Remember me
Forgot your username/password?
 
Print
Home>Articles

Why 52?

By Bill James
May 7, 2008

Revisiting a Win Shares Issue

 

            I recently received a question from a reader in the “Hey, Bill” section:

 

I'm reading the Win Shares book again and am wondering if you (or someone) has solved the "Why 52" issue about the relative value of a run saved or added. 

 

            Trailbzer

 

For those of you who might miss the reference, this refers to the decision, in designing Win Shares, to allocate 52% of the credit to pitching and defense, 48% to hitting and baserunning.   I wrote an article saying, in essence, “Yes, I know this doesn’t make sense, but it seems to be necessary.   I’ve tried and tried to make it work with a 50/50 split between offense and defense, and it just doesn’t work.”

 

It turned out that this was the result of a simple and stupid error on my part, having to do with how I handled pitchers’ hitting.   I originally set up the analysis so that 50% of the credit would go to run creation, 50% to run prevention.   However, hitting by pitchers was kind of a leftover, and I decided to give credit for pitchers’ hitting as if it was part of their pitching.   In other words, suppose that there were two pitchers who were otherwise the same, except that one of them created five more runs than the other as a hitter.  I decided that “five runs added as a hitter is the same as five runs saved as a pitcher”, so I simply adjusted their runs saved as a pitcher for their runs created as a hitter.  

 

On a certain level this was not illogical, but what I failed to see until a year or two after I finished the Win Shares book was that this had screwed up the 50/50 split.   I knew that SOMETHING was screwing up the 50/50 split, but I couldn’t figure out what it was.   Pitchers’ hitting is a significant factor in the game. . .something like 3-4% of the game, when pitchers hit.  When you move that 3-4% to the wrong side of the ledger you have too many parts on that side of the ledger.

When I started designing Win Shares and Loss Shares I went back to the beginning, and I handled pitchers’ hitting in an entirely different way.   Early on in the design of Win Shares and Loss Shares I realized that, using 52% for pitching and defense, I had too many Win Shares going to pitching and defense, and I needed to move it back to 50%.   So then I asked myself, “Well, what am I doing different here than I was before, that caused this problem to go away?”

 

Just a silly design mistake on my part.  

 
 

COMMENTS (6 Comments, most recent shown first)

studes
I think you can justify the 52% -- in fact, you can justify a higher figure -- because of the fact that Win Shares starts with an average team baseline. When you start from average (runs allowed and runs scored equal at the league average), a run saved is worth more than a run scored. It's easy to play with the pythagorean formula to see why.

Also, I do believe that including pitchers' batting in the pitching Win Shares instead of batting skewed some of the results. At THT, we include them in batting Win Shares.
5:31 PM May 13th
 
ChristopherTaylor
This brings up a question that has crossed my mind from time to time (admittedly not very often because it isn't very good and as an AL fan, I naturally believe NL ball an inferior brand of baseball) but is there any way to measure whether or not a pitcher (or a glove-first SS like John McDonald, which is the AL equivalet that comes to mind) hitting contributes more than one would expect given the components of his statistics. This isn't very clear, sorry, but with pitching, it has been well worked out (via DIPS theory) how the various components of a pitchers line go into his adjusted pitching line. Has this been worked out for batters? Also if it has, do the components of a hitter who is thought to be an "automatic out" not reflect hot terrible he is at the plate. I expect that, like the pitcher, who gets a lot of luck (otherwise known as aid from his defense) will do better as defense will concede things such as the double down the line, play the OF in and so on. It seems like there should be a sweet spot there too, for a player who is bad enough to induce the team to make adjustments, but to be good enough to exploit these weaknesses. This is an analogous situation to that of David Ortiz trading batting average for slugging points (not a great trade for him).

Thanks... and please keep up the good work, I've been reading your work since I can remember. As an 8 year old, I begged my parents for an abstract. You've influenced my thinking, not only about baseball, but in my life as well (I'm a scientist in Canada) and developed a love for baseball where all spots but Hockey are anathema.
4:32 AM May 10th
 
ChristopherTaylor
This brings up a question that has crossed my mind from time to time (admittedly not very often because it isn't very good and as an AL fan, I naturally believe NL ball an inferior brand of baseball) but is there any way to measure whether or not a pitcher (or a glove-first SS like John McDonald, which is the AL equivalet that comes to mind) hitting contributes more than one would expect given the components of his statistics. This isn't very clear, sorry, but with pitching, it has been well worked out (via DIPS theory) how the various components of a pitchers line go into his adjusted pitching line. Has this been worked out for batters? Also if it has, do the components of a hitter who is thought to be an "automatic out" not reflect hot terrible he is at the plate. I expect that, like the pitcher, who gets a lot of luck (otherwise known as aid from his defense) will do better as defense will concede things such as the double down the line, play the OF in and so on. It seems like there should be a sweet spot there too, for a player who is bad enough to induce the team to make adjustments, but to be good enough to exploit these weaknesses. This is an analogous situation to that of David Ortiz trading batting average for slugging points (not a great trade for him).

Thanks... and please keep w
4:29 AM May 10th
 
JeffreyWalters
Does this data support the phrase "offense wins games - defense wins championships"?
1:09 AM May 8th
 
Trailbzr
Hey, Bill. This is Troublmkr, er, Trailbzr. While you were writing the above article, I was contributing my own $.02 to the Reader Posts board, suggesting a way to estimate the relative value of an offensive and defensive run, and coming up with an answer about 48/52.

I had thought the issue was that a run prevented increased the marginal value of each run by making them more scarce, while a run scored does the opposite. Using the Pythagorean method in a league that scores five runs a game, a team that scores an extra run a game should win at 36/(36+25)=.590, while a team that prevents a run a game would win at 25/(25+16)=.610.

SHOULDN'T a run prevented have more marginal win value than a run scored?


6:07 PM May 7th
 
wavydavy
Makes sense. Presumably, runs created and runs prevented should balance out (usual double-entry bookkeeping math).

On a similar note, why do you attribute ZERO credit/debit to pitchers' fielding stats? Just in general, this surely has SOME effect on game outcome unless you assume that pitchers never perform on defense (obviously not true).

More specifically in the real world, are you saying that having Greg Maddux or Bob Gibson on the mound has no effect on a team's defense? Even more to the point, and more recently, did you see the 2006 World Series? I cannot believe that you can still say that pitchers' defense has no value (positive or negative) after watching the Tigers' pitchers act like Little Leaguers.
11:27 AM May 7th
 
 
©2024 Be Jolly, Inc. All Rights Reserved.|Powered by Sports Info Solutions|Terms & Conditions|Privacy Policy