May 05, 2004
Futile Stats
Jay Jaffe at Futility Infielder takes on the productive outs crowd, or as he refers to them, the flat earth society.
The more I mull over productive outs the more incorrect I find it. One big problem is that you are only looking at situations in which outs are made, (Productive Outs)/(All Outs in Productive Situations). But why not count all successes in this situation? ((Productive Outs) + (Base Advancing Hits and walks))/(All Productive Situations) If getting a runner on first with no outs into scoring position is such so tremendously important, why not count the guys who do it by drawing a walk or getting a single? I'm going to try to work on this over the weekend.
Posted by David Pinto at
01:09 PM
|
Statistics
|
TrackBack (0)
I've heard the argument that the runner on first leaves the opportunity for a double-play with a following batter. It's an utterly insane position to take.
Keep in mind that a man on first and a man on second greatly increases the hitting alleys along first and toward left field, if the defense is making any attempt to hold the runners. It's not so much a measurable aspect, but it contributes to the success of a batter with a man on first and second with zero outs versus a batter with a man on second and one out.
You've also got the potential for three runs with one swing of the bat (HR), as opposed to simply two.
All kinds of reasons that Buster Olney is an idiot. I heard he steals money from old ladies too.
I'm kidding...geez.
What's the point here? Is the point to incorporate the value of these productive outs into the overall value of the players?
If so, then the productive outs definition is too constraining. Perhaps one could take a run expectancy approach and simply do some arithmetic addition of how a batter's at bat has changed the run expectancy in every one of a player's at bat over the course of the season, and this would give you some number that represents a player's run contribution over the average player's run contribution.
How about productive plate appearances (PPA). Did the batter do something to either get on base, or advance a runner? I don't have valuse, etc. figured out, just the basic idea (probably I've stolen it from someone). Hitting to the right side, advancing a runner, while you are out, is say a +1. (double play, naturally is -1) A walk is also +1. A meaningless flyout to center field is a 0.
There also seems to be the assumption that a bunt (or a hit and run) will be properly executed and that's not the case.
Most players in MLB were stars coming up through the ranks and many don't know how to bunt (because they have never been asked to). Take Gabe Kapler for example. He was a stud until he got to the bigs where he is now just barely better than a replacement level player. Kapler also happens to be one of the worst bunters I have ever seen.
Sure - bunting is a skill and it can be learned but for many players it may be too late.
Personally - I never want to see Kapler bunt - not because I'm anti-productive outs but because he can't do it.
The idea of a PPA is fine. But you would have to measure the number of bases advanced. If a single/walk is a +1 then the next batter's single is +2 and a + 3 if the runner advances to third. A home run with a runner on first, would be a +7
Just by looking at that you can tell why giving up outs is a bad thing. But as the saying goes if your going to make an out do something with it.
yeah, the numbers in a PPA based on +1 for every base would get astronomical very quickly, as in your example, while a double play (say even -2) wouldn't be so bad in that light. Needs lots of work. +0.25 per base?
How about just a +1 if you do something good, advance runners while using up an out. +2 if there's no out. -1 for double play, (-2 for triple play, but they're really rare).
I'd still like to see the correlation of men left on base to team runs scored and to team winning %. If Productive Outs don't have a visibly stronger correlation, they're nonsense. (I had some methodological suggestions here that may be helpful in your studies). Trumpeting Productive Outs as a key to postseason success may have been silly, but using the stat in the regular season is daft.
Derek Zumsteg sez:
"Generally speaking, giving up an out to advance a runner does two things:
- it slightly increases the chance that one run will score
- it decreases the chance that many runs will score"
I'd be interested in seeing how slightly the chances increase to scoring one run results from advancing a runner.
We're all familiar with the run expectancy matrix:
0 1 2
0 0.555 0.297 0.117
1 0.953 0.573 0.251
2 1.189 0.725 0.344
3 1.482 0.983 0.387
1+2 1.573 0.971 0.466
1+3 1.904 1.243 0.538
2+3 2.052 1.467 0.634
1+2+3 2.417 1.650 0.815
The productive out people insist that one run is appropriate in certain situations...while there's no doubt that "productive" outs help kill rallies of substance, if they do increase the likelihood of picking up one run, is there ever a situation where this is desirable? Is a visiting team with a 1-0 lead in the ninth who gets a leadoff double better suited to play for one run with a bunt+sac attempt, or a bigger inning by swinging/attempted walking away? If their chances of going up 2-0 are better by moving a runner over, even though moving that runner over diminishes their chances of going up 3-0, is that ever a gamble worth taking? I remember an article over on BP that suggested that even one additional run late in the game more significantly reduces risk of loss. I guess I go back to the question on Derek's statement: how slight is the increased chance at scoring one run?
Does mgl's super linear weights incorporate outs that advance runners? Because productive outs aren't better than a walk, but a productive out is, by definition, better than a strikout. Of course, it's worse than a GIDP, too.
And yeah, if you had the play-by-play data, you'd want to have a baseSit/out specific run value for every event. The value of a BB is worth more with the bases loaded and two outs than with a runner on second and no outs.
Of course, the more you introduce situations into things, the less predictive value the metric has. Hitters tend to not have much ability to control when their homeruns happen, or when they ground out.
I did a study of productive outs over the past two postseasons at The Hardball Times, if anyone is interested is seeing what that turned up.
(You won't be surprised, but the last stat in the article might surprise you)