Tuesday, April 13, 2010
Hate the ERA rules
Isn't it a crime that Will Ohman comes in, gives up a single, double and then gets a groundout to end the inning, allows his inherited runners to score and comes out with his 0.00 ERA intact? What a joke. Another stat should be more widely used that incorporates inherited runners scored into ERAs.
Subscribe to:
Post Comments (Atom)
Well, there are stats that do that already, only they don't exactly replicate ERA, and they are not very user-friendly.
ReplyDeleteThe idea behind each of them is to look at the scenario into which the RP enters, then to look at the result after the pitcher left, and to calculate a value for the performance based on those factors. But each uses a different way of measuring the value.
The first, which is called WXRL, uses Win Expectancy to determine the value. The idea is that for every possible score and inning and for every possible base-out scenario, there is, based on historical precedence, a likelihood of winning the game. If you are ahead by 3 runs with 2 outs in the 9th and no one on base, your expectation to win is almost 100% (or 1.00). If you are ahead by only 1 run with no outs in the 9th and runners on 1st and 3rd, your expectation of winning is significantly lower.
The idea behind WXRL is to look at the Win Expectancy rate before the reliever enters and the rate when the reliever leaves and to calculate the difference. That number (positive or negative) is added up into a counting stat that gives you an idea of how good the reliever has been. The nicest thing about this is that it really rewards pitchers who come into high leverage situations and shut the door.
ARP is similar, only instead of dealing with Win Expectancy, it deals with Run Expectancy. Regardless of the score and inning, there is, again based on historical play-by-play analysis, an average number of runs that should score for each base-out situation. In a bases loaded no out situation, 2.417 runs are expected to score, but in a bases loaded two out situation, only 0.815 runs are expected to score.
If you compare the expected runs before the reliever enters to the actual + expected runs after the reliever leaves, you can calcualte the number of runs (positive or negative) that the reliever saved.
I've seen somewhere, and I can't remember where, the idea to then take the ARP concept and turn it into a useful tool for re-calculating ERA.
So let's take last night's example:
Matusz leaves with one run in, bases loaded, and one out. The expected runs still to be scored in that scenario are 1.65. So the idea is that no matter who is pitching for the rest of that inning, the next 1.65 runs will be charged to Matusz (plus the one that has already scored). If fewer score, he benefits (as do they), but if more score, he's only really responsible for those that were expected when he left.
Johnson then comes in. When he enters, the run expectancy is 1.65 and records and out and allows a 2-run single. So now 3 runs have scored, plus, with Johnson leaving, there is an expectancy still (with runners on 1st and 3rd and 2 outs) of 0.538 more runs scoring. So 2.65 runs (or the 3) were charged to Matusz, and the next (1.65-0.538=1.112) runs should be charged to Johnson.
Ohman enters. When he comes in, the expectancy is for 0.538 more runs. He then gives up a single and a double in which 2 more runs scored, and finally records the last out. The run expectancy now is 0. There were a total of 5 runs that scored. Matusz got 2.65 of them, Johnson got 1.112 of them. The other 1.238 of them are Ohman's.
Then use those numbers in calculating ERA instead of assigning 4 runs to Matusz, 1 to Johnson and 0 to Ohman.
Now, as I mentioned, this is not user-friendly, and it's not easy to calculate. So if you prefer a much easier idea, check this out:
http://www.thegoodphight.com/story/2007/6/28/174013/394
-Brian
That's exactly what I'm talking about. Only catch is, these non-user-friendly stats are only generally known by statheads such as yourself and possibly someone like me, who takes some interest in numbers as they relate to sports, as well.
ReplyDeleteTruth is, I don't think that's such a complicated stat, and so what if it is? Does anyone understand how the QB rating is calculated? And yet, it's the preeminent stat for the the QB. Pretty basic fans know QB ratings for their favorite quarterbacks and what a good rating is. So, it can be done.
My point is that ERA is somewhat outdated and not a true indication of a pitcher's quality ... not as much so as wins and losses, but still.
And on top of that, think about how the ERA is skewed for relievers relative to starters. Relievers often come into games only needing to record an out or two to finish an inning and have an immediate benefit for ERA. Starters have just the opposite - they always start innings and usually leave games with men on base credited to them.
I agree completely, except that you have to remember that in some cases the starters' ERA can benefit from having left runners on that don't score (thanks to a good bullpen). Just think how many of those runners would be charged to the starter if we used the Runs Expectancy chart.
ReplyDeleteI don't think anyone should use ERA to determine whether a reliever is any good.
It's pretty good for starters, but even then, it still needs to be put into context (park factors, league factors), and it needs to be adjusted for runners left on base when they leave.
There are stats I've seen called "Best Case ERA" and "Worst Case ERA", in which the pitcher's ERA is calculated based on no bequeathed runners scoring and based on all bequeathed runners scoring. In most cases, actual ERA is right in the middle, so it tends to even out.