Wednesday, April 19, 2006

Win Probability Charts

Finally, a web site is going to publish win probability graphs for every game played, daily. As the name implies, the graphs plot the percent chance of each team winning after every PA of a particular game. The site also sums up the win probability (WPA) added for each player during the game. The site is and the exact link can be found below:

Fangraphs Win Probability Page

For a more detailed description of Win Probability, check out this Hardball Times article:

The One About Win Probability

In this article from HBT, Dave Studeman warns not put TOO much weight on WPA and I agree. Within the context of a single game, a HR hit in the 9th in a tie game is about 4 times more valuable than one hit in the first. Does that mean we should give 4 times the credit to the player for hitting the 9th inning HR? I guess it depends on what that "credit" is supposed to represent. I see WPA as a measure of impact a player has on a game more than a measure of actual ability. People inclined to believe in clutch hitting and hitters would strongly disagree. I think we all can agree that they are very interesting to look at, however.

From the comment section:

but what is the point of these things?

Let me explain in a little more detail. It shows you two things:

1. The probability of either team winning at any point in the game. This is the "win probability" (WP).

2. The amount the probability changes for each baseball event (walk, single, HR, etc...). This is the "win probability added" (WPA).

Like I tried to briefly explain above, not all baseball events have equal value with regard to their impact on what team end up winning. Let's look at two very different situations:

1. A solo shot in the 9th while down 10-0:

The win probability added for situation 1 is practically zero; They had almost no chance of winning before or after the HR. We would say that the WPA given to the player who hit it was basically 0.

2. A grand slam in the 9th with 2 out, home team down 3-0.

For situation 2, the home team had a very small chance of winning before the HR, and 100% chance of winning after (walk off HR). As a result, the WPA of that HR was huge. We can use the "Win Expectancy Finder" link on my blog to estimate the WPA for that HR. The finder shows the number of times teams won in different situations from 1979-1990 and 2000-2004. In the 149 games where situation 2 occurred, the team behind won only 9.4% of the time. The WPA for hitting that grand slam would be:

100% - 9.4% = 90.6%

That's the highest value a single event can have. The sample is still pretty small and the actual run environment isn’t being considered, so there's definitely some uncertainty there. Still it's still a nice ballpark estimate and you can find some suprising situations if you look hard enough.


Anonymous southwest brewer said...

rluz - i've been checking out the fangraphs site since you mentioned it somewhere and these little daily game graphs have grabbed my attention.

I usually look at the Brewer graph and it goes up and down depending on who's making runs/hits/whatever....but what is the point of these things? I don't quite get what they are used for?

It always ends up at 100% when some team wins with a bunch of jags in between depending on who takes the lead. Do they add up to something over the course of a season?

2:19 AM  
Anonymous southwest brewer said...

Takes the runs/outs/innings/baserunners and compares to the past, right? OK so it's taking the game situation and comparing it to how many times teams in that same position won. If that's it, pretty cool.

Now if we can convince the Brewers to start getting that 100% mark at the end of games.

2:05 AM  

Post a Comment

Links to this post:

Create a Link

<< Home