Question about Game Score

[catbox_9]

What is everyone's opinion on the game score statistic that Bill James came up with. I'm sure there is tons of analysis that has already been conducted but I don't have any and was attempting to do it myself. Is it a very good predictor of whether or not a team will win?

[Zagi-CRO]

It could be a very interesting but...
Have you formula or link on his predictions for 2007. ?
I have own formula predicting win/lose and it works in 60-70%...

[Zagi-CRO]

I found this
http://www.tangotiger.net/winshares.htm

[SamtheBravesFan]

It could be a very interesting but...
Have you formula or link on his predictions for 2007. ?
I have own formula predicting win/lose and it works in 60-70%...

What he's talking about is a stat called Game Score, where it tries to quanity how well a pitcher has done in a game. Here's the definition of it from Baseball Almanac:

Start with 50 points. Add 1 point for each out recorded, (3 points per inning). Add 2 points for each inning completed after the 4th. Add 1 point for each strikeout. Subtract 2 points for each hit allowed. Subtract 4 points for each earned run allowed. Subtract 2 points for each unearned run allowed. Subtract 1 point for each walk.

Quite frankly, I don't see how it can predict whether a team is going to win or not. All I think it does is shows how well a starting pitcher did on any given day.

[Tango Tiger]

It's an evaluative stat that adds the pluses and subtracts the minuses. It would be hard for any such stat to not be well-correlated to winning (like OPS, or the multitude of personal toys published on this site).

The question is if it adds up those pieces in an intelligent fashion.

[SABR Matt]

The answer is, no it doesn't.

It makes no attempt to view the pitcher in terms of his defense-neutral real self, and I'm not sure I follow why going beyond the fourth inning should be essentially double counted...I mean naturally it's good to go deep into games, but I don't see how it benefits us to double count the extra innings.

[pizzacutter]

Catbox, like a lot of Bill's (and everyone else's) early stuff, it's been replaced by more sophisticated measures. Matt already pointed out a few flaws, but Game Score does have its usefulness. It's pretty easy to calculate, and it gives an idea of how well a pitcher did independent of his run support and his bullpen. It's certainly better than W/L. I haven't seen any studies on how well it predicts winning, but those might be out there.

[catbox_9]

The reason I ask about predicting team W/L is because I have attempted to do that myself. I'm not good with using computer programs to analyze data so I have to do it manually which is rather tedious. I have taken statistics classes so I know how to calculate margin of error and what not. I've looked at about 500 games so far and there appears to be significant evidence that if a pitcher records of score of say 39 or less his team is very likely to lose while a score of 61 or more gives him a good chance of winning. A score of 40-60 gives fairly mixed results and a very high % of those games end up as no decisions for the pitcher.

[Tango Tiger]

You mean in games where a starting pitcher gives up 5 runs or more, they are likely to lose, and if he gives up two runs or less, they are likely to win?

And, I'm pretty sure that the team that scores runs than they allow are "predicted" to win 100% of the time.

Predictions are based on what is unknown, given what you know. If you want to calculate the game score through 5 innings, and then tell me that that will be more predictive than just the actual score through 5, I'll listen.

Otherwise, what you are talking about is not "predictions" but "correlations".

[catbox_9]

Correlation is probably what I'm looking for. Obviously if a pitcher gives up 5+ runs they're likely to lose, but I'm looking to see hom much more likely are they to lose if they were to give up 4 runs vs. 5 runs, etc.

[Tango Tiger]

I suggest either:

http://www.tangotiger.net/innwin2.html
http://winexp.walkoffbalk.com/expectancy/search