I've been tinkering around with a pitching statistic for starters, and I think I'm ready to share it. It's a new take on an old stat called Game Score.
Game Score was developed by Bill James to give a single number to a pitching performance that should tell you how they did. I like the stat; it bases everything around 50, and naturally caps at right around 100. Here is the formula:
Start with 50 points. Add 1 point for each out recorded, (or 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.
Thanks to Baseball-Reference for the link.
While Game Score does give a pretty accurate guess as to how a pitcher did (the leaders yesterday were Shark and Kershaw at 86 each; Volquez' 24 brought up the rear), I had to distinct issues with it:
a) It does not penalize short outings. You can strikeout the first batter and end with 51 points, or go 6 innings, allow 3 runs, and end up with 46.
b) It counts things that are out of the pitchers hands (earned runs, runs, and hits).
DiPS theory teaches us that hits come in more-or-less a random distribution that we can control only slightly. Thus, the pitcher's burden is just to get outs and reduced the number of balls in play or over the fence. It is with this in mind that I set to re-write Game Score.
First, I got the latest run expectancy matrix and calculated how many expected runs are saved, on average, by striking a player out (increasing outs by 1 with no station changes). The answer was .457. Then, I calculated the run expectancy of a walk (increasing the relevant stations by 1 with no out changes), and the value was .523. Lastly, I changed the bases-clearing, run-scoring change of the HR, and got a value of 1.77. I came to these conclusions:
A HR was roughly 3 times more valuable than a walk. A walk was slightly worse than a strikeout was good. The values roughly went like this: K – 3 points, BB – -4 points, HR – -12 points. This kind of makes sense: If you gave up a HR, then struck out the side, you had a bad inning, but not an awful one – at least you minimized the balls in play. Of course, it's possible you actually gave up 6 runs in that inning with a series of base hits, but that is somewhat attributable to luck.
After that, I made a judgement call on how many points to give to outs. If a strikeout was worth 3 points, a regular out seemed to be worth 2 to me. This part could definitely be changed (and may be, to 3 points).
The only thing left was to make sure that the average SP performance was roughly 50. I took my numbers and assigned the average performance last year to them (18 outs, 5 strikeouts, 2 walks, .68 HR). That came up as 35 points, so I added a baseline of 15 points.
The problem with this new stat is that there is no natural line to prevent a pitcher from getting over 100 points. Kerry Wood's 20-K CG SHO is 129 points, in this system. In my defense, I don't care so much about this.
Which system is better? I'm not really sure. Today, I'm going to include every pitcher's game score under the old system and the new. You can decide for yourself which is better.
| Regular | DiPS | Delta | STDEV | |
| Kershaw | 86 | 90 | -4 | |
| Samardzija | 86 | 86 | 0 | |
| F. Hernandez | 80 | 81 | -1 | |
| Cueto | 74 | 64 | 10 | |
| Cain | 71 | 71 | 0 | |
| Sale | 71 | 78 | -7 | |
| Chacin | 67 | 49 | 18 | |
| Kennedy | 66 | 77 | -11 | |
| Verlander | 66 | 58 | 8 | |
| Weaver | 66 | 55 | 11 | |
| Anderson | 63 | 59 | 4 | |
| Niese | 60 | 47 | 13 | |
| Shields | 58 | 57 | 1 | |
| Burnett | 54 | 63 | -9 | |
| Lester | 54 | 58 | -4 | |
| Worley | 46 | 56 | -10 | |
| Wainwright | 42 | 69 | -27 | |
| Hudson | 39 | 26 | 13 | |
| Gallardo | 37 | 26 | 11 | |
| Hamels | 37 | 20 | 17 | |
| Sabathia | 36 | 44 | -8 | |
| Volquez | 24 | 33 | -9 | |
| 58.31818 | 57.59091 | 0.727273 | 11.17046 |
Comments
I always like new things. Do you feel like certain pitchers will have better scores in DiPS than Game Score?
WaLiQuote Reply
Groundball pitchers that work in the zone will probably benefit the most.
MylesQuote Reply
Nice work Myles. I’d be particularly interested in the big variances. Chacin going from well above average to just below average (or Wainright’s huge jump) are particularly noteworthy. It’s in these variances that you’ll be able to really determine what each system rewards and punishes. Was Wainright REALLY that unlucky?
26.2cubfanQuote Reply
Very true. Let’s take a look at each of those lines:
Wainwright: 6 IP, 11 H, 4 R, 3 ER, 0 BB, 6 SO, 0 HR
He had 12 non-K outs and 11 hits. That’s a BABIP around .500. It seems like that was unlucky; however, DiPS theory is only mostly true. If you or I were out there, our BABIP as a pitcher would probably be around .600 or something. It looks to me like Wainwright was pretty unlucky yesterday, maybe he also was struggling with stuff a little bit.
Chacin: 6.2 IP, 3 H, 1 R, 1 ER, 3 BB, 6 SO, 1 HR
Chacin had 14 non-K outs and only 3 hits, 1 of them a HR. That’s a BABIP of .143. He certainly got lucky, and gave himself 3 walks as additional opportunities to get burned. I think Chacin is maybe punished a little much, but his performance doesn’t scream “well above average” to me.
MylesQuote Reply
I didn’t realize that you lived in Slovakia, WaLi
http://www.theregister.co.uk/2005/01/31/avalanche_escape/
BerseliusQuote Reply
I’m glad someone else is thinking about this.
Note:
1. Those coefficients are similar to FIP.
2. See Version 3:
http://www.fangraphs.com/blogs/index.php/game-score-and-crowd-sourcing/
Anyway, I’d love it if you were to track the four versions, and see how well each compares to what your “eyes” see, espcially in those cases where there’s disagreement among the 4 versions.
TangotigerQuote Reply
I LOVE YOUR NEW SYSTEM! IT IS BETTER THAN “CATS”!
11111 rec’d 1!!111 LSA!!! 111
Adam WainwrightQuote Reply
Bullshit.
Gustavo ChacinQuote Reply
Meh.
Jeff SamardzijaQuote Reply
Yes, but how does this correlate to Phil Rogers’ +/- system?
Aisle424Quote Reply
@ Berselius:
Yeah. Funny story – I didn’t even realise I was stuck in an avalanche. I just thought it would be a good idea to get drunk while stuck in traffic.
WaLiQuote Reply
@ Myles:
You grossly overestimate my ability to avoid contact. I’m guessing I’d have a line around:
2/3 IP, 0K, 6BB, 18RA, 14 earned (outfielders got winded chasing frozen ropes), 4HR, 15H. Only outs on line drives hit directly at outfielders. Pulled with bases loaded.
Looks like my BABIP finishes up around .867. Sounds about right. All of this, of course, would be against rookie ballers. Thus marks the end of my Moonlight Grahmesque career in pro ball. Never got an at bat.
26.2cubfanQuote Reply
Suburban kid wrote:
Why not just have the day off after the first series? The math works out the same. I know the reason is “maximize profits” since Opening Day probably scores very high ratings, so they effectively have two of them. It’s just annoying, though. Not the end of the world, but annoying.
joshQuote Reply
Tangotiger wrote:
That’s a very interesting article.
I’m curious now as to how FIP came about those coefficients (I’ve been searching half-heartedly for a reason, but my links keep turning up DOA). In a perfect world, I would have weighted the 24 gamestates against the probability of those states occuring; I’d be willing to wager the bulk of the difference of the two methods is due to my quick-and-dirty version simply being inelegant compared to a more nuanced approach.
That being said, I have two thoughts.
1. I’m slightly discouraged that you “got there” first.
2. I’m pretty encouraged that I was thinking at least slightly on the same lines.
This makes me more confident than I was before that my system is at least an improvement over the vanilla score. I’m going to think about the other systems and see where that gets me.
As for tracking it throughout the season, I’d like to be able to do that, but it seems to be an arduous undertaking. Perhaps if I was a better coder, I could pull the data somewhere and make all the calculations automated; afterwards, I’d still probably need to crowdsource the debate much like you did.
MylesQuote Reply
http://www.insidethebook.com/ee/index.php/site/comments/tangos_lab_deconstructing_fip/
TangotigerQuote Reply
This makes me want to draw funny pictures! WOOT!
joshQuote Reply
@ Tangotiger:
Funny, I just came here to post that same link.
dmick89Quote Reply
My workfilter eats that link, I’ll check it out when I get home
MylesQuote Reply
@ Myles:
I find it funny that you can’t access that site but your work doesn’t mind you visiting this place.
Suburban kidQuote Reply
@ Suburban kid:
Even more funny when you learn he works from home and sets his own filters.
joshQuote Reply
@ Suburban kid:
Let’s try to make Myle’s work consider this NSFW with our comments.
JonKneeVQuote Reply
Dirty Anal Sex Viral Video
JonKneeVQuote Reply
@ JonKneeV:
Whores gushing anal cancer rimjob pics
joshQuote Reply
I have problems.
joshQuote Reply
josh wrote:
Hemorrhoids?
RynoQuote Reply
@ JonKneeV:
boobs + dicks + 2013 Cubs. Two dirty words and a reason to kill yourself, I’ve done my part.
MJKQuote Reply
Suburban kid wrote:
Of course it’s useful knowledge. It’s like pinch hitting for Rizzo or Castro because they’ve looked horrible in their last 2 at bats. Same thing. It’s believing in hot hand/cold hand without evidence.
Same thing as leaving a starter in because he’s looked so good. It would be like getting 6 shutout innings from Feldman and expecting him to pitch the 7th with equal success because “he’s on today”. That’s what too many managers think and they get burned for it. It’s not a commendable action to base your expected performance on what happened the previous innings in that game.
dmick89Quote Reply
MJK wrote:
Not sure which is which.
RynoQuote Reply
If by calling it the right decision to remove Marmol, you mean there are better relievers, fine. Then we have to say that the manager corrected his own wrong decision and I’m not sure he gets bonus points for that.
If the closer is the best pitcher, you only hurt your chances to win the game by removing him. Because the next pitcher obviously wouldn’t be as good.
dmick89Quote Reply
@ dmick89:
Whatever the optimal decision is, Sveum would only make it on accident anyway.
joshQuote Reply
Shitty new comic posted
joshQuote Reply
dmick89 wrote:
If hitter ABs lasted three outs, the same as most RP’s appearances, and Castro or Rizzo were swinging at balls in the dirt and over their heads for the first two outs, I’d conclude they are “cold” and pinch hit for them before the 3rd out.
But I agree with your other point. I did not think Marmol was the best pitcher. Maybe if I did, I would think differently.
Suburban kidQuote Reply