i keed, i keed.

i’m off on a tangent.

Dave, you are a good analyst and writer, but let me try and explain why that logic is wrong – sort of a gambler’s fallacy. It is like if A-Rod is struggling (i.e., he is “cold”) I’ll move him down to 8th in the order and just move him back to 4th when he starts hitting again. Or, better yet, I’ll move Mathis up to 4th in the order because he is so hot and if he struggles, I’ll just move him down. Therefore, I have not lost anything. Of course you have lost something. Every PA that Mathis has in the 4th spot, whether he does well in that PA or not, you have lost something.

I know you know that, but you are making the same mistake with your, “If they struggle, just take them out and bring in Gaudin or Joba.”

Now, you are right in that the research shows that pitchers don’t lose much when they pitch on 3 days rest in the post-season, but, as you know, that is a selective sample – those pitchers who the manager thinks can handle it. Which would be the case for Petttitte and Burnett if he lets them pitch on 3 days rest of course.

So, it is probably not a big deal to throw them on 3 days rest. But your comment, “If they struggle…” is misguided. We know (and I think you know) that when pitchers struggle, it has little if any predictive value for the rest of the game. So how would taking them out when they struggle have any relevance to the effect of pitching on 3 days rest? It doesn’t. And I wish you wouldn’t perpetuate the myth of “When a pitcher struggles, it means that he doesn’t have his good stuff on that day,” or something like that.

Let’s put it another way: Let’s say that we throw them on 3 days rest. Your contention is that we are not sure that that is correct because they may be .5 runs (per 9 IP) worse in true talent and they may not be. I agree with that. No problem there.

Then you contend that we can mitigate that problem by “taking them out if they struggle.” You are implying that if they struggle early that it means that they likely are indeed significant worse on 3 days rest and if they don’t struggle early, it means that they were not affected (much) by pitching on 3 days rest. Well, that is where you go wrong. That simply isn’t true. We can’t tell a pitcher’s true talent from him “struggling” or not in the early innings. And we certainly can’t tell the difference between a 4.00 pitcher and a 4.50 pitcher, which is what we are trying to do (we are not sure if pitching on 3 days rest makes you .5 runs worse or so).

Here is the likely scenario: We pitch Pettitte on 3 days rest. He is normally a 4.00 pitcher (a few tenths of a run per 9 better than the average starter – BTW, he is not nearly as good as everyone thinks, partly because of that “post-season mystique” nonsense). Let’s say that he has a 50% chance of being a 4.50 pitcher on 3 days rest and a 50% chance of being normal – a 4.00 pitcher. So, we are basically putting a 4.25 pitcher on the mound (which is probably better than Gaudin or Joba, so that’s fine anyway) and we let him pitch the normal number of pitches or innings.

Again, you contend that somehow if we “take him out if he is struggling” we can do better than just letting a 4.25 pitch his normal amount of innings. Here is what happens. Let’s say that half the time he struggles and half the time he doesn’t. At best when he struggles (IMO), there is a 55% chance that he is a 4.50 pitcher and a 45% chance that he is still a 4.00 pitcher who just got unlucky for any one or more of thousands of reasons how a goof pitcher can “struggle.” You will take him out. Assuming that it is correct to take him out if he in indeed a 4.50 pitcher, you are going to be correct 55% of the time, but you are going to be really wrong 45%. If he doesn’t struggle, let’s say that there is a 55% chance he is a 4.00 pitcher in that day and a 45% chance he is a 4.50 pitcher who is just having a good few innings so far (as mediocre pitchers do all the time). So you leave him in and you are making a mistake 45% of the time again.

In total, you are going to make a mistake (by taking him out when he struggles but is really a 4.00 pitcher and by leaving him in when he doesn’t struggle but he is really a 4.50 pitcher) .5 * .45 + .5 * .45, or 45% of the time.

That is not a very good strategy. Especially since as a 4.50 pitcher, he is probably not much worse if at all than Joba or Gaudin, but as a 4.00 pitcher, he is probably a lot better.

The correct strategy is to simply throw Pettitte and Burnette, if you choose to throw them on 3 days rest, for 5 innings or so and then start bringing in your relievers for 1 or 2 innings each, which is ALWAYS the correct strategy when your starter is not an ace, which neither Pettitte nor Burnett are. And that way Girardi has an excuse for taking them out after 5 (they were throwing on 3 days rest).

This whole “take pitchers out when they are struggling and leave them in when they are not, especially the latter, is nonsense, and is one of the biggest mistakes that managers make.

If anyone has the time, please look at all starting pitcher performance in the last X years. Then isolate the ones where the starting pitchers are mediocre or worse – say, career ERA of 4.50 or worse. Then look at all games where they pitched well for 6 innings (say gave up 2 runs or less) and were allowed to continue. The assumption is that their pitch counts are not that high and the manager things they were “pitching well” and therefore will continue to pitch well.

Now, we know that a manger can take out an otherwise mediocre starter after 6 innings even if he is pitching well and replace him with a decent reliever for an inning or two with an ERA of around 3.50. There are plenty of relievers like that. It’s not like you have to find a set-up guy or closer for that kind of ERA.

OK, since in our research sample we have games where the manager decides that staying with the starter is the best course of action. He thinks that the starter, since he is pitching great, is going to pitch better than any reliever, otherwise why let him continue to pitch. You can even only look at close games if you want (say a difference of 3 runs or less).

Now, yes I know there may be other considerations. He may want to boost the starters ego or “stretch out” his outing. He may want to rest the pen. He may have no reliever available, or something like that. But, by and large, and in most cases, I think the assumption by the manager is that the pitcher is throwing well today, even though he is normally a mediocre or worse pitcher. After all, isn’t that what everyone thinks? Even sabermetric analyst David Cameron said, “Just take them out if they are struggling,” which is the converse of, “Leave them in if they are pitching well.”

OK, back to the research. So now, all we have to do is look at the ERA of all those pitchers in the 7th or later innings. Surely it is better than or equal to the easy 3.50 he can get from just about any decent reliever for an inning or so. And SURELY it is better than our group of pitchers’ career ERA’s. After all, they are “pitching well” and that is why we left them in. If a mediocre or worse pitchers “struggles” through the first 6 innings, he will likely be taken out, right?

So, let’s see what that ERA looks like in the 7th and later innings? Anyone up to this research project? That should answer once and for all whether it is correct to leave in a pitcher who is not a great starter if he is “pitching well,” right? If it is correct, then when managers actually leave them in, we should see an ERA which is considerably lower than their typical career ERA’s, right and certainly better than a halfway decent reliever.

I’ll say this also: If we find that those pitchers’ ERA in the 7th inning and later is not even better than a BAD short reliever (4.50 ERA), then we are really in trouble, right? That would be a horrible mistake by a manager – to eschew even a bad reliever for a mediocre or worse starter who has been pitching well, especially in a close game. Yet managers do that all the time. All the time. Let’s see if they are making an error and if yes, how egregious that error is.

One more thing if anyone is up to this research (which is not that hard with retrosheet data). Since starters are more likely to have pitches well for 6 innings against bad offensive teams, in pitcher’s parks on cold days with the wind blowing in, you have to make some adjustments for that. The best way would be to park adjust the ERA’s in the 7th inning and later sample and to tally the total offense of all the teams pitched against. For example, let’s say that we found an ERA of 4.00 in the 7th and later innings. And let’s say that the teams that these pitchers faced in these games (where they pitched well for 6 and were allowed to continue) had an average offense of 4.50 runs per game whereas the average team in the league was 4.75, we would have to adjust that 4.00 ERA by multiplying it by 4.75/4.50.