Cardsfanboy, this thread is for you!

AO

Fascinating stuff. Not sure what to add, other than I think I reason to be suspicious that all these new Nationals fans around me are in for a disappointing remainder of the season.

Nice article Studes. I think these summary articles are a great jumping off point for the debates and discussions that usually ensue.

Again, I think a lot of the backlash against these is in using a term like "luck", which carries a connotation. Winning one run games requires skill but your overall performance in the subset of games that are won by 1 run clearly has a nice chunk of random variance involved.

Now, what part of the pythagorean therom of W% is the hypotenuse?

Thanks, Studes, that starts to address a lot of the stuff I've been whining about recently regarding Pythag.

Let's just assume for the sake of argument that the Pythagorean Theorem is in fact some kind of sacred altar to be worshipped at when it comes to predicting how a team will do.

If that's the case, then the Nationals should play around .500 ball the rest of the way and end up with around 88 wins.

88 wins probably won't be good enough to make the playoffs, but it's going to be pretty darn close, which means that according to the sacred Pythagorean Theorem, the Nationals are going to be in the hunt for a playoff spot right up until just about the bitter end. Better get used to it!

And anyone who says that they thought before the season started that the Nationals would be contending for the playoffs throughout the whole entire season is lying. Disappointing? Barring a total absolute collapse, this team has pretty much already exceeded my expectations.

Oh, and Jose Vidro should be coming back soon, which can't do anything but help the team.

A true .500 team, with 81 games to go, should be expected to win between 32 and 50 games 19 times out of 20. So, their skill will make them win 41 games, and the timing/circumstances/variations of events outside their control (i.e., luck), will make that number anywhere between 32 and 50, most of the time. Luck gives every team hope.

I also recommend this pythag article from Ben Vollmayr-Lee.

Thanks everyone, and Tango too. I forgot Ben's article.

The big question everyone wants answered, however, is how to explain that remaining 9% variance, which can be much more than 9% for some specific teams.

Which does rather presuppose that the remaining 9% variation can be explained. My belief is, once you get up above .9 R^2, you might as well stop looking for causal reasons. Of course, if you introduce additiona; variables into the regression, R^2 will improve (modestly). But it doesn't take more

The rest of the variance is explained by the outcomes of close games.

In a broad mathematical sense, there are two things that affect the unexplained variance:
* Specific run distribution patterns, as discussed above, and the timing of those patterns in specific games.

I've got to admit, I don't see how those two statements are related.

Did you check to see how blowouts affect the accuracy of Pythag? I'd be interested to see how Pythag%-Win% correlates to a simple count of "number of wins by 8+ runs". If I score 3 runs in 3 games, I'm not going to win many of those regardless how I distribute those runs. If I score 15 runs in three games, the distribution *really* matters as to how many games I might expect to win.

Of course, if you introduce additiona; variables into the regression, R^2 will improve (modestly).

I used adjusted R squared, which offsets this.

Did you check to see how blowouts affect the accuracy of Pythag?

Blowouts are typically captured by the distribution analysis. If you blow a team out or are blown out, the high score will be reflected in the actual distribution of runs.

I've got to admit, I don't see how those two statements are related.

Not sure how I can be clearer. It may have been confusing that my two bullet points referred to the larger math, but the subtitle referred to only the second of the bullet points. Does that help?

interesting that when you went to the kc link that had 1 run record listed by saves holds, that both boston and st louis still have a poor record in one run games yet good bullpens.

Isn't it arguable that if you have a good bullpen and a good offense that you will probably have a worse record in one run games as your pen stops the other team from scoring but your team keeps scoring?

The Sox are sure "lucky" that Juan Uribe has a cannon. Uribe got Polanco last night at first with a throw from left field. The winning run was on third and if Uribe doesn't make the play the Tigers win.

Coming up with a huge defensive play, a key strike out, taking the extra base on a hit to right, advance scouting so your fielders are in postion, or just playing fundamental baseball is going to give you the edge in 1 run games.

I don't understand the fascination with one run game records. It's like giving extra weight to batters left on base with 2 outs or something like that. Shouldn't good teams lose their games by one run. Would it be better if they lost by 2 or 3 or 5 runs? And shouldn't bad teams win most of their games by 1 run, because they aren't good enough to win by more than that?

I'm with you Dirk. I've never liked the idea of looking backwards after the game is over. It's just not the way we watch baseball.

What I prefer is looking at the score after 6 innings, AND THEN seeing how the teams went. So, when the Yanks are up by 1 after 6, how did they do, etc...

Coming up with a huge defensive play, a key strike out, taking the extra base on a hit to right, advance scouting so your fielders are in postion, or just playing fundamental baseball is going to give you the edge in 1 run games.

Nobody really disagrees with you. But the point remains that there's huge variation in one-run records from season to season, so why do those things work one year but not the next?

Why? Well how about because being a good defensive team, being a good baserunning team, advance scouting so your fielders are in position, etc. all help you just win games period. The impact of those things is mostly or entirely picked up by the pythagorean already.

Second, you can't really time when you're going to make the great play or take the extra base that decides the game, etc. Most 1-run games end with popups to SS or line drive singles after a double or other perfectly mundane baseball events, not great plays. Meanwhile, most great plays get wasted in multi-run wins (or worse yet, losses). OK, point #2 is really the same as point #1.

I'm curious how many "one-run" wins resulted from situations where a team led by more than one run entering the 8th or 9th but their bullpen gave up enough to make it a 1-run game.

I'd be interested in seeing home/road breakouts. I know the Twins are just phenomenal in one-run games at home from 2000-2004 (win percentage over 710!) The A's and Dodgers are also in the 700 range at home.

Obviously there's a big advantage in batting last -- every single team is over 500 at home in 1-run games over the last 5 years while only 17 are over 500 in games decided by 2 or more runs. But the Twins and Dodgers are among the top in Studes' "over-performers" list and their over-performance seems to be entirely home field over-performance. If it's particular strategies or the bullpen quality or the defense, it's not clear why they work at home but not on the road. (but maybe the home/road differential isn't so extreme relative to the expected record)

Meanwhile, the Reds are among the worst (overall, not relative to expectation) at home but are #2 overall on the road (I assume #1 by a wide margin relative to expectation).

By the way, the Cubs have played the second most number of one-run games at home over the last 5 years. Wrigley doesn't spring to mind as "home of the close game".

By the way, the Cubs have played the second most number of one-run games at home over the last 5 years. Wrigley doesn't spring to mind as "home of the close game".

I'm guessing most of those games the Cubs should've won by 3-4 runs, but Dusty managed to #### up just enough to make it exciting for the fans.

Isn't it arguable that if you have a good bullpen and a good offense that you will probably have a worse record in one run games as your pen stops the other team from scoring but your team keeps scoring?

Exactly, which I think is in basic agreement with Walt's statement:

Well how about because being a good defensive team, being a good baserunning team, advance scouting so your fielders are in position, etc. all help you just win games period. The impact of those things is mostly or entirely picked up by the pythagorean already.

Also, I think a football analogy might help. Suppose USC and San Diego St. play 100 times. San Diego St. might be expected to win, let's say, 3 times. How many times would San Diego St. "blow out" USC? Probably zero. So you'd have a USC record of something like:

OVERALL 97-3 .970 Win%
Blowouts 80-0 1.00 Win%
Close Games 17-3 .850 Win%

The difference is the effect in baseball is more subtle because the teams are closer in ability/WinLoss%, and scoring is lower, and thus subject to more luck, but the effect of 1-Run/Close games being biased towards a lower win% for good teams and a higher win% for bad teams is the same.

And anyone who says that they thought before the season started that the Nationals would be contending for the playoffs throughout the whole entire season is lying.

Actually, the DMB sims I ran on Dan's ZIPs projections had the Nats around 84 wins on average, which is contending for the playoffs in the NL East.

I'm guessing most of those games the Cubs should've won by 3-4 runs, but Dusty managed to #### up just enough to make it exciting for the fans.

Now, now, Baylor "managed" to #### up twice as much as Baker.

Since this is all tangentially related to the White Sox.

Record when scoring 2 runs or less

2004: 1-42
2003: 2-42
2002: 2-42
2002-2004: 5-126

This season
2005: 7-12

Here is an interesting stat I heard during today's Sox game.

The 2005 White Sox are 7-10 in games where they have scored 2 runs or less.

In the last 3 years the White Sox are 5-168 in games where they score 2 runs or less.

2004: 1-42
2003: 2-44
2002: 2-42

Uh, cheat and no neck, why don't you two get a room?

Uh, cheat and no neck, why don't you two get a room?

Cheat is right on the 2005 stats 7-12.

I should have listened to Hawk dagnabit instead of crunching the #'s myself.

studes : Fair enough on adjusted R^2..., but I still don't quite follow the second point, or understand the source of the syllogism that tells you the "missing R^2" must come from one run games.

Also, can you be more specific as to what variables you used to regress the "run distribution" part?

Surely the key factor in turning runs into wins-above-pythagoras is how your run distribution meshes with oppositions. Consider the results of the following homestand :

15-11
7-4
1-0

Now consider the following homestand
15-0
7-11
1-4

The RS, RA and distributions of scoring are exactly the same, but there's a two-win swing from between the two score lines that cannot be explained simply from results of the close game. Over an infinite season, the effect of these synchronicities between scoring distributions, would balance out, but over 162 games, there's no compelling reason why it won't win or lose you a few more games.

There's a word for this ability : luck.

To suggest that the a sensible selection of inputs must result in 100% R^2 (adjusted or otherwise) is a case of underestimating the fog.

What I prefer is looking at the score after 6 innings, AND THEN seeing how the teams went. So, when the Yanks are up by 1 after 6, how did they do, etc...

Could Tango (or someone) explain this? I'm not really following.

As dirk #12 says, you don't know why the game ended up as 1 run. Was it that a team was ahead by 5 and gave up 4 runs? Did they come from behind 1 to win by 1? etc

The whole reason we are talking about a game ending by 1 run is that we are inferring that it was a close game. That's really what we are talking about: how does a team perform during a close game? So, it's silly to look at the final score, and make that inference.

The better way would be to see how the win probability changes throughout the game. That'll tell you how close the game was.

A quicker way would be to look at the score after 6 innings. That's when strategies start to take place (which is what we are talking about as well). So, you want to know how well the Tigers 2003 played when leading, tied, and trailing by 1 after 6. I'm quite sure that you'll get results that make much more sense than what we are seeing now.

Over an infinite season, the effect of these synchronicities between scoring distributions, would balance out, but over 162 games, there's no compelling reason why it won't win or lose you a few more games.

Exactly. So if you choose your variables from a 162-game season in a way that captures these lack of synchronicities, then you will achieve an R squared close to 1. That's really all I've done.

This is not a big deal. 60% of all games are decided by three runs or less. So if you take runs scored and allowed, plus add the outcome of those 60% (which tend to be relatively more variable) of course you'll get an R squared of .99. The interesting thing to me was showing the relative impact of each variable.

Whatever those White Sox stats are supposed to mean, it'd be more useful if you removed the times they were shut out from the stats - nobody's winning any of those games.

every single team is over 500 at home in 1-run games over the last 5 years while only 17 are over 500 in games decided by 2 or more runs.

Isn't this due primarily to the fact they stop playing when the home team goes ahead in the 9th or later, and so will win these games by 1 run (unless it's a multi-run homer), while the visitors can win by as many as they want to score?

In a nutshell, James found that there is some evidence that some teams display an ability to perform better or worse in one-run games independent of their overall talent level. The teams that show this ability have two fundamental traits: they play small ball (sacrifice hits, stolen bases, fewer home runs, etc.) and have good pitching.

What would interest me is to see how these 1-run-small-ball-overperformers performed in 2 and 3 run games. Did the small ball hurt them overall, or were 1-run-wins pure Pythagorean surplus?

studes : Oh, right. I see. That makes sense. I misunderstood
a) what you were trying to show
b) how you were trying to show it.

I'm still interested in the methodology about regressing run distribution, though, if you've time.

I'm still interested in the methodology about regressing run distribution, though, if you've time.

What I did was pretty simple. I took each team's distribution of runs scored and runs allowed, and then basically multiplied the two together (percent of time scoring three runs times percent of time allowing less than three runs, etc.) and added them into a composite winning percentages.

I then regressed those composites against actual winning percentage, and found that this approach reduced the standard error from 3.87 wins (straight Pythagorean) to 3.53 wins.

Isn't this due primarily to the fact they stop playing when the home team goes ahead in the 9th or later, and so will win these games by 1 run (unless it's a multi-run homer), while the visitors can win by as many as they want to score?

Obviously that has an impact, but it only comes into play in games that were tied at some point in the bottom of the 9th and eventually won by the home team and I suspect that's only a relatively small percentage of 1-run games. I don't think it would be enough to explain why all teams are over 500 at home in 1-run games.

But maybe teams are winning more games in their last at-bat than I think. And maybe nobody but me is surprised to find every team over 500 at home in 1-run games.

Another aspect that has me scratching my head is that, as I mentioned, in games decided by 2+ runs, only 17 teams are over 500 at home.

Overall, for 2000-2004, home teams win 53.8% of the time and are 928 games over 500. 867 of those 928 games are 1-run wins. They win 61.6% of those vs. just 50.9% of games decided by 2+ runs.

Even with last at-bats and the game stopping once the home team leads, that seems an extreme difference to me.

Or at least if anyone had told me that the "natural" homefield advantage was almost entirely due to 1-run wins, I'd have been surprised.

But I don't read much sabermetric stuff.

It's not a recent phenomenon. Take it back to 1969 and the worst home team in 1-run games (other than the Pilots and even they were 500) is the Senators at 565. Among teams in existence that entire time, it's the Angels at 578. And half the teams are over 500 in 2+ run games. The respective win percentages are practically identical to above -- 608 in 1-run, 508 in 2+.

Those extra wins can't all be bottom of the 9th wins, can they?

That's mindboggling!

It sure doesn't seem that way annecdotally.

Obviously that has an impact, but it only comes into play in games that were tied at some point in the bottom of the 9th and eventually won by the home team and I suspect that's only a relatively small percentage of 1-run games. I don't think it would be enough to explain why all teams are over 500 at home in 1-run games.

I don't know, it might be. Remember we're talking not only about 9th-inning rally wins, but also the great majority of extra-inning decisions.

I suspect 9th-inning rallies and extra-inning games make up a significant proportion of 1-run games, and home teams have an obvious large advantage in these games.

Walt, I don't have the same data as you. Here's a link to all teams' home won/loss records from 2000 to 2004 in one-run games:

http://www.baseball-reference.com/games/situational.cgi?from=2000&to=2004&0=0&1=3&rsgtlt=gt&rs=5&ragtlt=gt&ra=5&2=6&trgtlt=gt&tr=10&3=9&mvgtlt=eq&mv=1&teams=team&years=each&submit=Run+Situation

On this page, there are twenty teams under .500 and an overall Win% of .616, according to the data I downloaded from Retrosheet. If I'm interpreting the data correctly, the home team had a .608 (2000 to 2004) record in one-run games that lasted nine innings and .643 record in one-run games that went into extra innings.

I see the same phenomenon you do, but not to such an extreme degree.

Sorry about that. Here's the link.

Plus a little more info, based on games from 2000 to 2004:

There were 3,298 full one-run games.

1,203 (30%) of those ended in the top of the ninth, which obviously means the home team won.

359 (11%) were decided during the bottom of the ninth, again in favor of the home team.

1,004 ended after nine full innings, which means the visiting team won.

732 (11%) were tied at nine, and were decided in extra innings.

Don't know what all that means, but I thought you might be interested.