Win Probability: Game 108

by Eric Simon on Aug 6, 2005 12:36 AM EDT

Here is the Win Probability graph for the Mets' 12-9 loss to the Brewers on Thursday (click to enlarge).

(What the hell is this Win Probability stuff anyways?)

A few notes about WPA:

- WPA stands for Win Probability Added
- A positive WPA indicates a positive contribution towards a victory
- A negative WPA indicates a negative contribution towards a victory (or a positive contribution towards a loss)
- Both teams start with a win probability of .500, or 50% (a 50-50 chance of winning)
- The losing team will have an aggregate WPA of -.500
- The winning team will have an aggregate WPA of +.500

What immediately stands out is Roberto Hernandez and his -1.108 WPA (Win Probability Added). I don't know that I've ever seen a WPA that low. How bad was he? If the Mets were virtually guaranteed a victory (1.000 WP) and one player single-handedly turned that victory into defeat, they would still only have a WPA of -1.000. Roberto Hernandez was 11% worse than that!! Basically, Bert blew much of the game, twice.

A quick note: I've actually been told that it's impossible for a player to have a WPA of less than -1.000 (or more than 1.000), and it does make sense if you consider the difference between one situation and another as a single play. The problem with this game was in the top of the ninth. Technically, Jenkins' single and Weeks subsequently getting thrown out at home should have counted as a single, net negative play (-.045 WPA).

This is where I diverge. I counted this as two plays: one, a -.262 attributed to Hernandez for the RBI single by Jenkins that tied the game at nine. I considered Woodward's throw home to nail Weeks at the plate as a second unique play, a +.218, all attributed to Woodward. As a result, Hernandez gets -.262 instead of -.045 (the net difference between the plays), and is thus able to catapult beyond the -1.000 WPA mark. In any event, I sent an e-mail for some clarification to Studes (from The Hardball Times and Baseball Graphs), who is the de-facto expert on this subject. I will let you know what he has to say about this scenario.

Tweet Comment 3 comments | 0 recs |

Do you like this story?

Comments

math

ES,

I'm too boozed to do any math or make any sense of this stuff.

The chart looks good; can't argue with it.

I gotta tell ya (since we're talking about an old game now), bringing in Bert Hernandez for a third straight game seems like a low-win-probability situation to me. Mets should be 57-52, at least...

by kingcritical on Aug 6, 2005 2:23 AM EDT reply actions

Hernandez/Woodward

I'm far from a WE expert, and I, for one, have always been confused about how WPA is distributed between fielding and pitching. However, it seems to me that it should be regarded as one play, because if the throw home had been late and Weeks had scored, the WPA would have been treated as one event.

by Alex Nelson on Aug 6, 2005 11:03 AM EDT reply actions

FWIW

Studes told me that every play involving a fielder should "correctly" be scored as two separate events:

1) the batter event, with appropriate credit applied/deducted from the pitcher
2) the fielder event, if any credit should be given/deducted from the fielder

In most cases, it is simpler to just record them as a single play, since nothing extraordinary happens 99% of the time. In extreme cases (such as this, I suppose), it is more accurate to record them separately, thus assigning credit/blame more precisely.

Your point is valid. If the throw home were merely late, the play would be treated as a single event and the entirety of the WPA lost would be attributed to Hernandez.

If Woodward's throw would have been on time, but he air-mailed it, let's say, I would probably still record it as two events, since Woodward deserves some (or all) of the blame for the second run scoring.

by Eric Simon on Aug 6, 2005 11:12 AM EDT up reply actions