THE BOOK--Playing The Percentages In Baseball

it might be interesting to look at day/night wOBA by time through the order for all games, not just the CG-ish ones.  No day/night split for the 4th time through the order would support the “differential tiring rate” hypothesis.  If the day/night split was common to all games, you’d tend to chalk it up to shadows or biorythms or something.

technically, I guess you’d want to back out the CG-ish games, so that you were looking at CG-ish v non CG-ish…

It should be possible to determine whether it’s shadows that make the difference if you analyzed only those games which start at one o’clock, as compared to those which start in the mid-afternoon. Likewise, it should be possible to isolate the temperature effects by comparing games played in different months of the season. For example, if heat is a problem, causing pitchers to tire more quickly, the effect should be more pronounced in July and August.

MGL surely looked at all games, not just those that had at least 27 batters faced.  He surely adjusts for the quality of the batter.  He may have either removed PH, or applied the PH penalty as well.  He probably applied a park adjustment, those that part is mostly unnecessary.

Every starter gets to face a batting order the first time through.  Almost all of them face the batting order a second time through, and then a very big chunk will face the batting order a third time.  Very few, as compared to other times through, pitchers get to face batting orders a fourth time through.  And what is more almost no SP is allowed to fail as badly against the batting order a fourth time through as they would be allowed against the batting order at any other time.  So anyway each bucket will have a group of players that fail to get to the next bucket.  I would imagine that the group with the highest failure rate is the the 3rd time through bucket and it is no surprise that they do the worst against hitters.  But within that bucket are pitchers who don’t fail and can move on to the next bucket.  Just like pitchers like that exist in the first and second bucket as well.  I think the more important job is to detect the pitcher who can move on to the next bucket instead of simply lumping all pitchers in the game into the same bucket and then saying get them out of there as soon as possible.  It seems to me that with that kind of strategy you are creating a huge inefficiency that a smart team could then exploit.

McCoy, I don’t get your argument…

MGL,
are your numbers for 1st time through the order in first and in other innings for night games correct? 339 vs 341 doesn’t seem to be high enough to speak of real depression, when you have 330 vs 341 for day games.

(sorry if it’s double post, previous post seems to be lost)

MGL,

You removed starting pitchers in the ninth inning from consideration. I would suspect a fourth time through the order to occur most often in the eighth inning. Perhaps there is a similar (though smaller?) bias in the 8th inning, similar to what has been found in the 9th? I wouldn’t expect this wOBA decrease to suddenly occur in the 9th without beginning a bit earlier.

Also, I’m assuming the sample only includes pitchers who make it four total times through the order. If you include others, I would expect a bias during the last time through the order as the better pitchers would last longer in the game.

You introduce a bias by looking for pitchers who faced at least 36 batters.  You can’t do it that way.

And you probably missed the earlier threads that shows that this phenomena occurs only in the 9th.

Perhaps the issue is that I don’t completely understand the methodology used here, reading the earlier threads would probably help on my end.

Regarding bias, I think it would depend on the question you are trying to answer. If you are talking about a comparison between first, second, third and fourth times through the order, then you need to compare apples to apples. If your overall sample includes a number of starters who didn’t make it to a second time through the order, one would expect higher than average wOBA against for those pitchers, and these wOBA would be included only in the “1” sample above. After all, there would be a reason why that pitcher failed to last very long. Similar for second vs third, etc.

If you included only pitchers who faced 36 batters, I agree you would be biased toward the better pitchers. But you would be comparing this set of pitchers against each other as opposed to comparing against a number of pitchers who only made it one, two or three times through the order.

That said, another potential answer for the dip four times through the order would be a partial run through the entire lineup. For example, a starter makes it through 40 batters, meaning he faced the 1-4 batters the fourth time around. You would be comparing three full times through the order (including the 7-8-9 batters, often including the pitcher) to the top of the lineup.

Adam, yes, pitchers that make it to the later innings are slightly better pitchers (although not as much better as you would think - the main criteria for making it into the latter innings are pitch count and how you did in the earlier innings).  I control for that in the numbers.  The numbers you see are after adjusting for the talent (using seasonal numbers) of the pool of pitchers in each bucket.  So, for example, if in the 8th inning, the wOBA is .340 and the pitcher talent is .344, which is the average MLB pitcher, I call that .340.  If, in the 9th inning, the wOBA is .330, but the pitcher talent in that inning (as you said, better pitchers than the pool in the 8th), is .340 (4 points better than league average), I call that .334.  Get it?

As Tango says, we already determined that it looks like this depression (in close games only) only occurs in the 9th and it is mainly in games where the pitching team is ahead by 2 or more runs.  Also in the 9th, the wOBA is very high in tied games.

This seems to be because of an extreme approach by the batting team.  While there certainly are differing approaches in every inning depending on the score, which could affect wOBA, I am not surprised that it is not particularly noticeable until the 9th.  For example, if you are down 2 runs in the 8th inning, although you would like to tie the game or go ahead, you are certainly content with scoring 1 run.  In the 9th, you absolutely cannot score 1 run.  In the 9th, when down by more than a run, you often see batters take the first strike automatically.  You will not see that in the 8th.

That being said, I’ll look at the 8th inning by score differential. I don’t think I did that.

Also, I still need to post the components in the 9th to see what is going on…

"MGL,
are your numbers for 1st time through the order in first and in other innings for night games correct? 339 vs 341 doesn’t seem to be high enough to speak of real depression, when you have 330 vs 341 for day games.”

I think they are correct, but I’ll check.  That would be fascinating if the 1st inning road team effect only occurs in day games!  We would have a whole new topic to speculate on!

Censored observations!!!!!

My bet would be that survival bias is at work.  Another example that’s not perfectly related but close enough is BABIP over a pitcher’s career.  You’ll notice that as you increase the number of innings pitched, the BABIP for that collective group steadily drops.  One could infer that, thus, good pitchers have control of their BABIP.  Now, regardless if that is true or not, the selection process present in sports necessitates that bad pitchers get weeded out.  We don’t know if their BABIP would be lower or higher, because we censored their time series early.  “Good” pitchers could be producing that BABIP out of pure luck.

Just sayin’.