THE BOOK--Playing The Percentages In Baseball

Hmmm… well Jeff gave us the league totals right at the bottom: around 68.6 win+loss points per situation. 

So, Bedrosian’s LI is 1.90.

***

Since Mills uses a scale from -1000 to +1000, as opposed to the scale of 0.000 to 1.000, you need to divide by 2, then divide by 1000 to get it into Fangraphs’ scale.  So, the 68.6 win+loss points is .0343 win movements.

If you go to recipe #1:
http://www.hardballtimes.com/main/article/crucial-situations

We see that each situation in 99-02 has a win movement of .0346.

***

If we look at Hoyt Wilhelm’s 1965 seasons (144 IP, 66 G), we see:
67.4 : AL League Win+Loss Pts per situation
77.6 : Wilhelm

That makes his LI 1.15.  In 1964, he had 510 PA, meaning that they were as impactful as 587 random PA.

Compare that to our current top crop of relievers who get an LI of 2.0, who get close to 300 PA, and therefore, are as impactful as Hoyt Wilhelm!

Studes talked about Billy Williams of the Cubs (and WRIGLEY!) here:
http://www.hardballtimes.com/main/printarticle/ten-things-i-didnt-know-last-week40/

I would guess that there is no park adjustments to win probability.  Be careful!

Right, Sagarin didn’t include park adjustments.  Still, I believe there’s something there re: Williams being undervalued by context-neutral stats.

According to b-r.com, which (I think) does park adjust, here are the lg/park OPS numbers for various top players of that ERA:
http://www.baseball-reference.com/w/willibi01.shtml
.730 Billy
http://www.baseball-reference.com/m/mayswi01.shtml
.729 Mays
http://www.baseball-reference.com/r/robinfr02.shtml
.724 Robinson
http://www.baseball-reference.com/c/clemero01.shtml
.724 Clemente
http://www.baseball-reference.com/a/aaronha01.shtml
.722 Aaron

In terms of wins, the difference between Aaron and Billy’s contexts is 0.2 wins per 600 PA.  10,000 PA means 3 wins.  And that’s between top/bottom.  Hardly seems right, unless Wrigley wasn’t WRIGLEY back then.

***

Here’s a simple way to convert OPS+ into wins:
PA * .013 * (OPS/100 - 1)

So, Billy’s career 132 OPS+ with 10,500 PA gives him +44 wins, without clutch, but with park adjustment.

Studes, how much does Sagarin give him (implies with clutch, but without park)?

I see several posts around that try to figure out player’s LI for the non-Fangraphs seasons.  There is a very easy way to estimate it, and it’s in post #1 above.

Here is an exact step-by-step for Willie Hernandez, 1984:
1. Go here: http://www.kiva.net/~jsagarin/mills/al1984.htm
2. Search for your player
3. Add up his win points and loss points and divide by the number of situations (in my case, that’s 30026+12613 divided by 561, or 76.0)
4. Go to the very last line on that page, and figure out the league average (3048554+3035018 divided by 89865, or 67.6)
5. Divide 4 by 3 (76.0/67.6=1.12)

Quiz that same season comes out to 1.51.

It works fairly well for most pitchers, but it doesn’t seem to work well with Willie here.

Anyway, that’s a good way to estimate it, until I finally produce the results.

Justin makes a related post here:
http://jinaz-reds.blogspot.com/2008/01/dirty-way-of-predicting-reliever.html