THE BOOK--Playing The Percentages In Baseball

And here’s everyone’s favorite former Mariner, Eddie Guardado:

http://www.fangraphs.com/statss.aspx?playerid=727&position=P

When he enters the game, his LI is 1.53.  But, on a PA-by-PA basis, he shoots up to 1.96.  When he leaves the game, he’s at 1.80, and that’s happened 16 times in 42 tries.

Remember, there’s a whole bunch of reasons that the LI can go up or down.  It’s not necessarily based on success.  For example, bases loaded 0 outs, tie game, bottom of the 9th, the LI is 2.4.  But, if the pitcher gets an out, the LI shoots to 5.4.  And, if there are now 2 outs, the LI is 6.4!  (And, if the pitcher gets that third out to send it to extra innings, the LI is reset back to 2.4.) Rather than taking pressure off, the outs are actually creating pressure!

Why is that?  Because, with the bases loaded and no outs, the expectation is that the game is close to being lost.  The home team has a 90% chance of winning here.  But, once they get the out, that runner on 3b is not so sure to score.  And with 2 outs, the game is more likely to go to extra innings than it is to be won right there.

Tango,

I noticed at fangraphs that they said they used your charts to calculate WPA. I was wondering if you had any plans in the future to come up with specific charts for each year/league?

I put some of the pitching statistics for 2005 in a spreadsheet and noticed that the average pitching stafff had +6.2 WPA wins.

Yup, that would be based on the 5.0 RPG environment.  I told David at Fangraphs that I’ll probably come up with separate charts from 3 to 6 RPG, so he can use them as appropriate.  (Say, a 4.7 RPG would be 70% 5.0 and 30% 4.0) I might do them in increments of 0.5 instead.  I don’t know yet.  The earliest I can look at this is after the season ends.  Unfortunately, things like paying my bills takes precedence.

As The Book shows, the win value of an out or a K is around .026 to .027 wins.

Go to Fangraphs, and take a random hitter or pitcher, and a random year.  Do this calculation for a hitter: LA / (AB-H)
where LA = loss advancement, the -WPA at Fangraphs
And for a pitcher do:
WA / (IP*3)
where WA = win advancement, the +WPA at Fangraphs

What do you get?  Is there much difference based on the quality of hitter or pitcher?

This is just something that occured to me. Some of the debate about the worthiness of WPA is that the leverage of the late inning situation is not independent of that of the early inning situation. (This is one reason why the best pitchers are usually starters instead of relievers). So, what about crediting each run its average contribution to wins? From what I recall, the first 5 runs have around the same win value, and then it starts to decrease. So, a guy who drives in the 2nd run in a 2-1 win would get the same credit whether he did it in the 1st or 9th inning--which is what many people think should happen.

If win values were given out based on the sequence of when they occured, then the second run in a 2-1 game would be equal to that of a 9-2 game.

The average WEA/WPA is set at 0 wins right?  I thought so, but just wanted to make sure.

If you were a GM, how would the exLI help you?  I see how regular LI (or PA LI) and how gmLI can help a GM, but what about exLI?

Well it now looks like theirs two ways of doing this...the WA-LA or setting WEA/WPA at 0 as average.

Chris, you can use my spreadsheet to derive win values for different run environments.  It’s pretty easy to create a master list of values for different run environments for each league, even for each league/park.

Mike, WPA is, by definition, WA-LA.  There is no “setting” anything to average.  WPA is, by definition, wins above average.