jump to navigation

Mets Run Support by Starting Pitcher August 1, 2014

Posted by tomflesher in Baseball.
Tags: , , , ,

Yesterday’s post discussed distributional wins and losses based on the Mets’ inconsistent bunching of runs together. Since the boys didn’t play last night, I had a pretty stable dataset to work with, and the opportunity to crunch some numbers to see if the hypothesis that we’re working with is true. In addition, I took a look at each of our current starting rotation’s run support numbers and found some surprising things.

First of all, no pitcher had a statistically significant run support number than any other. Although Dillon Gee‘s run support is .77 lower than the average pitcher, for example, the p-value is .44, meaning the probablity that that’s statistically different from 0 is just about 56%. Jacob deGrom has a similar number – .796 runs below the average, but a .42 p-value. The only pitcher with a positive effect on run support is Bartolo Colon, but his p-value is a whopping .72, meaning it’s more likely than not that his number is a statistical artifact.

The runs allowed are a bit more stable – deGrom allows 1.18 runs fewer than average with a .2 p-value – but Gee, Jonathon Niese, Colon, and Zack Wheeler all have statistically 0 effect on runs allowed. Their ps are, respectively, .91, .84, .64, and .79. Basically, this means that an effect would have to be really big to show up in such a small sample size, not even all 108 games are covered in the sample.

Another way of tracking pitcher run support is to track team wins and losses in the games started by those pitchers and compare it to the team’s Pythagorean expectation in those games. This is a bit more revealing; for example, the Mets are 6-8 in starts by deGrom, but would have a Pythagorean expectation of about .568, or about 8-6, in those games. Wheeler also ends up with a Pythagorean expectation better than his record, predicting the Mets would have won 11 rather than 10 of his 22 games. The other pitchers are more or less in line with their expectations, although, like Zack, the pitchers don’t always get credit for the wins they pitched in.

Behind the cut is the table of regression results for a linear model with a dummy variable for each pitcher’s starts, plus a totally useless Away game dummy to look for home field advantage. (Surprise: There is none for the Mets, but all pitchers do allow roughly .74 more runs on the road than at home.)



Quality Starts and Differential Luck July 12, 2014

Posted by tomflesher in Baseball, Economics.
Tags: ,
add a comment

Zack_Wheeler_on_July_25,_2013On July 11, Zack Wheeler gave the Mets a quality start by either definition – he pitched 6 2/3 innings and allowed only one run for a game score of 64. The  Mets managed to convert it into a win, which they’ve managed to do in 27 of their 46 wins thus far this year. Zack’s made 12 quality starts this year (by the sabermetric definition of a game score of 50 or more), but the Mets have managed to convert only 5 of them into Ws for Zack; the team is 7-5 in those games, while Zack himself is 5-2. That’s a far cry from the Giants’ freakish Tim Lincecum (9-0 in 12 quality starts) and the Angels’ Garrett Richards (10-0 in 15 quality starts). (The whole list of pitchers with quality starts so far is here.)

That got me thinking – which teams do the best at converting quality starts into wins? Which teams are the worst? What’s the relationship? I grabbed all of these numbers and put them together into a spreadsheet in order to play with them.

First, a quick review of terms: A cheap win is a pitcher win in a non-quality start. A tough loss is a pitcher loss in a quality start. “Luck” is whatever I happen to be measuring at the moment, but today ‘luck differential’ refers to the difference between the percentage of wins that are cheap and the percentage of losses that are tough; in other words, luck differential = 100*[(CW/W) – (TL/L)]. For an individual pitcher, these are fairly random occurrences – no pitcher in MLB today hits reliably enough to consistently earn himself cheap wins – but it seems that aggregating by team allows for the quality of batting to smooth out over a large number of games.

The Texas Rangers lead the league in this sort of luck differential, with 4 of their 38 wins coming cheaply for over 10% cheap wins but only 2 of their 55 losses tough (3.64); the Atlanta Braves have the worst luck differential in the league with a high proportion of tough losses (17/42, or 39.53%) and a low number of cheap wins (3/50, or 6%) for a total of -33.53. The Mets themselves convert less than 50% of their quality starts into wins for the starting pitcher.

These numbers are indicative of a general trend. The more quality starts a team has, the more negative its luck differential is (ρ = -.72 – an extremely strong correlation) and the more wins a team has, the more negative its luck differential is (ρ = -.20 – a bit weaker). Essentially, teams with more quality starts generate more wins (ρ = .56), regardless of the fact that sometimes they lose those quality starts, too. Surprisingly, the Mets have a -21.67 luck differential, one of the most negative in the league, probably due to the fact that they convert so few quality starts into wins.