OOTP Developments Forums - View Single Post - Good Performances By Bad Pitchers, Revisited

holes573 · 11-17-2024, 12:39 AM

I recently posted data suggesting that bad starting pitchers may be programmed to have good days too often (https://forums.ootpdevelopments.com/...d.php?t=358515).

Two key questions arose among some of the responders to that post:

1. Does OOTP really program pitchers to have good/bad days?
2. Are there better metrics for measuring the quality of pitchers in my analysis?

Let me address each of these questions with new data I just collected, playing the SF Giants in 1965 (OOTP25). I analyzed every inning a starter pitched against the Giants (and for good measure, I did the same for starts against the NY Mets and Houston Astros).

1. Does OOTP really program pitchers to have good bad days?

This is a challenging question because you need to distinguish between good performances that are just due to luck (chance) from those that are due to a programmed good performance.

The key to distinguishing these is that eventually, luck will often run out after a while, during the game. After a lucky pitcher does well in the first inning (such as allowing 0 ERs), they have to continue to be lucky in the second, third, etc. to have a good performance. Each new inning provides an opportunity for the pitcher to lose their luck and start allowing runs. In contrast, a pitcher programmed to have a good day should keep delivering good innings until they tire.

Among the 418 starter performances in my data, in 73% of the outings, 0 ERs were allowed in the first inning. Separately, I found that, regardless of what happened in the first inning, 81% allowed 0 ERs in the second, 78% allowed 0 ERs in the third, and so on. If good performance is due exclusively to luck (chance), then a start that has 0 ERs in both the first and second innings should happen 59% of the time (73% x 81%). And, allowing 0 ER in all three of the first innings should happen 46% of the time (73% x 81% x 78%). Below on the left are the these percentages for consecutive 0 ER innings, for each set of innings. If luck explains all of the good performances of 0 ERs over these innings, these are the percentages we should find when we actually start counting the times consecutive 0 ER innings happen.

But, that isn't what we found. The percentages on the right are what actually happened - how often our 418 starts had consecutive 0 ERs from the first inning on.

0 ERs for first: Luck Only Actual Data

1 inning ...............73%......... 73%
2 innings .............59%......... 58%
3 innings .............46%......... 46%
4 innings .............35%......... 36%
5 innings .............28%......... 32%
6 innings .............24%......... 28%
7 innings .............20%......... 27%
8 innings .............17%......... 27%

Starting about the 5th inning, the actual frequency of consecutive 0 ER innings starts outpacing the predictions based only on luck. In other words, some pitchers strung together long starts of 0 ERs, well beyond what chance/luck would predict. And, for those of you familiar with the concept, these differences are highly "statistically significant" - we should be 99.8% confident that the actual percentages are reliably higher than the luck only percentages. (BTW, these differences happened when facing a good team, like the Giants, and against a bad team, like the Mets.)

These seem like pretty strong data showing that the occasional good performances by pitchers happen more frequently than they naturally would - thus supporting the conclusion that OOTP has, as part of its complex algorithm, a good day/bad day variable at work.

2. Are there better metrics for measuring the quality of pitchers in my analysis?

Last time, I showed data that bad pitchers had essentially the same frequency of good performances as medium and good quality pitchers.

To do that analysis last time, I divided starters into good, medium and bad quality groups based on their 20-80 rating based on scouting. There were suggestions that better ways to do this would be with their 20-80 ratings when they are 100% accurate, or their Projected WAR ratings on the editor. I did the analysis this time with both of these better ratings systems, and found essentially the same results for both. For simplicity, I've presented below only one of these analyses (the one based on Projected WAR).

Here are the frequencies of starters having a "good day" (defined as having a 2.0 ERA or less for the whole outing, that is, allowing only 1 or 0 ERs for their whole outing).

Bad pitchers (1.8 Proj. WAR or less): 28%
Medium pitchers (1.9 - 3.1 Proj. WAR): 32%
Good pitchers (3.2 Proj. WAR or more): 35%

Like last time, the differences between these three groups are not statistically meaningful, so we should treat them as essentially the same. And, that same conclusion (they are essentially the same) held true when I look those starters facing SF, NY and Hou individually.

Also, like last time, I am surprised that so many Bad Pitchers are having good days. My feeling is that Bad Pitchers should have considerably fewer good days than the better pitchers.

Sorry for the data-heavy post, but the only way I think this can be adequately understood is to go to the numbers.

11-17-2024, 12:39 AM	#1
holes573 Bat Boy Join Date: Mar 2017 Posts: 17	Good Performances By Bad Pitchers, Revisited I recently posted data suggesting that bad starting pitchers may be programmed to have good days too often (https://forums.ootpdevelopments.com/...d.php?t=358515). Two key questions arose among some of the responders to that post: 1. Does OOTP really program pitchers to have good/bad days? 2. Are there better metrics for measuring the quality of pitchers in my analysis? Let me address each of these questions with new data I just collected, playing the SF Giants in 1965 (OOTP25). I analyzed every inning a starter pitched against the Giants (and for good measure, I did the same for starts against the NY Mets and Houston Astros). 1. Does OOTP really program pitchers to have good bad days? This is a challenging question because you need to distinguish between good performances that are just due to luck (chance) from those that are due to a programmed good performance. The key to distinguishing these is that eventually, luck will often run out after a while, during the game. After a lucky pitcher does well in the first inning (such as allowing 0 ERs), they have to continue to be lucky in the second, third, etc. to have a good performance. Each new inning provides an opportunity for the pitcher to lose their luck and start allowing runs. In contrast, a pitcher programmed to have a good day should keep delivering good innings until they tire. Among the 418 starter performances in my data, in 73% of the outings, 0 ERs were allowed in the first inning. Separately, I found that, regardless of what happened in the first inning, 81% allowed 0 ERs in the second, 78% allowed 0 ERs in the third, and so on. If good performance is due exclusively to luck (chance), then a start that has 0 ERs in both the first and second innings should happen 59% of the time (73% x 81%). And, allowing 0 ER in all three of the first innings should happen 46% of the time (73% x 81% x 78%). Below on the left are the these percentages for consecutive 0 ER innings, for each set of innings. If luck explains all of the good performances of 0 ERs over these innings, these are the percentages we should find when we actually start counting the times consecutive 0 ER innings happen. But, that isn't what we found. The percentages on the right are what actually happened - how often our 418 starts had consecutive 0 ERs from the first inning on. 0 ERs for first: Luck Only Actual Data 1 inning ...............73%......... 73% 2 innings .............59%......... 58% 3 innings .............46%......... 46% 4 innings .............35%......... 36% 5 innings .............28%......... 32% 6 innings .............24%......... 28% 7 innings .............20%......... 27% 8 innings .............17%......... 27% Starting about the 5th inning, the actual frequency of consecutive 0 ER innings starts outpacing the predictions based only on luck. In other words, some pitchers strung together long starts of 0 ERs, well beyond what chance/luck would predict. And, for those of you familiar with the concept, these differences are highly "statistically significant" - we should be 99.8% confident that the actual percentages are reliably higher than the luck only percentages. (BTW, these differences happened when facing a good team, like the Giants, and against a bad team, like the Mets.) These seem like pretty strong data showing that the occasional good performances by pitchers happen more frequently than they naturally would - thus supporting the conclusion that OOTP has, as part of its complex algorithm, a good day/bad day variable at work. 2. Are there better metrics for measuring the quality of pitchers in my analysis? Last time, I showed data that bad pitchers had essentially the same frequency of good performances as medium and good quality pitchers. To do that analysis last time, I divided starters into good, medium and bad quality groups based on their 20-80 rating based on scouting. There were suggestions that better ways to do this would be with their 20-80 ratings when they are 100% accurate, or their Projected WAR ratings on the editor. I did the analysis this time with both of these better ratings systems, and found essentially the same results for both. For simplicity, I've presented below only one of these analyses (the one based on Projected WAR). Here are the frequencies of starters having a "good day" (defined as having a 2.0 ERA or less for the whole outing, that is, allowing only 1 or 0 ERs for their whole outing). Bad pitchers (1.8 Proj. WAR or less): 28% Medium pitchers (1.9 - 3.1 Proj. WAR): 32% Good pitchers (3.2 Proj. WAR or more): 35% Like last time, the differences between these three groups are not statistically meaningful, so we should treat them as essentially the same. And, that same conclusion (they are essentially the same) held true when I look those starters facing SF, NY and Hou individually. Also, like last time, I am surprised that so many Bad Pitchers are having good days. My feeling is that Bad Pitchers should have considerably fewer good days than the better pitchers. Sorry for the data-heavy post, but the only way I think this can be adequately understood is to go to the numbers.