You're the GM (Game)

[brett05]

Then you're wrong. Of course the two are related due to their obvious dual impact. Let's say you have an ace pitcher. His WAR is so high because he pitches a lot of innings and gives up few runs. Well, that stat also is counted to your overall team stats so now the team is also not giving up runs much either. When teams don't give up runs, they tend to win better than if they gave up more runs.

Of course WAR and Pythagorean win totals are connected as they count the same thing they just count them for an individual (WAR) or the team (Pythagorean).

Explain a scenario where a team has a high WAR total and doesn't win a lot OR has a low WAR win total and does win a lot that isn't explained by "fluky luck in one-run/close games"

you do realize you moved the goalposts, right?

 

[DanTown]

That's the point. WAR judges players not teams. Wins judges teams That's what matters. WAR might get you within 15% of being at the total of wins, but that's not the discussion as far as I can see.

WAR judges a player. adding up all 25 players WAR judges the value of the team.

WAR, unlike other stats like HR/ERA/etc has already accounted for it's value relative to other players. So of course you can add up team WAR and get a good sense for how good a team is or will be.

Essentially, it's a prediction stat, NOT a counting stat. If you think a team is going to win/lose a lot of games, you're going to conversely say that it will have a high/low team WAR.

 

[brett05]

WAR judges a player. adding up all 25 players WAR judges the value of the team.

WAR, unlike other stats like HR/ERA/etc has already accounted for it's value relative to other players. So of course you can add up team WAR and get a good sense for how good a team is or will be.

Essentially, it's a prediction stat, NOT a counting stat. If you think a team is going to win/lose a lot of games, you're going to conversely say that it will have a high/low team WAR.

It can be but does not have to be which is why it's unrelated. I don't understand why this is a sticking point for TC and yourself.

 

[TC in Mississippi]

It can be but does not have to be which is why it's unrelated. I don't understand why this is a sticking point for TC and yourself.

Ah ha! You've just arrived at it. "can be but does not have to be". Is that not a definition of all speculative projections? Adding up the sum of WAR values is as useful a predictor as any other which is to say it is an educated guess. I'm not sure why the argument over how one goes about predicting the difficult to predict.

 

[brett05]

Ah ha! You've just arrived at it. "can be but does not have to be". Is that not a definition of all speculative projections? Adding up the sum of WAR values is as useful a predictor as any other which is to say it is an educated guess. I'm not sure why the argument over how one goes about predicting the difficult to predict.

Well, You're initial comment was WAR equals wins which I commented on that it does not have to, they could, but are not related. Then you and DT went into predictions and the like which is all fine, but WAR itself has no correlation with team wins. The example you used was the 2015 Cubs. It equaled out. That was flukey. Evidence is by doing it to other teams. I checked on three teams, one of which I have already mentioned the 2001 Mariners. The Angels and the White Sox of 2015 do not have their WAR equal their Win totals. And that was my only point of contention.

 

[DanTown]

Well, You're initial comment was WAR equals wins which I commented on that it does not have to, they could, but are not related. Then you and DT went into predictions and the like which is all fine, but WAR itself has no correlation with team wins. The example you used was the 2015 Cubs. It equaled out. That was flukey. Evidence is by doing it to other teams. I checked on three teams, one of which I have already mentioned the 2001 Mariners. The Angels and the White Sox of 2015 do not have their WAR equal their Win totals. And that was my only point of contention.

In a PREDICTION model, total team WAR does equal wins. You wouldn't project Team X with team WAR 45 to win more games than team Y with team WAR of 40.

Obviously due to the fluke nature of some stats like record in one-run games or injuries you can never say "if you end up with team WAR of X, you will win at least Y games". But as a predictive total, yes you can say that if player projected WAR adds up to 50 that you're also predicting the team to win 98 games.

 

[TC in Mississippi]

All data works better the larger the sample size correct? Let's take WAR and apply it to the entire league in 2015. Let's use the following numbers:

Wins at replacement level (0 fWAR)= 47.7 x 30 teams = 1431
Total position player fWAR for MLB 2015 = 569.7
Total pitching fWAR for MLB 2015 = 430.1
1431 + 569.7 + 1431 = 2,430.8

Actual win total in MLB 2015 = 2,429

Variance 1.8 games or .075%

I'd say in the larger sample size fWAR works pretty well using actual stats. The highest variance on team's WAR to actual wins was 7.3%. So on predictive level you would have to believe that projecting teams win totals by WAR is accurate +/- 5.5 games.

In the Cubs example the predictive WAR per ZiPS is 97 wins or a range of 91.5-102.5. I'm guessing, barring injury, that's going to be pretty close.

 

[brett05]

Again, I we are speaking of predictions which I am not disputing.
The dispute is that the individualized players war for a given year will equal the team's win total for that given year. This is a discussion on facts not predictions. And the fact shows that the numbers do not add up when speaking of individual teams as examples that I have provided and have gone uncontested.

 

[TC in Mississippi]

Again, I we are speaking of predictions which I am not disputing.
The dispute is that the individualized players war for a given year will equal the team's win total for that given year. This is a discussion on facts not predictions. And the fact shows that the numbers do not add up when speaking of individual teams as examples that I have provided and have gone uncontested.

I acknowledge that there can be a variance on individual teams actual wins vs the sum of fWAR. Going through all 30 teams I found the highest variance to be 10% which means that even in that case the fWAR to actual wins was 90% accurate or, in mathematical terms, statistically significant. I don't think any stat is 100% accurate 100% of the time. I'm saying with a 93% correlation rate that the sum of fWAR is a fairly accurate representation of wins which makes it a good projection model which is where I started when I added up the Cubs ZiPS fWAR totals for the 97 wins and you told me it didn't work that way.

edit: Actually I stand corrected and have edited the post accordingly. The highest variance was your Angels example at 10% meaning that the fWAR to the sum of actual wins was at least 90% accurate as opposed to 93% and yet the correlation is still statistically significant.

 

[DanTown]

Again, I we are speaking of predictions which I am not disputing.
The dispute is that the individualized players war for a given year will equal the team's win total for that given year. This is a discussion on facts not predictions. And the fact shows that the numbers do not add up when speaking of individual teams as examples that I have provided and have gone uncontested.

The numbers are close enough to say that they have a strong relation though. This is like a team having a WAR of 50, adding it to the 48 win number, they win 99 games and you throw out the predictive measure of the idea. The point of adding WAR to 48 wins isn't to get an exact measure of a team but to rather speak to how good the team is.

One of the reasons that WAR doesn't perfectly align is that managers potentially have a strong impact on W-L but they don't impact WAR as much. (I.e pitching matchups, batting order optimization, who does/doesn't play, etc).

 

[beckdawg]

The problem with WAR as a predictor is less to do with WAR and more to do with injury and poor projections. Take the Nationals last season. By all rights they were a WS favorite. But they lost multiple players to injuries and others didn't perform to projections. In a world where you have 100% accurate projections and 100% assurance of no injury WAR would be more predictive in this manner. You still wouldn't have a 1:1 correlation between the two because while you could accurately predict the number of runs scored and given up you wouldn't be predicting them on a per game basis. So, to use someone else's example, you might score 12 in a game and give up 0 and then lose a 1-0 game. However, those are the exceptions and they are relatively few per season.

With all that being said, it's frankly pretty silly to characterize WAR negatively because of this since no matter how you want to talk about future performance there will be issues. For example, we had really no reason to believe Verlander would be shit the past several years. If you want to use ERA, FIP.... whatever the contract he signed seemed like the right move. I think perhaps the best part about baseball is that unlike basketball it's a team based game where one person can't take over and unlike football year to year prediction is slightly more accurate. So, while it may not be 100% predictive, that to me is part of the charm of baseball because it make debates worth having.

 

[brett05]

I acknowledge that there can be a variance on individual teams actual wins vs the sum of fWAR. Going through all 30 teams I found the highest variance to be 10% which means that even in that case the fWAR to actual wins was 90% accurate or, in mathematical terms, statistically significant. I don't think any stat is 100% accurate 100% of the time. I'm saying with a 93% correlation rate that the sum of fWAR is a fairly accurate representation of wins which makes it a good projection model which is where I started when I added up the Cubs ZiPS fWAR totals for the 97 wins and you told me it didn't work that way.

edit: Actually I stand corrected and have edited the post accordingly. The highest variance was your Angels example at 10% meaning that the fWAR to the sum of actual wins was at least 90% accurate as opposed to 93% and yet the correlation is still statistically significant.

Everything you have shared does not make it a good projection model. It makes a reasonable attempt at explaining what has happened, not what will happen.

 

[TC in Mississippi]

Everything you have shared does not make it a good projection model. It makes a reasonable attempt at explaining what has happened, not what will happen.

Fine, how do you project win totals? What model are using for that? In the Cubs example almost every computer model or flat out opinion projects them for 95 + wins. Using WAR as a rough guide you get the same thing. Again how would you project win totals? What equations/stats would you use?

 

[TC in Mississippi]

The problem with WAR as a predictor is less to do with WAR and more to do with injury and poor projections. Take the Nationals last season. By all rights they were a WS favorite. But they lost multiple players to injuries and others didn't perform to projections. In a world where you have 100% accurate projections and 100% assurance of no injury WAR would be more predictive in this manner. You still wouldn't have a 1:1 correlation between the two because while you could accurately predict the number of runs scored and given up you wouldn't be predicting them on a per game basis. So, to use someone else's example, you might score 12 in a game and give up 0 and then lose a 1-0 game. However, those are the exceptions and they are relatively few per season.

With all that being said, it's frankly pretty silly to characterize WAR negatively because of this since no matter how you want to talk about future performance there will be issues. For example, we had really no reason to believe Verlander would be shit the past several years. If you want to use ERA, FIP.... whatever the contract he signed seemed like the right move. I think perhaps the best part about baseball is that unlike basketball it's a team based game where one person can't take over and unlike football year to year prediction is slightly more accurate. So, while it may not be 100% predictive, that to me is part of the charm of baseball because it make debates worth having.

I would argue that those Nationals projections were accurate based on data available. The fact that they didn't accurately predict outcome is irrelevant. The had more injuries than any model could have predicted, their manager actually cost them games as did the trade for Papelbon. They came in 6 games under their Pythagorean win total which means they should have won 90 games which would have tied them with the Mets for the division. Projections aren't crystal balls, outcomes are never assured, they simply tell you what should happen within an reasonable margin of error.

In the end baseball is going to baseball. It's the most unpredictable of all sports because talent usually weighs less in any predictive model you can devise than in other sports, even football which has more players. Predicting team wins is fun because of that unpredictability as you say but it doesn't discount the projections which are data based.

 

[Parade_Rain]

Fine, how do you project win totals? What model are using for that? In the Cubs example almost every computer model or flat out opinion projects them for 95 + wins. Using WAR as a rough guide you get the same thing. Again how would you project win totals? What equations/stats would you use?

With what's happened to all the teams in the NL Central, one could lick his index finger, stick it up in the wind and guess the Cubs should win 95+. No WAR needed!

 

[KingDarude7]

With what's happened to all the teams in the NL Central, one could lick his index finger, stick it up in the wind and guess the Cubs should win 95+. No WAR needed!

105+ *

 

[CSF77]

http://www.vegasinsider.com/mlb/odds/futures/

 

[brett05]

Fine, how do you project win totals? What model are using for that? In the Cubs example almost every computer model or flat out opinion projects them for 95 + wins. Using WAR as a rough guide you get the same thing. Again how would you project win totals? What equations/stats would you use?

I'm not speaking of projecting win totals for next year. I am speaking strictly on using KNOWN numbers and that WAR =/= WINS

 

[brett05]

http://www.vegasinsider.com/mlb/odds/futures/

I can;t click the link at work, but you did not just post win totals or odds to win the World Series as evidence toward any team doing anything did you?

 

[TC in Mississippi]

I'm not speaking of projecting win totals for next year. I am speaking strictly on using KNOWN numbers and that WAR =/= WINS

Well in that case I showed you the numbers. For the entire 2015 season, every team, there was a 1.8 game variance on WAR to actual wins of .075%. The highest variance on an individual team WAR was 10% on your Angels example. If the number in the largest sample size, the totality of MLB, has a .075% variance and in the smallest individual components the largest variance is 10% why you can't accede to the point? In every meaningful measure I've proven the correlation to a minimum of 90%. You keep saying to show facts and I keep showing them to you without success. Do you not believe 90% to be a statistically significant correlation?