MASTER NOTES: Lucky (and unlucky) pitcher wins

I yelled at my dog Leo last Saturday. Not because he did any business on the new basement carpet, and not because he stole yet another sock or pair of shorts out of the laundry basket, and not because he was yapping maniacally at the UPS guy.

No, I yelled at Leo because I had two pitchers lose wins I thought they had in the bag.

I’m 99.9% sure that losing “sure wins” has happened to you, too, if you’ve been playing for any length of time. And I’m about 40% sure you’ve yelled at your dog because of it.

I had two starting pitchers going on Saturday, Jose Berrios and Mike Fiers. Both pitched well, and both left with leads. In each case, I had penciled in a win, in a league where I’m in a standings clump where five wins separate seven points. At the time, those two wins would have meant two points in the category.

You can see where this is going.

In KC, the Twins brought in the reliably excellent lefthander Taylor Rogers, whom I have on a different roto team. Rogers got two quick outs, then allowed a sequence of hits and a HBP to score both of Berrios’ baserunners. Two extra earned runs for Berrios, and no win for him (or for me).

A similar story unfolded in OAK. Ryan Buchter came in to replace Fiers. He also got two quick outs, then gave up a homer to tie the game. No win for Fiers (or for me).

This got me thinking about when I should be allowed to feel righteous in my indignation about losing wins. And, by extension, not so guilty about yelling at Leo.

Four years or so ago, BHQ analyst Matthew Cederholm wrote a Research article describing a formula for Expected Wins, using the modified Pythagorean formula for Expected Wins:

eWins=Team Runs per game (R/G)1.80
---------------------------------  * 0.72 * Games Started
Pitcher ERA1.80+Team R/G1.80

The 0.72 is to account for the percentage of starts that end up as no-decisions.

It’s pretty cool, and does a good job identifying outliers in pitchers’ win totals versus their expected wins totals.

But Matt’s formula is aimed at season-to-season variations and expectations in wins. It finds pitchers who had too many (or too few) wins for the entire year, with the idea that those pitchers would regress to expectation the following year.

What I needed to salve my conscience, after snapping at Leo like he was Taylor Rogers or Ryan Buchter, was some idea of expected wins on a game-by-game basis.

I used another venerable BHQ concept, the Pure Quality Start (PQS). It’s a five-point system for scoring each individual start, based on innings, hits, Ks, walks and HR (the full explanation is here).

My idea was to find out how often starters get wins in PQS-DOMinant starts (PQS scores of 4 or 5), versus wins in PQS-DISasters (0 or 1) and what I call PQS-MEH (2 or 3). Then I could look at all the pitchers this year (or any year) and see which pitchers have been lucky or unlucky. I looked at all the starts compiled in the BHQ PQS logs, which can be found in the Leading Indicators page. I eliminated any pitchers who had averaged fewer than 4 innings per start, to get rid of “openers” who can’t win games.

The win percentages by PQS-level were:

PQS    Win  Loss  NoDec
=======================
DOM    60%   14%    26%
DIS    10%   57%    33%
MEH    34%   27%    39%

So pitchers should win about 60% of their PQS-DOM starts, and lose roughly the same of their PQS-DISasters. That seems reasonable.

The next step was to assess the starts of all 218 non-opener starters this season, and to give them xWins for their starts in proportion to those PQS outcomes. So if a starter had 10 PQS-DOMs, he would get 6 xWins (60% of 10 starts), with other xWins in proportion with his PQS-DIS and PQS-MEH starts.

From there, it was easy to stack-rank the pitchers by actual Wins minus xWins. Starters with high-positive totals would have been lucky; high-negative totals unlucky.

Based on this table, I should apologize to Leo. Yes, Berrios lost a win he should have had. But overall this season, he has actually been a little lucky, with three “W”s in just 4 DOM starts, a 75% clip. He has also won four of his 10 MEH starts, which is above normal, and even snagged one win in his two DIS starts, which is way out of line. Overall, he has eight wins this season but only 6.1 xWins. In other words, Jose and I have almost two wins more than what we deserve.

And Berrios is nothing compared with Fiers, who is just outside the top-10 luckiest starters in baseball this season. He has seven wins, when xWins says he should have around four. He’s three wins to the good! He did get a No-decision in one of his two PQS-DOM starts, but he also has five wins in his eight PQS-MEH starts, close to double what he should have.

The other lucky starters above Fiers are:

Pitcher          W  xWins   Diff
================================
Jake Odorizzi   10    5.6   +4.4
Lucas Giolito   10    6.4   +3.6
J.A. Happ        7    3.5   +3.5
Domingo German   8    4.7   +3.3
Max Fried        8    4.9   +3.1
Frankie Montas   9    5.9   +3.1
Andrew Cashner   7    4.0   +3.0
Jonathan Gray    8    5.0   +3.0
Zachary Davies   7    4.0   +3.0
A Senzatela      6    3.1   +2.9
Marco Gonzales   8    5.1   +2.9

And here are the unluckiest:

Pitcher          W  xWins   Diff
================================
Chris Sale       3    6.8   -3.8
Tyler Mahle      2    5.5   -3.5
Trent Thornton   2    5.3   -3.3
M Bumgarner      3    6.2   -3.2
Trevor Richards  3    6.2   -3.2
Jorge Lopez      0    3.2   -3.2
Trevor Bauer     5    8.1   -3.1
Jacob deGrom     4    6.8   -2.8
Yu Darvish       2    4.6   -2.6
Max Scherzer     6    8.5   -2.5
Sonny Gray       3    5.4   -2.4
Brad Keller      3    5.4   -2.4

Does any of this mean these pitchers are “due” to see corrections in their win totals? No. There are other factors at play, irrespective of the PQS-score he had. For instance, run support. It’s probably not a coincidence that the “lucky” guys include two starters from the powerhouse offense in NYY, and one each from MIN, MIL, OAK, ATL, and SEA.

Similarly, many of the “unlucky” starters pitch for teams with anemic offenses, like Mahle (CIN), Thornton (TOR), Bumgarner (SF), Lopez (MIA), Bauer (CLE, and two of my fantasy teams), and Keller (KC).

Look into the matter deeply, and you’ll also find some bullpen issues plaguing the unlucky while benefiting the lucky.

The moral of the story is: No matter how we slice it, wins are just going to be fickle and largely unpredictable. And no matter what he has to do with your carpeting, your laundry, or your UPS guy, your dog has nothing to do with your pitchers.


Click here to subscribe

  For more information about the terms used in this article, see our Glossary Primer.