How to Choose Your All-Stars
6:15 PM EDT Sunday, May 25 2008
The MLB All-Star balloting is open, but before you cast your photo, I'd like to inform you about the most important stat that you should be looking at, one that MLB doesn't show you. If you click on the compare stats link for AL second basemen, you will see something like this:
Now as much as I would love to vote for Robinson Cano, the only players you really can consider here are Jose Lopez, Dustin Pedroia and Ian Kinsler. Based on these misleading numbers alone, you may think that Ian Kinsler is the right choice, but you'd be wrong. If you check out these stats from fangraphs.com, you'll see the right choice, based on WPA, is Dustin Pedroia.
WPA stands for Win Probability Added. How it works is that you can breakdown a baseball game as a finite set of states. The game moves from one state to another, one batter at a time. For example, when the game starts, it is the Top of the 1st, 0 base runners on, 0 out, and the score is 0-0. At this point, each team has a 50-50 chance of winning the game (note that it is not actually 50-50, I'm sure the home team wins more often the road team, but it doesn't matter for the sake of this example). After the first batter, there are 5 possible states the game could be in:
- Top of the 1st, runner on 1st, 0 outs, 0-0
- Top of the 1st, runner on 2nd, 0 outs, 0-0
- Top of the 1st, runner on 3rd, 0 outs, 0-0
- Top of the 1st, 0 runners on, 0 outs, 1-0
- Top of the 1st, 0 runners on, 1 out, 0-0
Notice, by the way, that it doesn't matter how the hitter gets to first base, a hit, walk, error, etc. This is why OBP (On-Base Percentage) is a much better stat than AVG (Batting Average). And also, if the batter is out, it doesn't matter if he strikes out, flies out or hits a line drive. Many people look at the number of times a batter strikeouts, as if having a lot of strikeouts is bad. Striking out isn't good, but it's not any worse than a line out or pop out. The number of strikeouts a batter has is often over-emphasized in my opinion. A guy who has a lot of homeruns, walks and strikeouts is much more valuable than a guy who has a lot of hits, but few homeruns and walks. In fact, a walk is better than a single, because a batter is likely to make a pitcher use more pitches in the process of drawing a walk, versus getting a single on the first pitch. Same goes for a strike out, you are better off striking out on a full count than popping up the first pitch.
Alright, back to how WPA works. Once the game has moved into one of the 5 states, you can recalculate each teams odds of winning. I'll use fictitious numbers to illustrate. Let's say after the first batter, the game moved into state #1 from above. Let's also say that there have been 100,000 games in the history of baseball that have been in that state. And out of those 100,000 games, the road team won 55,000 of those games and the home team won 45,000. In that case, the lead off batter increased his teams' chances by 5%, from 50-100 to 55-100. As I mentioned before, the actual numbers are different, but that's the concept.
So you can keep track of how much a batter increases/decreases his teams probability of winning with each at bat. This running tally is what you see in fan graphs in the WPA column. And as you can see from the stats for AL second basemen fangraphs.com, Dustin Pedroia has done more to increase his team's chances of winning and in fact, Ian Kinsler, despite his good numbers, has had an overall negative impact on his team's chances of winning. How does that happen? Well, it's simple if you think about it. If Kinsler hits a 3-run home run with his team up 8-0, but then the next night strikes out in the 8th inning of a game with runners on 2nd and 3rd with his team down 5-3, what's his overall impact on the team?
Announcers of baseball games inherently know this. For example, as I write this, the Jose Vidro just singled to right center, scoring Ichiro and Jose Lopez, extending Seattle's 3-2 to 5-2. Michael Kay announced it as a big base hit and it was, but Fan Graphs actually gives you a way to measure it. As you can see from the game's Live Play Log, with that hit, the Yankees' chances of winning (a.k.a Win Expectancy, or WE) went from 21.7% to 10.9%.
As you can probably tell, I love Fan Graphs. What surprises me is that major networks that cover baseball like Fox, ESPN, etc. have failed to incorporate WPA and WE into their broadcasts. I think fans would really like this. It could be displayed very similar to the way poker is covered on TV, with each teams chances of winning or losing displayed as each play happens. Instead of the score of the game being shown as NYY 5 - SEA 3, it could be NYY(66%) 5 - SEA(33%) 3. And as each batter comes to the plate, instead of displaying misleading stats like AVG, HR, and RBI, they would show AVG, HR, RBI, OBP, SLG, OPS and WPA. These additional stats give a much clearer picture of how well a given hitter is doing.
Here's an example of how misleading the traditional AVG, HR, RBI stats can be. Compare these two hitters:
AVG HR RBI
Garrett Atkins .338 8 30
Melvin Mora .239 5 23
Obviously Garrett Atkins is having a better season and is therefore more valuable to his team? Well what if I told you Garrett Atkins' WPA is -0.49, whereas Melvin Mora's is +0.63? Now who's having a better season?
