Sunday, 1 June 2008

Why You Don't Intentionally Walk the Bases Loaded

The Jays made the same mistake twice--choosing to follow inherited baseball wisdom rather than going with cold hard statistical probability--and ended up losing this series with the Angels because of it. BJ Ryan has been top steppin' for most of this year and a blown save was bound to happen eventually, but I'm really not too impressed with the managerial decision that set the stage for today's loss.

Here are the numbers explained by Tom Tango:

Anyway, looking at the RE chart for the AL, 1998-2007, below, we see we have an RE [Run Expectancy] of 1.487 with second and third and one out, and 1.695 with the bases loaded. Of course, that is with league-average batters and pitchers. Presumably, you are walking a well-above-average batter who is also a lot better than the following hitter, and/or the on-deck batter (plus pitcher) is a heck of a double play threat. Still, .208 runs is a lot of cheese to make up. Simply walking the batter to “set up the double play,” which is a common explanation for the intentional walk, is not going to help you reduce the other team’s run scoring potential, since the 1.695 runs that score, on the average, when the bases are loaded already include the GIDP. Run Expectancies for base/out states, AL, 1998-2007
        0 outs  1 out   2 outs
XXX 0.560 0.299 0.114
1XX 0.968 0.579 0.249
X2X 1.214 0.738 0.349
XX3 1.459 0.982 0.385
12X 1.585 0.988 0.469
1X3 1.907 1.271 0.544
X23 2.048 1.487 0.643
123 2.494 1.695 0.838
If you get into a situation where one run wins it like Saturday night, you will probably lose whether you've got runners on 2nd and 3rd or the bases drunk with only one out. That doesn't change the fact that you have a greater liklihood of losing with the bases loaded scenario. It only makes sense to load the bases then to avoid a strong hitter so you can pitch to a weaker one. We walked two weak-ish hitters to face a superior one. What?

Either of the two intentional walks would've been acceptable had they been issued to say, Vlad Guerrero or Jay killer Casey Kotchman. Instead, in the bottom of the 10th on Saturday (tied, 2-2), Brian Tallet intentionally walked Jeff Mathis, a .198/.258/.346 hitter this year so Jason Frasor could come in to face pinch hitter Juan Rivera. Rivera, now a bench player who was as good of a hitter as Alex Rios before breaking his leg and losing all but the last month of 2007, singled to win the game.

Today in the bottom of the 9th (Jays up 3-2), BJ Ryan intentionally walked Mike Napoli, a .220/.307/.505 hitter, to face Rivera again with the bases loaded and one out. Napoli has an impressive amount of pop in his bat to go along with a K rate over 30%. In hindsight (and thensight for yours truly), he should have been pitched to with Ryan on the mound. Rivera came in to pinch hit again, K-ing, and then it went off the rails. Ryan beaned Howie Kendrick (though it sounded like the ball hit his bat rather than his elbow) to send in the tying run then gave up a game-winning single to Macier Izturis.

For someone who's vehemently against the bunt for all the right reasons (statistically, you're likely to score fewer runs if you do it), Wilner definitely isn't up to snuff on the numbers behind intentionally walking the bases loaded. I left a comment on his blog in the wee hours this morning about how it's not such a great idea. His response:

MW: I don’t think this is the case in point. According to your chart, the Jays were already in a situation where the expected outcome was almost a run and a half, which is a loss anyway. By issuing the intentional walk, a force is created at the plate, and the run expectancy only rises by .2 of a run. Again, you’re over one already, so what’s the difference?

The numbers in the chart do take into account new scenarios (DP, play at the plate) after loading the bases and RE still goes up. Anything that makes that number go up is bad, right? My logic is that even if the situation is dire, you should still choose the option with the lower failure rate.

-- Johnny Was


John Olerud said...

I think having the force play at the plate is a pretty big deal in this scenario. I'm going to have to go with Wilner on this one, against the numbers...I'd be interested to see what those numbers are like close and late, and if there's any change. Obviously having the bases loaded increases the probability of forcing a run in with a walk or hit batter or an error in all scenarios, but in a tie or one-run game where that run on third is the one that matters, I (intuitively) get the feeling that it's easier to scratch that run home without the force play being on (via a squeeze, for example).

gabriel said...

I agree, tentatively with Johnny O there. What matters is not run expectancy per se, but the likelihood of Run #1 and Run #2. Runs 3, 4, 5, etc. go into run expectancy, but don't matter a whit in this kind of situation.

The Southpaw said...

Ok, fair enough, but I think you guys are still overlooking the hitters who were intentionally walked. And up by one today, I don't know how you can possibly argue for the strategy that increases in the opposing team's RE. Saturday in a tie game maybe it makes sense, but don't forget that for each new positive scenario (potential DP, force at home) there are negaive accompanying scenarios (the walk, hit batsman, etc). The numbers are take all possible outcomes into account and Tom Tango one of most respected basement-dwelling number crunchers in the business.

eyebleaf said...

i didn't particularly agree with walking Napoli either. BJ Ryan should walk no one. He is the man who shows no fear. But it was typical that he blew the save yesterday, wasting arguably AJ's best start as a Blue Jay. He was filthy yesterday.

oh well. onward and, hopefully, upward

sp said...

I'm a play-by-the-numbers guy as much as anyone else but I'd at least like to have the chance at a DP.