Devil,

Here's where I think we disagree. Theoretically, I agree with you. You should be able to choose any player in the major leagues, and stick him at DH.

Somehow, I just don't think it works that way. Perhaps its because players don't want to DH. Perhaps their production suffers when they DH. I'm bothered by the fact that so many teams seem unable to get decent production from their DH's, year after year.

That said, I'm not disagreeing that a thirdbaseman has value over a DH. I just don't think that starting the season knowing that David Ortiz (or Edgar Martinez a few years ago) is as much of a disadvantage as some make it out to be.

That's why I still think (though I'm open to persuasion) that all full-time DHs start at a pretty severe handicap in terms of their value to a team, but once that's said, the quality of the other guys on their club isn't an issue.

I completely agree.

All you need to do then, at least for an estimate, is get an average of the worst regular defensive player on each team. My guess is you're looking at between -15 and -20. Dock that from the DH, and ignore the more specific context of teammates.

Sam,

I don't know if you have been following the situation closely, but there are two reasons the Red Sox waived Manny Ramirez: his contract and his attitude. They would have been less likely to waive him if he was a gold glove left fielder, but that has nothing to do with the fact that the Sox have Ortiz.

If you follow a "chaining" approach (think of hockey and what happens when you lose one of your 6 starting defenseman, and the 7th one comes in), the next effect on the fielding side is a loss of about 20-25 runs (2 wins) per 162 games.

Essentially, a poor fielder will make about .04 plays less than an average fielder at a particular position. For a low-leverage position (1B or LF with about 3 opps per game), that comes out to about 19 plays, or 15 runs. However, the average 1B is himself about 9 runs worse than the average fielder, making a poor fielder worth 24 runs relative to the average fielder. Call it 2 wins.

So, that's the penalty I've always used: whatever a DH contributes, subtract 2 wins per 162 GP. That's the effect *to the team* of having the fielding alignment they do have.

Dizzy -

For what it's worth, I think that the Red Sox would never have waived Manny if he played a GG left field.

If he weren't a -20 defender, his contract would be pretty good. The whole problem with Manny stems from his not giving enough value for the money. The value problem is intimately related to his defense, which is terrible.

(I'll add that I think the waiver game was a very dangerous one, and it's quite telling that the Sox haven't actually pulled the trigger. Replacing Manny with Vlad Guerrero would be good, but you better be damn certain you can get Vlad Guerrero because otherwise you'll be stuck with a worse ballclub. Cheaper, yes, but worse.)

My first two paragraphs are two different views to looking at the issue. Both result in essentially the same result.

That said, I'm not disagreeing that a thirdbaseman has value over a DH. I just don't think that starting the season knowing that David Ortiz (or Edgar Martinez a few years ago) is as much of a disadvantage as some make it out to be.

Fair enough. I think the way Tango Tiger does it is probably the most fair.

As for why it's harder to find a good DH than it theoretically should be? I would guess a lot of it is that, like you said, a lot of guys don't want to DH. Some guys, like Jason Giambi for instance, also happen to hit much better when they play in the field, as others have mentioned. I also think a lot of teams like to DH players that are injured but still healthy enough to swing the bat as well as older, declining players.

whatever a DH contributes, subtract 2 wins per 162 GP.

If you trust Rhoids' Game State Wins, which are quite inaccurate but at least take clutchness into account, this puts Ortiz and A-Rod into a dead heat before considering A-Rod's defense. So if A-Rod were a below-average defensive 3B, that would put Ortiz ahead.

Of course, if the race is that tight, you have to take a closer look at those Rhoids numbers. But this is no slam-dunk either way.

You just don't know how good a player Alex Rodriguez has been this season unless you see him play everyday.

If you trust Rhoids' Game State Wins, which are quite inaccurate but at least take clutchness into account, this puts Ortiz and A-Rod into a dead heat before considering A-Rod's defense. So if A-Rod were a below-average defensive 3B, that would put Ortiz ahead.

No, they only take hitting into account. There's no positional adjustment at all (AFAICT).

I don't think that Rhoids site is very trustworthy at all. Jorge Sosa and Huston Street are the pitchers of the year? Rich Harden as the best starter in the AL?

If ARod were a below average fielding 3B, he'd probably be an average fielder at a neutral position.

The way I see it, I keep hitting and fielding separate. Compare ARod and Ortiz to an average hitter, and compare them to an average fielder at a neutral position.

Ortiz's fielding is, by default, -2 wins. ARod, according to the fans, has been very good, would come in at +1 wins for his fielding.

You certainly need to look at clutchness for an MVP award. Typically, the gain is at the +/- 1 win level. So, a really good clutch hitter would gain an extra win on that score, and a poor one loses 1 win.

Bottom of the 9th a home team has something like .55% chance of winning

Where'd you get this? Also, he shouldn't get all credit for the walk-off homeruns. Many times guys will be on base, and if they weren't it wouldn't be a walk off homerun. Anyways, I think your rough and dirty way to calculate the value of a walkoff HR could be way off.

I'll add that I think the waiver game was a very dangerous one, and it's quite telling that the Sox haven't actually pulled the trigger.

Well, I'm not saying you should believe every media report, but FWIW, there were several stories this past off-season that the Sox thought they had agreed to a deal with the Mets to trade Manny for Floyd, only to have Omar Minaya back out because he "misunderstood" what they had agreed to re. how much money the Sox would send to the Mets. So, at least from their POV, it appears the Sox were willing to, and thought they had, pulled the trigger on a deal. At the time, according to the reports, Theo was plenty mad at Omar over the whole thing.

LOL, I can see what Mr. Jenkins is talking about when I read this thread.

Matt, I'll offer some advice. There are at least two people out here that its not worth your time to talk too at all.

If someone is really dim enough to state "a run is a run is a run" and thus believe the same value occurs in a home run in a 14-0 game in the ninth inning as a 4-4 game in the ninth inning there is just no use in trying to explain it any further.

Moreover, you have expressly explained that in that silly hypothetical, Manny Ramirez has more value than Ortiz (or in another form the Change in Win Exp. of Manny HR1 + Change in Win Exp. of Manny HR2 Change in Win Exp. of Manny HR3 > Change in Win Exp. of Ortiz HR; and Change in Win Exp. of Ortiz HR > max (Change in Win Exp. of Manny HR1, Change in Win Exp. of Manny HR2, Change in Win Exp. of Manny HR3), where Win Expectancy is mean adjusted WE independent of matchup context)

They are just going to keep telling you "I think ..."

Now I don't know if I agree that a walkoff values at its mean three times a non walkoff. I'd think there are context value adjustments to make, and Grand Slams have more value than Solo Shots because of the booked value of runs compared to expected runs, etc.

But you are making the right points in a sea of insults, so I'll tip my cap.

I agree with everything Tango says in 17 (as though I'm stupid enough to disagree with Tango) except for (proving that I am that stupid):

You certainly need to look at clutchness for an MVP award. Typically, the gain is at the +/- 1 win level. So, a really good clutch hitter would gain an extra win on that score, and a poor one loses 1 win.

I think our definitions of "clutch" are so nebulous that including them in discussions like this is often troublesome. But it is something -- along with the performance of the team -- that I think should be considered when the race was very close, as long as it isn't considered glibly.

The AL race last year was one of the closest in recent memory, in my mind, with Vlad, Tejada, and Santana all being around in the +50 to +55 range in runs against average. For me, the fact that Vlad turned it on down the stretch and dragged his team into the playoffs was a tiebreaker -- that and the fact that I am way biased.

No, they only take hitting into account. There's no positional adjustment at all (AFAICT).

I should have been more clear. They have Ortiz leading A-Rod by 2.1 wins as far as hitting goes. If Ortiz is -2 wins on fielding, then A-Rod's fielding provides the margin.

Jorge Sosa and Huston Street are the pitchers of the year? Rich Harden as the best starter in the AL?

Go to the data dump and sort it by Game State Wins. That gives a more reasonable ranking. (AL pitchers: Buehrle, Santana; NL pitchers: Carpenter, Clemens). On the other hand, their 3rd best AL pitcher is Zito, and there are plenty of other weird results.

Back to A-Rod and Ortiz: it is possible that while the usual spread of clutch hitting is +/- 1 win, Ortiz has been so super-clutch as to add 2 wins. Or maybe not -- I'll defer to Tango's expertise, but I wouldn't reject the possibility out of hand.

I think our definitions of "clutch" are so nebulous that including them in discussions like this is often troublesome.

I'm thinking of "clutch" as something like win expectancy added above replacement minus VORP/10, or whatever. You get the idea. Of course, BL's point stands that all the WE figures are with reference to an average context. A walkoff HR against Rivera should be worth more than a walkoff HR against Foulke (2005 edition).

One of my big pet peeves with ARod is that 'one out, man on third' situations seem to be kryptonite to him - he has to lead the league in strikeouts in that situation.

In a way I think Ortiz's candidacy would be stronger if the Yankees were closer in the standings. Right now the lead seems fairly comfortable, reducing the perceived urgency in Boston's recent games.

Hawk, I don't think we disagree anywhere. I also take "clutch" to simply be:
(change in win expectancy)
minus
(linear weights)/11

Tango, I've already measured Ortiz' change in Run Expectancy at 16 runs. When you look at his extraordinary splits by game score differential, especially compared to ARod's (which may be more typical of a great hitter), I think his clutch value has to be in the neighborhood of 3 wins.

Here are some more numbers (somebody tell me the equivalent of "code" on this board so I can format it correctly):

By the way, those numbers are breakdowns by final game score, not by score at the time. Which is what you want to measure. A 2-run homer in the first inning of a game won by 11 runs turns out to be low-leverage, because the opposition pitching staff turned out to be awful that day. It shouldn't be lumped together with a 2-run walk-off homer.

If these numbers represent something real, it's differing ability to hit good pitching, and different degree to which crap pitching is feasted on.

Whoa. If I'm reading that table right, A-Rod's OPS is wins of 6+ runs is 1.790?!?! Is that possible?

By the way, those numbers are breakdowns by final game score, not by score at the time. Which is what you want to measure.


Oh?


A 2-run homer in the first inning of a game won by 11 runs turns out to be low-leverage, because the opposition pitching staff turned out to be awful that day. It shouldn't be lumped together with a 2-run walk-off homer.


So Ortiz's 3-run home run to put the Sox up 10-0 becomes higher leverage when the Sox pitching gives up 8 runs, while ARod's 3-run home run in the same situation is less so (since Torre brings in the Gordon/Rivera hammer whenever the other team gets within 6 runs)?

I'm unconvinced.

If you're going to talk about the leverage of the innings, it's gotta account for game situation at the time -- win expectations before the at bat and after the at bat. Not win expectations once the final score is known.

Eric,

This may become a fun thread.

If indeed Ortiz's performance was that outstanding (I have no idea) in terms if timing, it would surely be one of the ages. So, most of the time, the clutch effect is within 1 win, but of course, there's no limit to it.

***

Now, on to the "game state at that point in time" v "retroactive effect on game".

The game state argument is very simple. All future events are unknown, and therefore, our expectations of them are centered around past results of what is to come. This is exactly how a fan looks at the game (he gets excited with a HR in the first inning, since he won't know the game will end 10-0). This is exactly how a manager looks at the game (he'll bring in Mo with the bases loaded in the 8th and 1 out, even if the Yanks will end up hitting 7 runs in the 9th). This is exactly how players approach the game (they'll hug the line in late/close games, even if they end up scoring tons of runs later).

In short, the game state captures, in a quantitative fashion, our expectation of what is to come, regardless of how it will actually come to pass.

On the flip side is the retrospective outlook. Once the game is finished, you then go back, and start to assign the appropriate win levels to each event. If a guy hits a grand slam in the top of the 9th (say like Vlad in the playoffs), only to have his team then blow the game later, that event did absolutely nothing to the Angels. The net effect of that was the same thing as striking out. Since the Angels didn't win, the win impact is zero.

Taking it to the logical conclusion, any event in a team loss is worthless. Any batting out in a team win should not be counted as a minus (so, it prevented the team from winning by 4 runs instead of 3... no change in the outcomes, since a win is a win). Any pitcher giving up 10 runs in a win won't get any negative in the deal.

So, those are the two viewpoints. People usually don't like either (usually the second viewpoint), so they decide to "meet halfway", which not a good position to take. Once you clearly define what it is that you want, once you establish your question, you have no choice but to accept the answer. If you don't like the answer, you change the question. Now, all of a sudden, you have a question that is not necessarily the one that is being argued.

So, some people will say "look, I want to give Vlad some credit, but his team did lose, but I don't want to give him so much credit, because his team did lose, and I want to treat the known outcome of the game, so that if the Angels didn't blow it later in the game, then Vlad's slam should count for more, after the fact".

If you want to construct the question that way, then I can't argue with your answer. But, don't then say that this represents the answer to the original question.

Btw, David Smyth (BaseRuns, PythagoPat) argues for the second method by using "Asbolute Wins and Absolute Losses".

He supports giving Vlad 0 wins for his late inning heroics in a loss, but he gives out Absolute Losses to the batters who make an out, and the pitchers who give up runs. So, Vlad will still come away with some absolute losses (since he wasn't 5-5), but other hitters on the team will get more Absolute Losses.

Over a season, this line of thinking will probably come close to the Game State method. However, just because it approaches it, doesn't make it right. I still don't believe that you can give Vlad nothing for his performance, but, David's method does support his question. As long as you believe that that's the way baseball works, then the Smyth approach is one that you can argue.

A-Rod now has 19 game winning hits this year, which is apparently the most in the AL (MLB too?). I can't find the article though. Does anyone know the one I'm talking about?

As for clutch...

I don't see how it's fair to say A-Rod isn't clutch or Ortiz is clutch because of how the game ends up. Would A-Rod's HR last night have been any less clutch if Mo had blown the save? Was it somehow less clutch because we ended up winning by 4 runs instead of just 2? To me, the only fair way to decide if the hit is "clutch" or not is to look at the game situation when it's hit.

I agree with Devil on this one. If you could, I would look at impact of each individual act on the probability that a team is going to win a game. To me, if a player leads off the game and hits a homerun, the impact on his team's chances to win, when he hit the home run, are exactly identical regardless of whether his team wins a result.

This is why I count homeruns in extra innings of tie games as more valuable than homeruns in the first inning, regardless of the final outcome.

I'm still confused as to why we'd look at anything other than "at the time the event happened, how much did the player increase the chances that his team would win?" That, in all senses of the word, is how much value a player adds to the team. His team had a 30% chance to win before he came up, now it's 60%. Over the course of the season, did the situational nature of Ortiz's hits give him more win value added as compared to Alex Rodriguez?

I would look at impact of each individual act on the probability that a team is going to win a game

Yes, this is exactly what I do.

I'm still confused as to why we'd look at anything other than "at the time the event happened, how much did the player increase the chances that his team would win?" That, in all senses of the word, is how much value a player adds to the team.

Well, if you have some simple cases, and that measure gives the wrong answer in those cases, then the measure must be broken.

For example: does the value of a pitcher's performance change depending on when the offense scores? Is eight shutout innings really more valuble when his teammates score in the top of the ninth, rather than in the top of the first?

Is the value of a leadoff home run against Pedro Martinez the same as the value of a leadoff home run against Charlie Brown?

If somebody has different answers to those questions than you do, then their opinion of the value of this measure is going to differ from yours.

At its root, this approach assumes that value and average value can be interchanged freely.

VoiceofUnreason, you're presenting a false set of choices. You're saying that we know the team will score in the future, which we certainly do not know in a regular game.

We have a measure to look at the value of a pitcher's innings. We can say, "When the game started, the pitcher's team has X amount of chance to win, when he left the game, his team had Y chance of winnings. An average pitcher would have given the team Z amount of chance to win." The value of a pitcher's innings does change with respect to the run scoring environment of the offense. It doesn't affect the quality of the pitcher, which is why I don't think we should take wins into account when looking at the Cy Young. Do I maybe think we could be more contextual when looking at pitchers as MVP candidates? Sure.

Starting pitchers are different than hitters, in that other events in the game aren't given, many more are in his control. In Ortiz's case, the context is given, and then we are able to look at how the game changes given the events that he causes.

Wow, I just realized that A-Rod has a great chance to win the batting title as well. I had no idea.

I think Ortiz gets the MVP. Ortiz has the "leader" rep. while A-Rod has the $252 million slappypanties tag.

Look no further than 1995:

Player/Pos.       BA/ OBP/ SLG   R    2B  HR  RBI  *OPS+ 
Albert Belle LF .317/.401/.690  121   52  50  126   178
Mo Vaughn 1B    .300/.388/.575   98   28  39  126   145

Red Sox win the division with 86 wins, Tribe does likewise with 100 wins.

Vaughn was considered a leader, the other problematic. Belle played the more challenging defensive position. Guess who took home the hardware? The writers viewed all of Vaughn's "intangibles" as more than valuable than the difference in production and position that Belle had.

Best Regards

John

Wow, I just realized that A-Rod has a great chance to win the batting title as well. I had no idea.

Somewhere, Placido Polanco weeps bitterly.

No, Joel, I'm not saying that we will know what the team will score in the future. What I am saying is that we don't know the value of an event until the game is complete[1].

Example question: what is the value of a leadoff home run in a game that gets rained out?

If you insist that value must be measured at the moment of the event, you're going to get a different answer than that which you would find if you instead wait until the game (or season, if you prefer) is over.

I see at least three distinct, but reasonable views. One is that the change in win probability is value, a second that it is an expression of average value - but what we want is actual value, which requires additional information that we don't have until the game is complete, a third view that it is an expression of average value - but that average value alone is sufficient for our purposes. The relative appeal of each of these will depend on an individuals internal understanding of value.

[1] Actually, I didn't say that, specifically, but it is the direction I currently lean in this discussion.

Polanco would be clinging on to first at .332, if he counted. I hope, for his sake, Young or Arod passes him.

The biggest problem I see with the "wait till the end of the game" argument is that managers don't do that, they make their decisions given current game states, meaning that a game that is not so close ends up being close. Suppose in a tie game Vlad hits a slam that puts the angels up by 4, we will know exacly what he did to their chances of winning. Because of this Soscia puts in a pitcher other than K-Rod, the game may end up being closer than it would have been otherwise.

Look, if I fix your car door that had a big dent in it, make it brand new, and then you go out and get it banged up again, did I not perform a service? Was it not valuable? How much do you pay me? Less because you went and banged it up? If I had never fixed it, would banging it up be bad (If Ortiz hadn't hit that home run in the clutch, fixing the situation, would giving back the runs have even mattered?)

That's how I think we must think of a play in baseball and its contextual value: What state was the game in when it got here and what did player X do to change that state? We have to use value-added because it affects the future events and how we interpret them as well.

John, I don't think Ortiz is considered a leader, he's considered to have been clutch this season. Given that, we believe it adds value to his tangible value, within the context of the season and not his skill level. We are arguing as to how to measure those tangible benefits with respect to the game states (quite interesting i think), and we were arguing how to quantify the deficit that being a DH ought to be on your value. Nothing in this thread has been particularly about intagibles, and this argument is nothing like the stupid Vaughn/Belle one.

Nothing in this thread has been particularly about intagibles, and this argument is nothing like the stupid Vaughn/Belle one.

Are you sure? Heck, until it's quantified "clutch" is the greatest intangible of them all.

See: Derek "The Clutch God" Jeter AKA Captain Intangibles

I don't think this thread has addressed the mindset of the BBWAA. Are we discussing who we feel should be the MVP or who will be named the MVP?

Best Regards

John

It's not any better to say a win is a win is a win than it is to say a run is a run is a run.

A players value can be determined in terms of anything runs, wins, penants, extra hot dogs sold. It would be cool if somebody we to take the time to calculate penant probability added.

or playoff probability added

John,

As several people have shown, the idea that Ortiz has performed well in important situations is not just an "image." Its not based on "intangibles." They are based on actual performance, as measured by statistics, in situations which most people agree are important. We may disagree on the exact amount that the performance situation is worth, but it is not at all similar to the Derek Jeter situation.

I'm still confused as to why we'd look at anything other than "at the time the event happened, how much did the player increase the chances that his team would win?" That, in all senses of the word, is how much value a player adds to the team.

I agree that that's a better way to measure clutch performance than the alternate method that Tangotiger was talking about. I don't think the second sentence is necessarily true; that is the value of the player's performance in those situations, but in many cases, the value is provided by the situations.

In other words, if a great player gets to hit in a lot of clutch situations, he will look like a Clutch God even if he performs no better in clutch situations. His performance is leveraged due to the situations; it's real value to the team, but I don't think it's self-evident that the player deserves all the credit for that value.

This is the same issue we've talked about with closers.

Pennant probability added: I think that this may be a possible approach from the one and only Sean Forman.

Wrong link for Forman. This is the right one.

the one and only Sean Forman?

That's like being the one and only Robert Machemer, I guess.

BTW, I agree that the first way you'd want to look at this is Win Expectancy. However, taking final game score is probably a better approximation than taking score at time. For instance, ARod homers in the bottom of the first with one out and a man on first, increasing WE from 60% to 75%. Ortiz homers in a tie game, top of the 8th, 2 outs, man on 1st (as he did two nights ago in Toronto), it's 42% to 88%. The winning margin in the first game is very likely to be larger than 2 runs.

And if it is 2 runs, you can argue that the 60% to 75% improvement in WE understates the importance of the homer. So even after calculating delta-WE, you'd want to at least look at the splits by game outcome to see if they are informative.

Re ARod's 1790 OPS in 93 PA in games won by 6 or more runs: he has hit .544 / .613 / 1.177 in those games. In contrast, he's 1-16 with a SF and no BB or XBH in the Yankees' 4 extra-inning losses.

You know, I talked with Vinay about this a long while ago, and my earlier statement was incorrect as to what clutch is.

A: performance, regardless of context (Linear Weights)
B: LI (leveraged index)
C: PA
D: context-based performance (change in Win Expectancy)

A*B*C + clutch = D

So, if a guy is +.01 wins (according to LWTS) per PA, has an LI of 1.2, 500 PA, and +6 wins, what's his clutch rating?

If his LI was 1.0, he'd be +1 wins in clutch. But, in this case, he's:
+.01 * 1.2 * 500 = +6, and so is worth 0 wins in clutch.

He was just lucky enough to have been presented with opportunities to shine, and he played at his normal level and shone at his normal level. (Similar idea for relievers.)

Eric,

I think you are proposing two different questions here. You state that you prefer Win Expectancy, and therefore, there's only one way to go there.

At the same time, you are asking about "more informative", and certainly having different breakdowns will inform you better (Joe Carter's Series winning hit was huge, but it's one event that completely overshadows anything else he did, and therefore doesn't inform you as you may want).

Your statement here shows the conflict in the question: And if it is 2 runs, you can argue that the 60% to 75% improvement in WE understates the importance of the homer.. The importance of the HR is exactly what it is. If it is understated, then the expected win% should have been higher, and it's not.

I see this as an issue where people want one answer, but there are multiple questions on deck here.

Tom

Tango,

I understand the logic, but how is the change in Win Expectancy determined? Does someone go through each at bat and figure it out?

I understand the logic, but how is the change in Win Expectancy determined? Does someone go through each at bat and figure it out?

Yes, you need play-by-play data (though for pitchers, inning-by-inning data suffices). Studes and GGC put together a pretty cool spreadsheet for tracking WE by hand.

I think what is really important is how Curt Schilling feels about this matter. Has Curt enlightened everyone on who is more deserving?

Re ARod's 1790 OPS in 93 PA in games won by 6 or more runs: he has hit .544 / .613 / 1.177 in those games.

But isn't the fact that A-Rod had a 1790 OPS in those games a big reason that his team won by 6 or more runs? Maybe if his OPS had just been 1200 in those games his team would have only won by 4 runs. There are some chicken-and-egg drawbacks to this approach.

Dizzy, yes, that's right. I did this for 99-02, and published the top/bottom guys at one point. Others have gone through 72-04. It's the same process to determine Leveraged Index (LI). I will be publishing the win prob charts for all half innings, and scores from -5 to +5, in our book, so, you can do this too.

Vinay: if you separate fielding from pitching, then you do need to go play-by-play.

Hawk, I'm with you here. A guy hits a slam in a tie game, and all of a sudden it's a 4-run game.

The problems with all the non-as-it-happens approaches, is that they have.... problems. The as-it-happens simply reflects exactly what fans, managers, and players see and do. All the after-the-fact approaches are theoretical and fun to talk about, but, I see no application for them as an analyst, or use for them as a participant. I just don't see the benefit.

The problems with all the non-as-it-happens approaches, is that they have.... problems. The as-it-happens simply reflects exactly what fans, managers, and players see and do. All the after-the-fact approaches are theoretical and fun to talk about, but, I see no application for them as an analyst, or use for them as a participant. I just don't see the benefit.

I mildly disagree for one of the reasons that you stated earlier.

If you look at the comments in this thread, you see the problem. Everyone wants a "number" for clutch, or else they consider an actual result as "intangible".

Value is only measured against an objective. All of the following measure value:

(1) Change in WE at the time of the event.
(2) Change in WE above a certain threshhold at the time of the event.
(3) Perceived change in WE at the time of the event.
(4) Value to Win with perfect knowledge.

The first defeinitively measures performance based on psychological context. The second and third defentively alter plan for resource utilitzation (which effects contextual WE for later intervals). All four secondarily measure context, and win goal objective.

Its not a search for which is best. If you want to get into a real dogfight, (3) is probably the best as it subsumes psychological environment, competive environment, and resource utilization, but you don't have to reach that question. Because leveraged performance will often appear in all four categories, and when it does not its metricizing different value.

IMHO, this is one of the problems with sabermetrics. Everybody wants to be a combinatorist so they can name some new stat, and get credit for the result.

Combinatorics is just instroducing error, because the weighing functions tend to be context dependent. Every single one of them depends on the plan. If you have Earl, pitch them to they drop Weaver, you don't get the same advantage in (3) than you would with Bobby spread the work around Cox. So the combinatorics end up causing the problem.

Figure out all four numbers to the best of your ability. If need be, figure out the team dependent utility derivived, but you don't need that type of precise quantification you are wasting cycles. They are all valuable components. Calculate what you need to get value. In the Ortiz v. Arod situation we do need to calculate some of the number sthat Mr. Van derived.

And another highly leveraged home run just flyes out of the park off Ortiz's bat.

"Look no further than 1995:"

and Belle put up most of his numbers in the 2nd half, when Cleveland already had a 30 game lead. in the 1st half of the season, no one talked about him as MVP...it was Lofton and Baerga.

Tango, we're probably not in any substantial disagreement over this, maybe none at all. My example of the first-inning 2-run HR "understating" the change in WE was meant to refer to a game that ended 2-0 and could have been expected to be a low-scoring game. The true delta-WE of that event in a pitcher's park with the wind blowing in, in 1968, Gibson vs. Drysdale, is a lot different from the WE in Coors. Final outcome run totals and win margins are thus a potentially informative proxy for the extra work we'd ideally be doing with WE itself.

I think you can go (in theory) very far with a WE approach without introducing any difficult philosphical conundrums. The first one you run into can be summed up by this question -- what changes, if any, do you make in total run expectation, and hence delta-WE calculations, for offenses facing Aaron Small this year? You run into a goofy situation where Small's performance in 2006 (as he argues for or against this year being a fluke) alters our WE analysis of the 2005 MVP race.

I think we are in agreement then. You absolutely have a different WE chart for Coors 95 with a bad pitcher and Astrodome 75 with a great pitcher.

Look no further than 1995:

Edgar Martinez says you are right. It seemed as if he carried his team to the playoffs that year. He was fantastic.

----"Btw, David Smyth (BaseRuns, PythagoPat) argues for the second method by using "Asbolute Wins and Absolute Losses"."

Well, I saw a gap where AFAIK nobody else had really attempted to put together a method to answer the value question from that after-the-fact perspective. So I made an attempt.

I don't really have a definite preference one way or the other. Or, to put it another way, if if I am voting for MVP, I want more information, rather than less. So, I'll look at the "ability" perspective, the "real-time" perspective, and the "after-the-fact" perspective, and decide how much weight to place on each.

It is entirely possible that the after-the-fact analysis will yield a significantly different leader-list than the other approaches, and that is information I would not mind having.

If forced to choose only one approach to use for MVP, I would go along with Tango's real-time point of view.

I think we are in agreement then.

I don't think you are. You both seem to be going for a harmonious approach, rather than trying to answer the question.

Eric,
you posted the charts, and stated Ortiz was +3 wins based on his clutch performance.

Now Tango says that clutch performance isn't measured very well. You agree that it isn't (I think).

So is Ortiz +3 wins based on clutch or not?

And who is the MVP?

Tango? Eric?

(hint: it's ARod).

Oh, and that's not to say you have to insult one another, but don't walk away from an unaddressed issue with some "not that far off" handshaking.

It can still be a nice discussion, but *what's the answer*?

what's the answer

Allen Iverson.

It depends on the question. If it is "What is clutch?"

The answer is definately not "The Clausen Brothers"

But, likewise the answer is not a single stat.

Both approaches, and as of yet uncompiled approaches provide part of the statistical answer.

If you want to combine them, then each player is going to have a different weighing formula depending on the events and needs of his team.

You don't use macro tools to make micro measurements.

Wouldn't DH replacement level simply be the MLE of the best minor league veteran type out there? (Rick Short this year, for example)

I'm not really walking away, but neither do I have the desire to debate all the nuances. I find that people appreciate a tool to build a house, rather than an already-built house.

That said, I can only support a real-time win probability method since this is what reflects the perception of players, fans, and coaches.

There are two ways to calculate real-time win probability: the easy fast way, where you assume that each team has an equal talent base at all times of the game, or the long complicated way, where you consider the exact makeup of the entire context.

The easy fast way gives you the same win probability, whether Bonds is on deck, or Santiago.

The long complicated way? You start by pre-assigning win points to all 50 players, and you consider what the park and temperature is. If you have Clemens, you give the Astros a .70 chance of winning the game. Clemens is pre-assigned say +.17 wins, various players are pre-assigned +.04, or -.02, or whathaveyou. It's all based on the actual context and our expectation as to what will happen. You would even consider if Lidge pitched in the last 4 games, so that maybe the Astros have a lesser chance of winning if he's not available, etc.

If Clemens pitches exactly like he's supposed to, he'll end up with +.00 in-game win probability wins. If he pitches better, he might get +.04. His total win probability will be +.17 or +.21, etc.

This makes more sense as well for the batter. With Bonds on deck, the win probability of the current batter shoots up, because that's what's really happening. However, since we pre-assigned Bonds an enormous amount pre-game (say +.10 wins), it will all balance out at the end of the game (if Bonds hits exactly like he was supposed to.... it's hard to do, since there are thousands of combination of game states, and Bonds will only be involved in a handful of ones).

Knowing the park, the pitcher, the guys in the bullpen, the guys on deck, the speed of the runners, etc, all affect the win probability.

In the end, the quick fast way will come awfully close to the long complicated correct way.

Wouldn't DH replacement level simply be the MLE of the best minor league veteran type out there? (Rick Short this year, for example)

I would look at the projection instead of the MLE. Rick Short's MLE for this year was around a .340 average. You can't predict that. For next year, he's not likely to hit .340 in the majors, or .380 in the minors, he's probably a .280-.290 hitter in the majors.

Its still silly to see Rick Short and Kevin Orie, 2 minor league vets having great seasons, stuck in the minors while Vinny Castilla gets the major league at bats.

Yes, that's one thing that p-sses me off about "replacement level". It is based on true talent levels, and *not* on past performance levels. Picking, after the fact, the best (or worst) performances, and then setting that as the bar, is selective sampling at its worst.

And, don't get me started on calculating hitting, pitching, and fielding replacement levels, and then combining them for a player replacement level. This doesn't approach reality in the least.

"Thats like adding an extra 4-6 homers to his traditional stat line compared to Arod who I don't think has hit any."

Not that this is relevant anymore but that shot against Schilling in Fenway was just as good as a walkoff IMO. His bomb to tie the game against Wickman in Cleveland was pretty huge too, despite not being a walkoff. Those are just two off the top of my head, I'm sure there are others.

However, taking final game score is probably a better approximation than taking score at time. For instance, ARod homers in the bottom of the first with one out and a man on first, increasing WE from 60% to 75%. Ortiz homers in a tie game, top of the 8th, 2 outs, man on 1st (as he did two nights ago in Toronto), it's 42% to 88%. The winning margin in the first game is very likely to be larger than 2 runs.

But that doesn't mean that the winning margin would have been (n-2) without that HR. There are *so many* decisions that are directly relevant to the game state, and that would have been different had the run-scoring pattern been different.

-- MWE

bump due to todays BPro article:

http://www.baseballprospectus.com/article.php?articleid=4463

Premium unfortunately. But this is relevant:

Batter          Team   WINS   VORP
--------------- ----   ----   ----
David Ortiz      BOS   7.12   80.3
Alex Rodriguez   NYA   4.59   93.1

They use win expectancy added (the method I prefer), so ARod would have to be about .5 wins in the field to beat Ortiz (using Tango's number)

VORP is really the wrong stat to compare to that win expectancy number, because the baselines are different. MLV is runs above the average batter, so that makes more sense:



This really shows the clutch difference. ARod is a win or so better even before adjusting for position. But he's 2.5 wins behind after counting game situation. I still think his defense is enough to give it to him (and as I mentioned in #49 on this page, I don't think it's fair to just credit a player with his WINS total).