Wednesday, April 8, 2009

I'm still giddy

Fan Graphs has some very cool stuff.

Their signature item is the "Win Probability" chart. They do this for every game every day in the Scoreboard section. I first came across the idea of "Win Probability" or "Win Expectancy" in Jim Albert's and Jay Bennett's 2001 (revised 2003)book Curve Ball: Baseball, Statistics, and the Role of Chance in the Game.

In chapter 10, "Measuring Clutch Play," they present Table 10-14 on page 291. The caption reads Probability of Home Team Victory Given the Score Difference (Home Team Minus Visiting Team) at the End of Each Inning.** They give most of the credit for the original research to a Canadian fellow named George Lindsey, and cite his 1961 paper "The Progress of the Score During a Baseball Game" in the American Statistical Association Journal. This idea has been taken up by several researchers and sabermetrically-inclined fellows, and I'm sure the original chart has undergone numerous improvements and revisions. Much discussion can be found following the links on FanGraphs. (Here's one.) Suffice to say, this view of the game is gaining traction, and like other sabermetric notions, is creeping into mainstream baseball talk. I won't bury you with math, I promise. But I'm giddy about the start of the season and the excitement of our big win yesterday, so I'm wallowing in all possible post-game enjoyment. I mean, when's the last time you saw a GRAPH with a trend line going up and ending with a GIANTS WIN? You don't have to be a nerd to think that's cool.

**You can actually see Table 10-14 by using Google Books preview feature. Go to Ch. 10 and click on the p. 291 link. You'll have to scroll around a bit. This is a cool thing but the interface seems clunky to me. (I tried to scan the page from my book and post it--way old school tech, I know--but it came out too faint.) Be sure you are looking at the revised edition. If I come across another source of this type of chart on the 'net I'll be sure to blog about it!


Zho said...

According to this graph, the probability of a Giants victory was 100% before the ninth inning started. Interesting.

M.C. O'Connor said...

You have to look closely. The win probability actually takes a hit in the 9th (Hinshaw), then bounces back up, and you can see a very small upward-trending tail at the end (presumably out no. 3).

(My inserted graph is smaller than the actual one. q.v.)

The win probability at home with a 5-run lead at the end of the 8th is about 99.6%.

Zo said...

I take it that this does not apply to teams that use Matt Herges as a closer.