Monday, January 7, 2008

Game Sccore + (Part II)

Part III in my Game Score + research (Part I and Part II). The last post showed that using the extra outs adjustment factor and graphing the results using a pitch per out scenario could provide a necessary adjustment to the Game Score framework that added in the pitch count to make a realistic comparison between Red Bartlett's 58 pitch efficient pitching performance to Kerry Wood's 20 strikeout dominant performance. In conclusion to Part II, I wanted to see how this adjustment factor effected a team's performance over a season and what if I used correcting factor for actual pitches per out rather than using 3?

So the extra out adjustment factor showed to be a good adjustment when you consider the best pitched games but what about a team's performance over a one year period? Looking at the Colorado Rockies in 2007, the following graph indicates how the adjustment factor looks for a team during the season.

Pretty linear (note Aaron Cook's 74 pitch game below the x-axis). The correlation of this is equal to 0.77 while Game Score is equal to 0.46 (data not shown). So this shows for an average pitching staff, a pretty good correlation between pitch count and Game Score+, although there is a greater amount of negative game scores with this adjustment. During the regular season the low Game Score was 4 which translated to -23 GS+ (which wasn't the worst adjustment, this goes to a -65 GS+ pitched in Game 53 by Hirsh who throw 103 pitches in a 4 2/3 effort). So to conclude this adjustment factor shows good correlation to an average team between pitch count and the new metric Game Score+.

The final data crunch is to look at the good pitching performances and the 2007 Rockies using actual pitch per out instead of a constant 3. To review, I took all pitching performances and adjusted all the pitching performances to 9 innings. For those pitchers who didn't go 9 innings, these extra pitches were divided by 3 (assume 3 pitch strikeouts which if you recall is based on the 81 pitch perfect strikeout game) and provided the "extra outs" metric which is subtracted from Game Score. Instead of assuming a 3 for every pitcher then I just used pitcher's actual pitches per out ratio that was established for the innings they did pitch.

The final question is then instead of using 3 pitches to account for each out for pitcher's adjusted 9 inning game then what happens when we use their actual pitches per out ratio they were actually having during their appearance?

Well this method certainly drops the large negative games scores and still holds a pretty good correlation factor of 0.57 versus 0.77 using the 3 pitch per out factor. For the good pitching (data not shown) it drops the the correlation factor to 0.02 versus 0.25. Because of this I think using the 3 pitch factor seems to be the best bet.

Finally if you recall the whole purpose of this exercise was to try and determine if Aaron Cook's 74 pitch complete game was worthy to be considered a great game. If you recall Bill James' Game Score for this game was 67. Using my Game Score + adjustment factor you increase his score to 69. How does this relate to the best performances from 2002 - 2007? Well he barely eeks in to around 120th and doesn't even come close to matching the 2006 performance by Jeff Francis who pitched a complete game Game Score 91 versus Game Score+ of 75 (due his taking 129 pitches or requiring 48 extra pitches or 16 extra outs).