything. I’m not sure its possible to overlook it.  In fact, what many are stating is that the junk science being spouted by some, is IGNORING critical components, and components of available information.  More important, they are rationalizing or defending conclusions by improperly categorizing events as being part of some “luck force.”

For all intents and purposes, there have been many things over history that were partially determinate or then indeterminate, for which we have found ways to develop more precision.

If we got in Treder’s time machine and went back to pre-recorded history, I bet we could have a nice FarmerForum.  I bet you would have some people that would determine that Field Factors play an important role in crop growth.  They would announce it with pompt and circumstance, like everyone hadn’t already come to some conclusion that were you plant your crops are important.

But some of those yayhooos would claim that storms and temperature were all luck, and you couldn’t influence crop growth based on that information. 

Some of the farmers might look at the moon or other environmental clues and get a better feel on seasons and rain patterns.  They may not write up there results, but they may have better outcomes.

Then the yahoos would exclaim, that the reason their farmers did bad was because of the Will of Zabronius, who would know it was going to be below freezing and kill the crops, how would now that it would rain so much during that time.

That is precisely what you have here.  Today, you have some yahoos, that are armed only with statistical techniques for large samples.  They will perform a few operations and overstate their results.  Sometimes they will even introduce error into the results.

Then they will make large self-serving conclusions that aren’t supported by even their erroneous choice and application of techniques.  Then they will use these conclusions and the spectre of “luck” to rationalize outcomes they don’t like.  If you tell them and show them how they are wrong, they then use “luck” like its a weapon.

That is rhetoric, not science, and as a practioner of rhetoric, you should understand this distinction. 

Science is not overlooking anything.  The sophists are ignoring relevant factors.  And they have invented a mythological force they call luck that they use as a weapon whenever they get in a corner.

Then when

There are two sources of uncertainty levels:
1 - the uncertainty of a true mean
2 - the random variation around the true mean

In both cases, we can never get an uncertainty level of 0.

The true mean is determined based on data on hand.  Unfortunately, players don’t get to pile up 1000 PA per game.  As well, we can’t give each pitcher 100 different fielders each game.  Furthermore, a player, as a human being, is subject to whatever affects humans at any moment in time.  So, to get to a true mean, for that particular moment in time, is in fact impossible.  You don’t know what Bonds is thinking or feeling, you don’t know how Pedro woke up, and you don’t know how much RJ hates the umpire.  All you can do is say that Bonds’ true mean is x, and my certainty level is y, z% of the time.

For any particular game, you have to add up the uncertainty levels of all the players in question, as well as your expectation as to how often they will play that game.  So, again, you’ve got a wide band around your true mean. 

For the second source, this one is far more wide, in terms of random variation.  Even if you could establish the exact true mean of every player and every team, “anything can happen” at any moment in time.  And with such few trials (80 PA per game), the random variation will be prominent.

What you are left with, in terms of a manipulator, is:
1 - try to establish the true mean of every player
2 - try to establish a model that interacts these players
3 - try to select the players that gives you the best chance to win a series of these models

What will happen, in any given series, no one would bet the farm on.  Give someone 99 series, and sure, take the favorite, and you probably could bet the farm on it. 

Otherwise, the great thing about the drama of baseball and hockey is the uncertainty of the true means, and the random variations around those means.

There are two sources of uncertainty levels:
1 - the uncertainty of a true mean
2 - the random variation around the true mean

In both cases, we can never get an uncertainty level of 0.

For all intents and purposes, I can and would agree.

However, this is the problem.  First of all, you have to understand that even if your uncertainty is 0 or infinity (depending upon one’s range), there are event spaces where P=1, if P<1, by definition it is a partially determinate event, although that doesn’t really matter at the extremes of the intervals.

The only reason that I am pounding this minor point is that the articulation of your true statement is still going to be misunderstood.

These folks will never even think about P(x|Y) and will never calculate that d(P(x))/dt or moderate anything based on a time scale.

But more important, as long as P(S) < 1 at any time based on a calculation just based on retrievable information from ESPN.com, these folks will luck as the proximate cause of any undesired result and skill in some broad technique (e.g. market inefficiency) as the proximate cause of desired results.

Then we will get little simplified Davisesque pieces like this to create some type of proof system on the existence of their luck god.

what many are stating is that the junk science being spouted by some, is IGNORING critical components, and components of available information. More important, they are rationalizing or defending conclusions by improperly categorizing events as being part of some “luck force.”

I think this is the crux of the second part of your post; all else appears to be elaboration on this theme. If I’m wrong about this, let me know.

Whenever we can’t foresee a result, we have two options: look for a previously neglected causal factor; recognize that some events are not predictable except in probablistic terms.

Of course we should always look for potential causes. People who discover those generally win prizes. But sometimes there is no identifiable cause, it’s just “luck” (I prefer “unpredictable”). It’s possible to err in either direction; nobody should dismiss “luck” as a factor.

To Tango’s post 25 I’d add that there are other sources of unpredictability as well, e.g., wind, temperature, spin, etc. So many factors that I seriously doubt anyone could even identify them all, much less account for them with certainty. I agree with Tango that this very lack of certainty is what creates the drama.

And you lose me completely. You don’t overlook anything. I’m not sure its possible to overlook it. In fact, what many are stating is that the junk science being spouted by some, is IGNORING critical components, and components of available information. More important, they are rationalizing or defending conclusions by improperly categorizing events as being part of some “luck force.”

Yes. When the topic first arose, I recall it went something like this: People pleaded luck on Beane’s behalf, and said essentially he’d done all he can do. I responded that that was an unsatisfactory approach, and indeed an approach that would be unacceptable to anyone striving for excellence. No manager or executive wants to hear from his subordinate “We just got unlucky again.” I stated at teh time that Beane’s job was to minimize the space in which luck could operate. You don’t just shrug your shoulders and concede, you move forward.

So what I believe I and BL are objecting to is not the phenomenon that there are variables beyond our capacity to control, but both the breadth some people give to those variables and (more importantly for this site) the use of that breadth of those variables to excuse the failure of the golden calf. Beane’s job remains, AFAICT, to produce a team capable BOTH of winning the regular season and succeeding in the playoffs. His teams not only have not done the latter, they’ve not experienced ANY winning in the playoffs. This differs even from Cox whom some slam b/c of his failures to win the WS. Something is amiss; or put differently, were I his employer, I’d want something better from Beane than “the playoffs are a crapshoot,” particularly since they don’t seem to be a crapshoot for WHOMEVER PLAYS US!

Whenever we can’t foresee a result, we have two options: look for a previously neglected causal factor; recognize that some events are not predictable except in probablistic terms.

No, when stated in this binary system, you have two choices:

  (1) Discern the causal factors; OR
  (2) Determine there is no net benefit from further research.

You are re-articulating and stating something further down the decision tree. 

The event space is closed, there are always causal factors.  Some are external to your control, some are too expensive to control, some are intractable to produce a success in your time interval for decision.

You examine all these things to determine how to build or alter your plan for the optimum output.

But when you sit back and only look at a players at bat outcomes divorced from available factors to consider, and make a long range projection, you have already made one analytical mistake.

When you take this result and then imply that events that are later known, and causally connected, such as:

  (1) Factors toward improving playoff outcome;
  (2) Impact of management and coaching;
  (3) Impact of med. policies and plans; and
  (4) Ignoring skills.

You are making critical errors.  Then when you claim these four things are luck, and use that as a rationalization, you have become annoying.

Use this article as an exemplar.  You are taking performance results that were available over a year ago, and citing a probability to win ratio that is not accurate.  Then that is used to build to the conclusion that the “Playoffs are a crapshoot” by some analogy to fair dice.

And then if some team that was picked to win based on pet theories winds up losing, it will be because of this “luck force.”

I can geometrically expand the errors and the problems if we would like to go through the history of DIPS.

Of course the longer your projection, the more uncertainty you have.  That is an innate characteristic of a partial differential equation.  These projections are dependent on external events, like decisions of others, and time.  I wouldn’t even want to try to figure out the change in rate on how many variables this amounts too.

But, unlike us, GMs aren’t all external, they can exhibit control.  Moreover, don’t use some dated model when you have more information.

If the goal is to show the bounds of a GMs influence on a playoff series, that is theoretically calculable.  But showing a probality table from a sophomore level math class not only doesn’t show that bound, it doesn’t even add any relevant information to the discussion.

Its a rhetoric piece only.

If you employed him JC, would his past record and attempts to explain, “playoffs are a crapshoot”, lead you to fire him?  Would you say that we’ve got to get better by a set number of years, or else?

If you employed him JC, would his past record and attempts to explain, “playoffs are a crapshoot”, lead you to fire him? Would you say that we’ve got to get better by a set number of years, or else?

Fair question, PV. I wouldn’t fire him straight away, since there are certain things I don’t know. But I do consider Beane a very good GM, and unless I had an alternative I thought could be better, I’d retain him but also try to change certain of his biases.

I guess I’m just angry today, I’d probably say “OUT, NOW!!!”

I can’t understand why my DMB teams always finish in last place.

If you employed him JC, would his past record and attempts to explain, “playoffs are a crapshoot”, lead you to fire him? Would you say that we’ve got to get better by a set number of years, or else?

That depends on my organizational goals.  I’d think at this point, the Beane administration looks like its good at Sheparding talent.  I would probably give him a shot at re-implementing a new talent base.  Then I think I’d give him a very short window to reach a reasonable output.

Otherwise, before he let the next wave of talent depart, I’d think about investing in that base, and putting in a management team that could take me to the next level.

I’m not sure I’d be happy at a Sisiphus cycle, but I wouldn’t fire Sisiphus before the rock was close to the top.

And I’d let him know that.  There wouldn’t be anything hidden.  And I’d probably let him know I’d keep him along as VP of Player Development.  But no more books unless you win a fukking championship.

Something is amiss; or put differently, were I his employer, I’d want something better from Beane than “the playoffs are a crapshoot,” particularly since they don’t seem to be a crapshoot for WHOMEVER PLAYS US!

I definitely agree with this.  And something else that has gone unsaid is that it is likely that all of the folks getting slammed - mostly Beane, a little Cox/Scheurholtz - probably are trying to figure it out and correct.  The “crapshoot” comment from Beane strikes me as a frustrated guy at a frustrated point and not organizational philosophy.  If, indeed, Beane’s philososphy is that there is nothing that can be done to prepare for the playoffs then, yes, absolutely he should be fired.

When I made a comment about semantics above, I did so from my view of luck.  Luck is not a win/lose in the series.  Luck is all the events taking place.  Say, Zito makes a pitch to a location that Manny Ramirez will hit well 33% of the time.  If you do this 100 times, you’ll get him out more often than not.  In the playoffs, there will be a single event (or just a few).  If, in that one instance, Manny hits a HR, you were unlucky from one POV.  However, as JC and BL have both said, you can’t simply say that you should have won.  The goal of a champion is to have a good enough team that even when a number of those breaks go against you, you can still win.  Obviously, if they all go against you, you won’t win.  But if all such events go against you, then you probably had the wwrong numbers going in.

And, not to take a real old-school approach to this new-fangled learning, but what is entertaining about a system that is random.  If it really is a crapshoot, then why play the games?  Why not throw some dice?

And, not to take a real old-school approach to this new-fangled learning, but what is entertaining about a system that is random. If it really is a crapshoot, then why play the games? Why not throw some dice?

The physiological response to uncertainty in a time constrained interval is the production of adrenaline, which has a euphoric effect.

The psychological response to certainty is often boredom, ennui, depression and feelings of powerlessness.

If you can’t view the full text of Voros’ article, it’s because you’re caught in a flux state between the old templates for the site and the new templates. I must have inadvertently coded the article to with the new status which allows featured material to display on both the frontpage and on Primer.

If you’d can’t view the full article, you need to access it from either the old setup or the new setup.

Although I closed off interaction between the old setup and the new setup, this particular article slipped through the cracks.

I should be done with the redesign over the weekend, so this problem shouldn’t crop up again.

Sorry about the confusion.

Here’s where I disagree with JC and Backlasher: the argument that “the A’s have lost 4 series in a row, the probability of that happening is so low”, it can’t be “just luck”.

That’s the wrong way to look at it IMO.  It’s the same logic that says, there wasn’t any luck to Joe DiMaggio’s hit streak, because the probability of such a streak is so low it couldn’t be due to luck.  But, it’s not nearly so improbable when you consider how many opportunities all players had to do it, one of which happened to be DiMaggio.  Maybe the long streak was 2% Joe D, Rose, Cobb, Brett, etc .... and 1% Gwynn, Williams, Lajoie, etc ...

Sure, DiMaggio may be slightly more suited to creating a hit streak than other players of similar overall ability, just as the A’s may be slightly more suited to losing a playoff series than other teams with similar regular season winning ability.  But still, DiMaggio was lucky—not that he got those hits, but that he was the one who got those hits, not someone else who could have done so.  Similarly, the A’s have been unlucky—not that they lost all the time, but that they were the team that lost, not someone else who could have done so.  I hope that makes sense.  Teams making the playoffs once or more 1999-2004, with series record

Marlins 3-0
Angels 3-1
Yankees 11-4
Red Sox 5-3
D-Backs 3-2
Mets 3-2
Cards 4-4
Cubs 1-1
Giants 2-3
Braves 3-6
Mariners 1-2
Astros 1-3
Twins 1-3
Indians 1-3
White Sox 0-1
Dodgers 0-1
A’s 0-4

That’s pretty much what you’d get if the playoffs were a crapshoot once you got there.  The A’s just happen to be the one at the bottom.

Something is amiss; or put differently, were I his employer, I’d want something better from Beane than “the playoffs are a crapshoot,” particularly since they don’t seem to be a crapshoot for WHOMEVER PLAYS US!

That’s a fundamental mischaracterization of the point.  The point is that, given the amount of teams making the postseason, if it’s a crapshoot, even if they are all equally good, we expect one of them to lose 4 series in a row.  That one of them did lose 4 in a row does not invalidate the hypothesis.  If Beane didn’t write that book, it would be like if the Twins lost every series instead of all but one, and the A’s were 1-4: no one would use “but they’ve never won” as a counter, because the team that’s never won isn’t part of the whole ridiculous saber rules/sucks debate and is just the bad equivalent of the Marlins, the extreme case that’s what you get with 17 teams in the postseason (hopefully I didn’t forget any).

In summary:  The playoff results may tell you something about Beane and A’s, but only if you are interested in Beane/Oakland in particular and not anyone else.  But if you take the names off the team, it all looks like workings of luck (or as Mark prefers, uncertainty, that’s what is meant by luck in this context).

If the goal is to show the bounds of a GMs influence on a playoff series, that is theoretically calculable. But showing a probality table from a sophomore level math class not only doesn’t show that bound, it doesn’t even add any relevant information to the discussion.

To some extent, I agree with everyone in this discussion. IMO, the amount of common ground is being obscured by the lack of a specific question to discuss.

So, let’s say you are actually trying to forecast a playoff series. To a first approximation you can take the regular season records of the teams. To a second approximation you can look at the actual pitching matchups and take into account the roster changes. If you want, you can continue with any of a jillion other factors.

The question is, are any of these other factors big enough to have a noticeable impact on the odds of winning the series? Examples: the manager’s flexibility to changing his approach, the batters’ skill in dropping down a bunt, the closer’s ability to hold up under extreme pressure, etc. Of course all these will affect the odds. The question is how much.

This article is about the first approximation. Contrary to what BL says, I think the first approximation is extremely important. So is the second approximation. That doesn’t mean a GM should pay no attention to the other factors. He should. But until I am convinced otherwise, I will continue to believe that the cumulative effect of the other factors is relatively small in most cases, and therefore that first or second approximation calculations have a good deal of merit.

So what I believe I and BL are objecting to is not the phenomenon that there are variables beyond our capacity to control, but both the breadth some people give to those variables and (more importantly for this site) the use of that breadth of those variables to excuse the failure of the golden calf.

I understand that. And what I’m objecting to is the constant claim that there “must be” some variable hidden to Bobby Cox (and the rest of us) for which he failed to account and which, if he had only been “better/more prepared”, would have led to more World Championships.

My view is that if there is such a variable, it’s incumbent on those who think so to identify it and prove it, not just multiply hypotheses. If they don’t do that, then the rest of us are entitled to set aside that claim until someone does. That which is unknown at this time is unpredictable/lucky; Stan Kasten is fully justified in saying that postseason play is a “crapshoot”.

The event space is closed, there are always causal factors.

As this is stated, I disagree. Even in paradigmatic determinate systems like Newtonian gravity or quantum mechanics, future states of the system can be unpredictable except in probabilities. This is NOT because there are unknown factors we have failed to identify as causes, it is because they are intrinsically unpredictable.

We may actually be in agreement, however, because you go on to say, correctly, that

“Of course the longer your projection, the more uncertainty you have. That is an innate characteristic of a partial differential equation. These projections are dependent on external events, like decisions of others, and time.”

To bring my point full circle, it’s far too easy to say that there must be some X factor which (fill in your favorite underperforming manager) didn’t understand. Recognizing unpredictability takes courage.

I agree with DCA and with this statement by fret:

“... I think the first approximation is extremely important. So is the second approximation. That doesn’t mean a GM should pay no attention to the other factors. He should. But until I am convinced otherwise, I will continue to believe that the cumulative effect of the other factors is relatively small in most cases, and therefore that first or second approximation calculations have a good deal of merit.”

Re #37:
According to Stephen Jay Gould, Joe D’s streak is not a very good example for the point you’re making:

My colleague Ed Purcell, Nobel laureate in physics but, for purposes of this subject, just another baseball fan,[2] has done a comprehensive study of all baseball streak and slump records. His firm conclusion is easily and swiftly summarized. Nothing ever happened in baseball above and beyond the frequency predicted by coin-tossing models. The longest runs of wins or losses are as long as they should be, and occur about as often as they ought to. Even the hapless Orioles, at 0 and 21 to start this season, only fell victim to the laws of probability (and not to the vengeful God of racism, out to punish major league baseball’s only black manager).

But “treasure your exceptions,” as the old motto goes. There is one major exception, and absolutely only one—one sequence so many standard deviations above the expected distribution that it should not have occurred at all. Joe DiMaggio’s fifty-six–game hitting streak in 1941. The intuition of baseball aficionados has been vindicated. Purcell calculated that to make it likely (probability greater than 50 percent) that a run of even fifty games will occur once in the history of baseball up to now (and fifty-six is a lot more than fifty in this kind of league), baseball’s rosters would have to include either four lifetime .400 batters or fifty-two lifetime .350 batters over careers of one thousand games. In actuality, only three men have lifetime batting averages in excess of .350, and no one is anywhere near .400 (Ty Cobb at .367, Rogers Hornsby at .358, and Shoeless Joe Jackson at .356). DiMaggio’s streak is the most extraordinary thing that ever happened in American sports.

Backlasher makes some excellent points.  Teams are in flux and matchups are crucial.  I would only add that matchups themselves are in flux.  The reason a guy who goes 0-for-16 is sometimes “due” is that he says, dang, I’ll go with the pitch for once and slap it into that big gap they’re giving me in the opposite field, instead of pulling everything I see.  And then he gets hot for a while and then everyone adjusts and counter-adjusts some more.

With reference to Beane, however, Beane is in flux.  (That doesn’t sound good.)  Four straight years, the A’s expired in a fifth game, and while that’s not “luck” exactly, it’s not getting blown away either.  The smaller the margin of your loss, the less you can ascribe every aspect of it to stuff that’s in your control. 

Later, in 2004 and ‘05, we are getting more and more reason to doubt that you can build a dynasty on half the budget of your competitors.  The cupboard is barer and barer, and the A’s are receding from the playoff picture.  If the trend continues, then the dream of building a champion on a shoestring was either beyond Beane’s powers or is inherently extremely unlikely.  To many observers, the failures of the ‘04 and ‘05 A’s confirm their opinion that Beane wasn’t all that good in previous years.  I don’t know if that’s fair or not.

And thankfully someone comes along to perfectly illustrate my point.  Games with time sequences; a few over statements, a few understatements, a few misstatements, and a blurring of the picture.  Then a moving definition to suit the rhetoric.

It’s the same logic that says, there wasn’t any luck to Joe DiMaggio’s hit streak, because the probability of such a streak is so low it couldn’t be due to luck.

Make up your mind what luck means, because you are changing the meaning multiple times to suit your argument.

If you define luck as uncertainty, which is what you say at the end, then there is absolutely no uncertainty in DiMaggio’s hit streak.  It already happened.  The probability that DiMaggio will have a 56 game hit streak is 100%.  You can’t zoom around the event point and make a cogent statement.

If you have a perfect simulator of all events, the likelihood of DiMaggio having that hit streak is still 100%.  If it doesn’t happen that way, it just means you treated something as random, that has already occurred.

Now if you want to start taking away things that are external to Joe D.‘s control, and treating them as random events then you certainly may have a lower expectation that Joe D. will have that hit streak.  If you want to start taking away things were Joe D. has partial control, your percentage drops further.  But now you are starting to introduce error.

That’s pretty much what you’d get if the playoffs were a crapshoot once you got there.

Its also pretty much what you would expect if you analyzed the strengths, skills and abilities of the various teams at a relevant time.

It might be what you expect if you used Santarria too, but I don’t know.

So you just stop there, you are stuck with multiple plausible explanations of the event space.

But why stop there.  Why not use the other information at your disposal and arrive at a more sound conclusion.  Is it now important to ignore those factors because it leads you to a conclusion you don’t like.

That’s precisely what I mean.  You have some gross order conclusion, and you use it rather than a more precise conclusion using more data because it fits your dogma.

As I said, I can do this too.  I can conduct an outcome space that shows that the worst franchise in baseball COULD have won 3 world championships.  Then I could not look at anything else and conclude the Yankees are the worse team in baseball because that was bound to happen.  In fact at some point we expect it to happen.

I can then use my specially derived and biased Steinbrenner stats to show that I predicted the Yankees are the worse team in baseball.  Then just say things are bad luck.

just as the A’s may be slightly more suited to losing a playoff series than other teams with similar regular season winning ability. 

And if you get rid of the unsupported term “slightly” then you have articulated the roll of the GM.

And if you want to use the loaded term “slightly” then you might want to derive the amount of influence the GM has on that outcome.

But the argument is not a probability table from an undergrad math textbook, and it sure as he11 isn’t “he couldn’t expect ...” when most of the time its his job to prepare and “expect ....”, and have an alternative plan.

The point is that, given the amount of teams making the postseason, if it’s a crapshoot, even if they are all equally good, we expect one of them to lose 4 series in a row.

Then that is a dumb point and is just as circular as everything else.  Because you are using that point to prove the playoffs are a crapshoot.

The playoffs are not a crapshoot and teams aren’t equally good.

We expect the bad teams to lose more than the good teams. 

That is what happened.

Normal distributions don’t exist because the world is mostly random.  Normal distributions exist because there is a difference in skills.

That’s like looking at any curve and assuming the bottom portion exists only because of luck, and that intelligence is not a skill, but a random output of test taking.

Its only a small sample because you are constraining the sample to purposefully try to imbibe uncertainty.  How about 0-9 in series deciding games (or whatever the number is).  There is certainly something unique about that space.  Nothing terribly selective about it.  How often do you expect that if all teams are of equal talent?

Not that it really matters, we are on the other side of the time point.  We can look at the A’s critically.  Or we can throw our hands up at every factor and say “That’s luck.”

The playoff results may tell you something about Beane and A’s, but only if you are interested in Beane/Oakland in particular and not anyone else. But if you take the names off the team, it all looks like workings of luck (or as Mark prefers, uncertainty, that’s what is meant by luck in this context). 

The only thing you should be interested in is how teams with specific attributes perform in this environment. 

If I take the names off the teams, it still looks like to me there is someboday kicking ass at 11-4 and somebody stinking up the joint at 0-4.  Because remember, if you are truly looking at the distribution you have to get a permuation of 8 from 30, which I hope you believe is arrived at from skills rather than luck.  You have to have someone that is both selected and loses in all four trials.  So even taken alone, it still doesn’t look like that much of a crapshoot.  It looks like you have teams that are successful and teams that aren’t, and there may be causal basis for those successes and failures.

Didn’t we discuss the Joe D thing a few months ago, and I showed that it wasn’t that out of the realm of possibility?

Even in paradigmatic determinate systems like Newtonian gravity or quantum mechanics, future states of the system can be unpredictable except in probabilities.

Disagree to all your liking, you are flat wrong on this one.  As I said, every one of you is playing games with the time point. 

Your rebuttal talks about future states.  And your point is about past systems.  I have done this 100 times, but I’ll do it again.  I can’t more expressly state that past events have a closed event space.

You can argue that till you are blue in the face.  The probability that the Red Sox win the 2004 World Series Championship is 100%.

And to the argument that you are making, it is a philsophical question that I have avoided.  From a pure analytical standpoint the only thing thereoms such as a Heisenberg Uncertainty principal tell us is that I cannot know certain attributes of a particle because the bounds are near infinite even in small time spaces (and my act of observing is a causal factor)  whether this is an intrinsic characteristic of randomness is totally irrelevant to anything under discussion.

The probability that the Redsox won the 2004 WS is 100%.  The probability that the Redsox won, based completely on observable events, is not 100%. 

If we could observe all electrons, and understand when and how they move, if all of these things were observable, then the Redsox winning the 2004 WS would have been a foregone conclusion, and the probability is 100%.

Didn’t we discuss the Joe D thing a few months ago, and I showed that it wasn’t that out of the realm of possibility?

I must have missed this discussion. If someone has a link I’d appreciate it.

As I said, I can do this too. I can conduct an outcome space that shows that the worst franchise in baseball COULD have won 3 world championships. Then I could not look at anything else and conclude the Yankees are the worse team in baseball because that was bound to happen. In fact at some point we expect it to happen.

I can then use my specially derived and biased Steinbrenner stats to show that I predicted the Yankees are the worse team in baseball. Then just say things are bad luck.

I do NOT think you could do this. I think your “Steinbrenner stats” would demonstrably have so little to do with the actual game of baseball that any reasonable person would reject your argument out of hand. Every stat* thrown around at this site is miles better than whatever you would come up with, in a way that would be clear to the casual observer.

And if you want to use the loaded term “slightly” then you might want to derive the amount of influence the GM has on that outcome.

This is the crux of the disagreement, I believe. I say the amount of influence is, if not slight, then not very large either. You say it is large. I have nothing to back up my opinion except my experience and intuitive understanding of the game, which could certainly be mistaken. But, philosophical arguments about the nature of luck will not convince me that I’m wrong. I would need to see actual evidence.

0-9 in series deciding games

IMO, this is cherrypicked. If someone comes forth with a study about series deciding games, I could change my mind. Needless to say I haven’t seen such a study.

*With the possible exception of Productive Outs.

Didn’t we discuss the Joe D thing a few months ago, and I showed that it wasn’t that out of the realm of possibility?

Yes. I once made an estimate of this on my own and I disagree with Gould.

The probability that the Redsox won the 2004 WS is 100%. The probability that the Redsox won, based completely on observable events, is not 100%.

Exactly. Past events are certain. Future ones never are. In all the discussions here I’ve been treating the matter ex ante. I thought that was clear, but if not it should be now.

To bring my point full circle, it’s far too easy to say that there must be some X factor which (fill in your favorite underperforming manager) didn’t understand. Recognizing unpredictability takes courage.

No, “recognizing unpredictability” in any space that are dealing with is pretty much a cop out.  What you determine is if there are things that are in your control that must be improved. 

It is a perfectly reasonable conclusion to determine its not feasible to exercise control over an element.

It is a mere copout to look at an output and say that’s luck.

The probability that the Redsox won the 2004 WS is 100%. The probability that the Redsox won, based completely on observable events, is not 100%.

No, the probability they won is 100%.  The amount of observable characteristics is incomplete.  That is an instrumentation problem. It does not effect the probability of a past event.

And I’ll state this one more time, these distinctions might not be that important, but it has become important with the way that “luck” has this little catchall amorphous characteristic in the saber world.

You have posters here that will use it three different ways even after expressly defining it in their post.

Sure, I can say, Joe is lucky that I didn’t get a DUI after driving home with 12 beers.  If you want, you can say Jim is unlucky for getting stopped and cited after 1.5 beers.  That is fine.  Where the cops are posted are external to your control.  How they will react once you are stopped is only partially within your control.  But none of these things imply that the decision to drive after imbibing alcohol is a crapshoot.  Its still a bad decision, it still influences the outcome of you getting a DUI.

You don’t just say, “Oh well, that is luck” because 2 years ago you determined that at 3 AM any person is likely to get stopped 3% of the time.  You might determine that alcohol can influence your driving to the level that a stop is more probable and its thus an improper decision.

I should stay out of this, but I want to give a concrete example to see Backlasher’s reaction.  In Game 5 of the 1969 WS, the Orioles led the Mets 3-2 going to the bottom of the 7th inning.  Al Weis—he of the 7 career homers in 800 career games, .219 career average, and .275 SLG, hit a game-tying home run off Dave McNally, one of the best pitchers of his time (in the midst of a run of four straight 20-win seasons).

We’ll never know if the Mets would have won that game w/o Weis’s home run, but it forced Weaver to PH for McNally in the top of the 8th, and the Mets scored twice in the bottom half to win the game and the WS.

From the POV of Frank Cashen and Earl Weaver, doesn’t that have to be seen as “bad luck”?  They had built a team, and had a match-up, that should have resulted in a series-losing home run about as often as I would hook up with Cher.  What more could they possibly have done to create a favorable situation, strong odds of a successful outcome?

My point?  There sometimes IS bad “luck”—an outcome that no one could expect a GM or manager to foresee even with the best, most insightful decision-making.  It wasn’t bad luck from McNally’s POV, or good luck from Weis’s.  One performed, the other didn’t, and that’s not luck.  But it surely was bad luck for Earl Weaver, and incredibly good for Gil Hodges, that it was Hodges’ player who performed and Weaver’s who did not.

No, the probability they won is 100%. The amount of observable characteristics is incomplete. That is an instrumentation problem. It does not effect the probability of a past event.

That’s what Einstein said about quantum mechanics. He was wrong.

No, “recognizing unpredictability” in any space that are dealing with is pretty much a cop out. What you determine is if there are things that are in your control that must be improved.

It is a mere copout to look at an output and say that’s luck.

Sure, but my point is that it is just as much a copout to claim an unknown causal factor without proving it.

Sam,

  Check out #51 for my response.  Of course there is bad luck.  Any outcome you don’t like is due to bad luck as long as it is not a completely determinable outcome.  Also, you might factor in your level of expectation.

And specifically, I don’t know what the Earl could have done.  It sounds like optimal decisions were made.  I’d have to do some research, to see if there was something that could have been done, And I’ll state this one more time, these distinctions might not be that important, but it has become important with the way that “luck” has this little catchall amorphous characteristic in the saber world.

You have posters here that will use it three different ways even after expressly defining it in their post.

Sure, I can say, Joe is lucky that I didn’t get a DUI after driving home with 12 beers.  If you want, you can say Jim is unlucky for getting stopped and cited after 1.5 beers.  That is fine.  Where the cops are posted are external to your control.  How they will react once you are stopped is only partially within your control.  But none of these things imply that the decision to drive after imbibing alcohol is a crapshoot.  Its still a bad decision, it still influences the outcome of you getting a DUI.

You don’t just say, “Oh well, that is luck” because 2 years ago you determined that at 3 AM any person is likely to get stopped 3% of the time.  You might determine that alcohol can influence your driving to the level that a stop is more probable and its thus an improper decision..  Just based on the facts, the first place I’d check is to see how the bullpen was constructed.  It sounds as if the Mets found a way to exploit a weakness, and I do know Earl had the propensity to use non-optimal bullpen utilization.

Depending on what I found, I may set some watchpoints for future inquiry.  And if that event or other events that could be caused by perceivable deficiencies manifested themselves over a number of trials, I would begin to start looking really hard at the situation.

Sure, but my point is that it is just as much a copout to claim an unknown causal factor without proving it.

What are you talking about Field.  If you see a fire, do you need to prove how it started before you go put the damn thing out.

The outcomes occurred.  They are suboptimal.  You can investigate cause, you can put the damn thing out, or you can say “we don’t need no water let the lucky muther fukkkker burn.”

You want to indict because somebody yells fire at a fire and begins to investigate or asks for a bucket of water.

Just to clarify my statement about Einstein, I am again referring to the ex ante prediction. Of course, after the fact, we know the event happened.

I thought the whole point of this discussion was to (1) identify currently unknown causal factors, or (2) admit that a system is unpredictable at present. Otherwise we’re just monday morning quarterbacks.

That’s what Einstein said about quantum mechanics. He was wrong.

To my knowledge, Einstein never spoke on this.  If he said that the probability of past events is discrete at either 0 or 1, he was right.  I don’t know who is telling you different, but they are wrong.

Sure, but my point is that it is just as much a copout to claim an unknown causal factor without proving it.

BL: I agree w/Mefisto on this, if I understand the point right. His point is NOT that there may be some currently unknown causal factor, but that the cop out arises in simply adverting to the existence of some causal factor you don’t then strive to show. Is that right, Mark?

If you see a fire, do you need to prove how it started before you go put the damn thing out.

The outcomes occurred. They are suboptimal. You can investigate cause, you can put the damn thing out, or you can say “we don’t need no water let the lucky muther fukkkker burn.”

If this is an analogy, I don’t get it. Putting out the fire and determining its cause are two separate and distinct tasks.

After the fact we know there was a fire. Nobody ever has or ever will dispute this. The dispute is over the issue “could/should we have predicted this particular fire in advance?”. If someone believes that we should have, he must show how and why. Otherwise we remain justified in believing we can’t predict when and where particular fires will occur even though we know they will.

But none of these things imply that the decision to drive after imbibing alcohol is a crapshoot. Its still a bad decision, it still influences the outcome of you getting a DUI.

Agreed.  So you want to treat the presence/absence of external factors influencing the outcome as the marker of whether it was “luck.”  That is, if there were external factors (e.g., where were the police stationed?) that might be lucky (good or bad) for the two drivers.  Then, however, you point out that there is still an important element within the control of the individual (she chose to imbibe), so it’s not ALL luck.  There was “a bad decison, [and] it still influences the outcome of you getting a DUI.”

But I don’t think there will always be such a “bad decision.”  Sometimes, you get an unfavorable outcome even without anything we can identify as a “bad decision.”  No bad decision by the GM, no bad decision by the manager, no bad decision by the pitching coach, etc.

It also makes a big difference what question we’re asking.  If the question is, “Were the Orioles unlucky?” I would say no—McNally’s performance is part of the equation, and that wasn’t really luck.  Him throwing a pitch Al Weis could hit out of the park was the equivalent of choosing to imbibe.  But if the question is, “Was Earl Weaver unlucky?” I think the answer is yes.

Mark: About Einstein, to what book or essay or whatever are you referring to?

His point is NOT that there may be some currently unknown causal factor, but that the cop out arises in simply adverting to the existence of some causal factor you don’t then strive to show. Is that right, Mark?

Exactly. That’s twice in recent weeks, JC.

One of us is improving. It’s probably best to let someone else decide who.

Mark: About Einstein, to what book or essay or whatever are you referring to?

I’ve read it many times. I can’t access my books here at the office, but a quick search came up with this unimpeachable site. I’ll quote the key passage:

“Einstein was very unhappy about this apparent randomness in nature. His views were summed up in his famous phrase, ‘God does not play dice’. He seemed to have felt that the uncertainty was only provisional: but that there was an underlying reality, in which particles would have well defined positions and speeds, and would evolve according to deterministic laws, in the spirit of Laplace. This reality might be known to God, but the quantum nature of light would prevent us seeing it, except through a glass darkly.

Einstein’s view was what would now be called, a hidden variable theory. Hidden variable theories might seem to be the most obvious way to incorporate the Uncertainty Principle into physics. They form the basis of the mental picture of the universe, held by many scientists, and almost all philosophers of science. But these hidden variable theories are wrong. The British physicist, John Bell, who died recently, devised an experimental test that would distinguish hidden variable theories. When the experiment was carried out carefully, the results were inconsistent with hidden variables. Thus it seems that even God is bound by the Uncertainty Principle, and can not know both the position, and the speed, of a particle. So God does play dice with the universe. All the evidence points to him being an inveterate gambler, who throws the dice on every possible occasion.”

To a first approximation you can take the regular season records of the teams.

That’s not what I’d call a first approximation.  And as such its only good if the bounds of the output are reasonably well set.  If its Pedro Martinez versus Jose Lima, then it pretty much sucks.

I say the amount of influence is, if not slight, then not very large either. You say it is large. I have nothing to back up my opinion except my experience and intuitive understanding of the game, which could certainly be mistaken. But, philosophical arguments about the nature of luck will not convince me that I’m wrong. I would need to see actual evidence.

No, the crux of the problem is that you don’t understand I’m not trying to convice you of anything.  I could care less.  If you want to dance around the time spectrum and make excuses then go right ahead.

The issue is very high level and very easily stated.  Many of these fanboys and would be statisticians play games.  They remove information, they introduce error, and then at all signs of evidence that shows they are wrong—- They cry luck.  That is the fukking issue.  Its the issue when I started.  Its the issue now.

But when you say something like this, you will get people like Treder saying, “They don’t recognize how much luck influences our lives, bar bar bar.”  I understand completely.  Its just every time, you throw out the big point, you got a bunch of folks taking it down to quantum mechanics to try to show you are wrong and that you don’t understand.  I’m happy to show there is no mistake except on their assimilation of this information.  I’m not taking this to a metaphysical level; its not necessary, but I am correcting misstatements when its used as an affirmative argument about these events.

I don’t really care how much influence you think any set of persons have.  What I care about is maximizing opportunity using reasonable constraints.  I’ll talk about when people don’t do it.  I’ll investigate it more, and I’ll fire back at those saying its just luck, because its convenient for them to make that conclusion.

Every stat* thrown around at this site is miles better than whatever you would come up with, in a way that would be clear to the casual observer.

Really.  You mean even DIPS.  LOL.  How about the sophmore probability table to show that the playoffs are a crap shoot.

And that passage does not refute the fact that a past even either occurred or did not occur.

I don’t need to read your cite.  That is a true statement no matter who says otherwise.

That’s what Einstein said about quantum mechanics. He was wrong.

If you are refering to the Einstein Bohr debates (1927, 1930), you might want to review the first chapter of _Collective Electrodynamics_ by Carver Mead.

BL: I agree w/Mefisto on this, if I understand the point right. His point is NOT that there may be some currently unknown causal factor, but that the cop out arises in simply adverting to the existence of some causal factor you don’t then strive to show. Is that right, Mark?

And I have addressed this.  That is the problem.  Every attempt to investigate a casual factor meets the rebuttal, “But that is luck” implying that luck itself is a casual factor.

That is the problem.  That isn’t a hidden variable theory and Bell does not read on this situation.  Luck is used for sophistic purposes.

There is no null hypothesis that must be excepted until debunked.  There is an outcome.  It is suboptimal.  YOu may discover that one actor could not do anything within their control to influence the result, but that must also be proven just as with any other outcome.  And that is not “luck” or if it is “luck” it is not probative.  But most important, its not a license to generate an excuse machine.

And sophmore probability tables don’t change that.  And the ontology of a particles observable and randomized nature is not probative to any of these analytical points.

And that passage does not refute the fact that a past even either occurred or did not occur.

For at least the third time, nobody ever has or ever will disagree with this. My disagreement is not with your statement—it’s true—but with its relevance. What I’m talking about is the ability to predict in advance.

Every attempt to investigate a casual factor meets the rebuttal, “But that is luck” implying that luck itself is a casual factor.

I agree that it’s wrong to discourage investigation into possible causes. But the person who suggests possible causes must do more than simply throw out some hypotheticals. He must make some sort of evidentiary showing.

If you are refering to the Einstein Bohr debates (1927, 1930), you might want to review the first chapter of _Collective Electrodynamics_ by Carver Mead.

I’ve never read the debates themselves, only others’ interpretations of them. Does Mead think the standard view of Einstein’s position (which Hawking’s quote expresses) is incorrect?

Normal distributions don’t exist because the world is mostly random. Normal distributions exist because there is a difference in skills.

Nicely stated, Backlasher.

This article is about the first approximation. Contrary to what BL says, I think the first approximation is extremely important. So is the second approximation. That doesn’t mean a GM should pay no attention to the other factors. He should. But until I am convinced otherwise, I will continue to believe that the cumulative effect of the other factors is relatively small in most cases, and therefore that first or second approximation calculations have a good deal of merit.

Perhaps it’s semantics, but I disagree here.  I don’t know precisely what you mean by the “second approximation” but fine-tuning one’s estimates of the true winning percentages of teams based on personnel changes, pitching matchups, and the like, can actually have a pretty huge impact on our expectations.

For example, consider the 2003 NLDS between the Cubs & Braves (sorry, BL, but it’s a nice example of my point).  The Braves won 101 games and the Cubs won 88.  Given that, I get a “first approximation” that the Braves would win a best-of-5 series of 65.1% (using the Log5 matchup formula, ignoring home-field advantage).

But, the Cubs that faced the Braves had two huge advantages over the 88-win Cubs—they didn’t start Shawn Estes or Juan Cruz and they got to play Kenny Lofton and Aramis Ramirez every game.

When Wood, Prior, Clement, or Zambrano started, the Cubs won 57.9% of the time (vs. 54.3% overall).  Before the Cubs acquired Lofton and Ramirez they were a 0.500 team (literally, they were 50-50).  After they acquired them, they won 61.3% of their games.  Combining these two things, we’d expect a Cubs team with just W/P/C/Z starting and with Lofton/Ramirez to win 64.7% of the time (that’s 105 wins out of 162).

So, making these adjustments take the Cubs from winning 35% of the time to winning 55% of the time.

I’m not saying that what I did is right or is where you’d stop or that you couldn’t or shouldn’t do the same sorts of things with respect to the Braves too.  My point is simply that, in my opinion, if making this type of adjustment can swing the odds of winning a series by 20%, then I’d simply question the merit of such first-order approximations.

I haven’t looked at the debates either (nor the sources you reference; my studies in physics haven’t taken me in the direction of MC Hawking).

But I’m completely confident in my answer to your question: “I’m not sure.”  I believe the answer to your question is “yes, Mead believes that the standard view of Einstein’s position disagrees with Einstein’s actual view.”

“... working-level physicists use a totally distinct interpretation to analyze collective systems: The coherent state is taken as the starting point, and the statistics are put where Einstein would have placed them—in the random thermal excitations outof the perfectly coherent state.  The success of this method can be viewed as a vindication of Einstein’s conjecture.”

But I may be wrong that this is related to the Hawking quote.

In Soviet Russia, craps shoot YOU!

One of us is improving.

Either that or it’s just random.

Look at it this way:
How much certainty can anyone here predict the outcome of a playoff series?

About 50-50.  It’s a bit easier with the Cards-Pads because hte Pads are weak for a playoff team, but all other series are a 50-50 propsect (about).

The way scales tip in a 50-50 prospect is random/luck/mistake.

I appreciate BL’s brine tanks, but due to the streak nature of baseball, it isn’t possible to account for what is going to happen in 5 baseball games (that is a short series).

Thus, as JC rightly says, you decrease the impact “luck” can have on your team, and then *roll the dice*.

I’ll have to take a look at Mead’s book. Here are two examples, 25 years apart, consistent with Hawking:

Heinz Pagels, The Cosmic Code, p. 61
Brian Greene, The Elegant Universe, pp. 105-108.

But I don’t think there will always be such a “bad decision.” Sometimes, you get an unfavorable outcome even without anything we can identify as a “bad decision.”

A wise man once said, “You can do everything exactly right and still lose.” 

In Sam’s example, I would say that the Orioles were unlucky - in that specific event.  No way should McNally have been expected to give up that home run and if they faced each other 100 more times, he probably wouldn’t.  That is safely an example of what I think of as luck.  Something very unlikely happens. 

However, this does not mean the Orioles were unlucky to lose the series.  They still had many opportunities to not only win the series but win that game.  To win, you need to be ready to deal with such an unlikely and unfortunate event. 

Thus, as JC rightly says, you decrease the impact “luck” can have on your team, and then *roll the dice*.

And this really is the point.  You decrease the effects of unknowns as much as possible.  I think it is completely plausible, even likely, that some people do this better than others.  That is, that there is a skill to the postseason.

is completely plausible, even likely, that some people do this better than others. That is, that there is a skill to the postseason.

I definitely agree that some people do it better than others.

However, the skill in the postseason, IMO, due to the skill level required to get to the playoffs, is overturned by luck. 

Or a bad play at an inopportune moment (Graffanino’s error).

Word Count:

Original Article: 1,109
Response of my stalker: 5,270

This might help:

http://www.lights.com/publisher/

The probability that the Redsox won the 2004 WS is 100%. The probability that the Redsox won, based completely on observable events, is not 100%.

I observed their victories in games 1, 2, 3, and 4. It’s still 100%.

I’m sure that’s not what you meant, though. Let me take a stab at it anyway.

Given:
1. what we had observed at some time prior to the World Series - maybe the end of the LCS, maybe on June 27, 2004, maybe at some point in 1952;

2. what the possible outcomes were from those observed conditions;

3. what degree of influence UNobserved events would have on the possible outcomes;

4. the extent to which the possible outcomes could be misestimated because we’re not as good at 1-3 as we need to be;

the estimable (at that point in time) probability of the Red Sox winning the World Series was < 100%.

IOW, when you say “observable”, I think you mean “observed”. Anything is observable. Few things had been observed.

I think Backlasher’s points are:

(a) The actual probability is still 100%, given that it did happen;

(b) We’re so horrible at 3 and 4 - or, worse, that we dismiss 3 and 4 as “luck” or insignificant when we have no basis in fact to make that judgment - that anyone toting their analysis as definitive is wrong.

And I’d agree, to an extent. (Naturally, I can’t completely agree, or else it wouldn’t be a traditional strawman.)

Everything is on a relative basis. Making predictions of playoff win probability based on win %, a basic formula, and a basic adjustment for home-field advantage, is more useful than some things, less useful than others. It’s more useful than a different formula - say, one that says the team with the worse record wins all the time. But it’s likely less useful than a prediction based on the roster construction of each team. That method, in turn, is likely less useful than one based additionally on the umpiring crew working the game. And that, in turn, is likely less useful than one based additionally on the current health condition of everyone involved.

Each of these things is observable and measureable; and each could improve the accuracy of the prediction. The improvement might be slight; it might be huge. But the fact is that we don’t know.

That doesn’t render the prediction useless. It just means it’s not definitive. It also means that we really don’t know how good it is in absolute terms. In relative terms we can make that judgment, but in absolute terms we can’t.

Going back to the original article, outlining the cursory review, I find this:

So we geeks <u>are essentially right</u>. (And don’t we just love to announce it to the world when we are). The playoffs <u>are, to a certain extent, a crapshoot</u>. Baseball’s current playoff format <u>ensures</u> that even the very best teams are likely underdogs to win the World Series once they make the playoffs.

While Voros does emphasize that this is cursory, and you shouldn’t bet anything on it, such emphatic language as underlined above does not belong in a cursory review. And such emphatic language often follows the analyses that appear on this and other sites. And it often accompanies a conclusion about what is or isn’t “luck”.

There is no luck. There are unobserved causal events. If you “observe” luck, it’s because you failed to observe the causal events. Wild deviations from your expectations have more to do with inadequate expectations than anything else.

Except Aaron Small, of course. That’s just luck.

DiMaggio’s streak is the most extraordinary thing that ever happened in American sports.

“Baseball isn’t statistics, it’s Joe DiMaggio rounding second base.” -Jimmy Breslin.

After the fact we know there was a fire. Nobody ever has or ever will dispute this.

You pay my retainer, and I will dispute this to the last penny.

“While Voros does emphasize that this is cursory, and you shouldn’t bet anything on it, such emphatic language as underlined above does not belong in a cursory review.”

Since when exactly does “to a certain extent” qualify as emphatic language?

You pay my retainer, and I will dispute this to the last penny.

That sounds about right.

Yeh, V.

THIS NMR IS CONSISTENT WITH THE SPECTRA WE"D EXPECT FOR THAT MOLECULE!!!

Now, I’m not saying the species we have isolated *is* this structure, but the NMR data is consistent with the data from NMR analysis of that species.

I can’t be any clearer than that.

This seems an opportune time to remind everyone about a study carried out by Vinay about 18 months ago. There do seem to be characteristics shared by teams that are successful in the playoffs. Bad news for statheads — it’s all about pitching. Well not quite all. Good fielding is important, plus speed and adventurous (but not necessarily successful) base running. The only hitting skill that shows up as important in the playoffs is to not strike out.

While Voros does emphasize that this is cursory, and you shouldn’t bet anything on it

Well, if you were going to bet on Vinay’s information, you’d get the Cardinals, Astros, Angels and White Sox to win. Any three of four should be enough to win money. Looks pretty good at the moment.

um, normal distributions would still exist even if there were absolutely zero difference in talent.

um, normal distributions would still exist even if there were absolutely zero difference in talent.

Correct.

The amount of disposable income, and the country’s attitude toward entertainment dollars have moved so far from where it was in the 1960s and before, would allow baseball to thrive with just the league setups - I think.

I would argue the opposite. Anything that would cause a dramatic increase in fans of good or very good teams who’ve given up and are waiting for NFL/NBA/NHL to start is a bad thing.

OTOH, yes, if you have more games to show on tv, then you make more money.

If the *only* problem with fewer playoffs would be fewer games to show, they could just make the schedule 166, 168, or 170 games.

Response of my stalker

I think your stalker has been two timing you, dude.

This seems an opportune time to remind everyone about a study carried out by Vinay about 18 months ago.

A nice start, but there were sample size issues. I still think “errors committed” is a nice dummy variable.

Here are the stats for all post-season series 1969-2004 except for the World Series

0-4 1/2 game difference   16-31
5-9 1/2             17-12
10-14 1/2             11-4
15+                11-2 (Mets over Reds in 73 and Yanks over Mariners in 01)

World Series 1903-2004
0-4 1/2             17-18
5-9 1/2             14-21
10-14 1/2             13-7 (but the last four the best record lost)
15+                16-3 (The most famous being the Sox over Cubs in 06)

Unless you’re ten games better it’s pretty much a crapshoot and the team that’s 5-9 1/2 better in the same league is 60% likely to win. (Small sample size however) The 14-21 spread in WS games might be due to the strength of the leagues at the time. Teams that won the year before tend to beat a team with a significantly better record. (Yanks over Met’s and Mariners in 00 and 01. Oakland in 74)

Isn’t it the case that—in Best of 7 Series—teams that lose the first three roll over in game 4 by a disproportionately large margin (last years ALCS notwithstanding).  Not just that the 0-3 hadn’t won a series until last year, but that they had only rarely extended the series to 5 or 6 games (and never to 7).

This RSB thread includes the results I got reviewing best of 7 world series.  Down 3-0 was very different from down other ways.

(Only world series were included in my numbers because I wanted to avoid any interference that might be caused by “setting up the rotation for the next series”); but baseball-reference has all the data you need to reproduce my results and expand them to include division play.

Kiko (#72): My “second approximation” was exactly the type of roster composition issues you are talking about. Sorry for not being more clear.

BL (#65): Suppose I say, “After accounting for roster composition, pitching matchups, and home field advantage, everything else is luck.” That is wrong. Now suppose I say, “After accounting for those things, most of the rest is luck.” I have no chance of proving that statement. The most I can do is show some sort of statistical equivalence (a la Dial in #86 if I understand him). As well, a GM shouldn’t take this approach. He should do everything possible to maximize his advantage, like JC says. But the statement might still be true, and I think it is true. (As for “dancing around the time spectrum,” I’m looking at things from the POV of right before the series begins.)

Now for this:

Every stat* thrown around at this site is miles better than whatever you would come up with…
Really. You mean even DIPS. LOL.

Yes. I’m making the absurd claim that DIPS is better than a stat showing the 1996-2000 Yankees to have been the worst team in baseball. LOL indeed.

OK, Backlasher, I get it.  Voros et al are using tools from Stats 101 or Stats 102 but that isn’t The Final Word on analyzing baseball.  But you portray these as having some sort of diabolical agenda.  Going by you, Steve Treder is the Timothy Leary of sabermetrics.  Heh, maybe he is, but I don’t think these guys have malicious intentions.

I’m guessing that this criticism by Voros was directed towards you: (please remember that word before someone starts chiming in with stuff you need a PHD just to get the introductory textbook to)  I’m not sure if this is a fair criticism now, but I think that it was at one point.  I actually like to consider myself a fairly smart guy, although the last formal class that I took in stats and probability was at the 101 or 102 level and I went to some podunk backwater of a college.  It took me a few months of reading your posts to figure out what the heck you were talking about with regards to some topics.  Part of it may have been your writing style, I dunno.  It can still be opaque and occasionally condescending.  (Not that you’re the only one here like that.)  I still don’t grasp everything that you’re talking about, but I have a better idea.

Incidentally, have you ever read anything by Nassim Nicholas Taleb?

BTW, I enjoy craps, although I’m too risk averse to do it more often than a couple of times a year.

Okay, I’m going to take another stab at my explanation of #37.  You can’t base and prove your argument about a random process on the same piece of evidence.

If we say that the outcome of each postseason series is a random binomial event of approximately p = 0.5 (say 0.4 to 0.6), then we would expect an outcome similar to the table I presented in #37.  Sometimes the A’s would come out on the bottom, sometimes the Twins, sometimes the Yankees or Cardinals or whoever ...

The way it actually turned out, the A’s lost 4 series and won none.  You can’t take the fact that they lost to identify them for the analysis, and then use that SAME fact to show how improbable it is that they lost unless you hypothesize this about the A’s in particular before any of the games are played.  Otherwise, you’re just taking an event that was extremely likely to have occurred (any team would lose 4 series) and then arguing that for that particular team it is unlikely.  Sure it is, but not when you chose that particularly team for analysis after knowing that it had lost.

The fighting over whether the Red Sox are a “saber” team or not is a remnant of that. BL et al want the Red Sox to be a non-saber team, because if they are a saber team their whole theory about the A’s sucking (which is a fine conjecture, but not supported or disproved by the playoff results because of the selection process I detail above) is a bunch of hooey.  The other side—with less strong personalities, so I don’t know who’s leadign the charge—wants to Red Sox to be a saber team because that would prove conclusively that BL’s theory is hooey.

In both cases, this is just Mendel classifying his peas.  BL will classify a winning team as not-saber, because he knows that saber teams don’t win, while the others will classify a winning team as saber, because the know the saber teams should win.  What’s funny is that neither group seems to realize—or at least, they fail to admit—that they are doing the exact kind of insincere analysis that they accuse the other of.

There is no luck.

While I think BL and others make some excellent points, this sentiment astounds me. 

In both cases, this is just Mendel classifying his peas. BL will classify a winning team as not-saber, because he knows that saber teams don’t win, while the others will classify a winning team as saber, because the know the saber teams should win. What’s funny is that neither group seems to realize—or at least, they fail to admit—that they are doing the exact kind of insincere analysis that they accuse the other of.

True enough.  What is most important, probably, is how good of a team you put out there and how well they play.  That is, talent and skill matter.  I could manage 25 primates with exactly the same organizational philosophy as Bobby Cox and I’d be unlikely to win a single major league game.  That is, the philosophy of the team isn’t singularly important.  If we have 5 saber teams, it isn’t surprising that we’d have the A’s (who, I’d argue were essentially saber before Beane arrived), the Red Sox, the Blue Jays and the Dodgers - great teams, good teams, bad teams and horrendus teams that happen to share a philosophy.

Now, I’m not saying the species we have isolated *is* this structure, but the NMR data is consistent with the data from NMR analysis of that species.

You need to take a COSY, dude.