The Third Time (Through the Lineup) Is a Charm

On Monday I wrote about Chris Tillman’s success the third time through the lineup this season and how that seemed at odds with his established ability to limit hard contact. It turns out there isn’t much of a relationship between those things, so it’s likely that Tillman is due for some regression when facing batters for a third time in a game going forward. It’s likely because excellent research done by Michael Lichtman shows that a pitcher’s past deviations from the typical decline in performance each time through the lineup is not predictive of future performance. He estimates a pitcher would need about eight full seasons of significantly deviating from the average decline each time through the order before that data would become reliable in a predictive sense.

With that said, let’s look at the starters who have been the best the third time through the lineup this year. Most are regression candidates, but there are a few exceptions.

Player	1st OPS	2nd OPS	3rd OPS	Season
Carlos Carrasco	0.603	0.638	0.407	0.569
Joe Kelly	0.711	0.853	0.488	0.713
Kyle Hendricks	0.668	0.594	0.502	0.588
Garrett Richards	0.584	0.482	0.513	0.529
Felix Hernandez	0.468	0.600	0.515	0.548
Gio Gonzalez	0.713	0.734	0.521	0.671
Chris Tillman	0.753	0.684	0.532	0.661
Alex Wood	0.667	0.661	0.534	0.648
Hisashi Iwakuma	0.663	0.616	0.543	0.617

The full sample was the 175 starters with the most innings pitched against hitters the first time through the lineup this year. Carrasco and Kelly were the only two with a 3rd time through OPS more than two standard deviations better than the mean. A huge part of that is the small sample size we’re dealing with here. Carrasco has only faced 59 batters in their third plate appearance of a game, and Kelly has faced only 78.

Given that Lichtman’s research tells us that we should expect a pitcher’s performance to decrease in line with the league averages, Carrasco’s third time through OPS “should be” around .675. Actually, it would be helpful if you knew the league averages and how they change each time through the lineup, wouldn’t it?

	1st OPS	2nd OPS	3rd OPS	Season
League Average	0.678	0.708	0.751	0.710

So to be clear, Carrasco is still pitching like a well above average pitcher. He’s just likely to have less success in the later innings next year. Using his second time through OPS as Carrasco’s norm (as Lichtman’s research suggests we should), he’s pitching like a top 25 starter in terms of OPS against. Of course that’s a far cry from the top five starter he has been over the last 30 days. Not that anyone will be drafting Carrasco in the top five at the position next year. And given his struggles earlier this year he probably won’t even be going near the top 25. There’s a good chance he’s a value on draft day next year even though this post is warning of some regression. It’s all relative.

Kelly is a different animal because he hasn’t followed the typical pattern at all. He’s likely been a little unlucky the second time through the lineup, but he’s more than made up for it the third time through. But because those two even out, you probably don’t have to worry about Kelly regressing significantly. You could say the same for King Felix. He hasn’t followed the usual pattern, but his variations are smaller than Kelly’s and only minor regression should be expected.

It’s the other guys on this list you need to be worried about. You can probably worry less about Hendricks than the others because his sample size is so small the third time through (72 PA). And again, all the sample sizes are small. We’re talking about 150ish to 225ish plate appearances against guys getting their third plate appearance in a game. But to say you can worry less about Hendricks because of the extremely small sample size is to say that the regression won’t hurt as much because he has less distance to fall. It’s not to say that the small sample sizes mean the regression isn’t coming. It is. That doesn’t mean you have to avoid Richards, Gio, Tillman, Wood and Iwakuma next year. It just means you’ve got to fade them a bit. Maybe their ADPs will be depressed for other reasons and they’ll still be worth drafting. But this should factor into your valuation of them next draft season.