clock menu more-arrow no yes mobile

Filed under:

Felix Hernandez, king of the quality start

To illustrate the consistent greatness of Felix Hernandez, the Mariners' broadcasters unveiled the "ultra quality start" and the "mega quality start". How do these statistics measure pitcher quality? Plus, we respond to a question left in the comments on last week's article.

Felix Hernandez is fun to watch. Year after year, Hernandez is a perennial Cy Young contender, and when his stuff is working (as it was in Saturday's start against the A's), even the best hitters in the world seem overmatched.

Hernandez is great by any objective measure. Since 2009, Hernandez has struck out at least 200 batters every season. Batters have managed a measly .240 average against Hernandez in his career. His earned run average (ERA) is 22% better than league average over his career. And, as Dave Sims and Mike Blowers pointed out on the ROOT Sports telecast, Hernandez has the third-highest WAR total for a pitcher over the last five seasons. By FanGraphs' estimation*, Hernandez was worth 28.4 more wins to the Mariners than his theoretical replacement, trailing only Justin Verlander (33.6) and Cliff Lee (30.0).

* - Granted, by Baseball-Reference's WAR metric, Hernandez is also behind Clayton Kershaw. Still: really good company.

Sims and Blowers also used quality starts to highlight Hernandez's consistent excellence. As originally defined, a quality start is awarded to any pitcher who goes at least six complete innings while allowing no more than three earned runs. But a pitcher who throws exactly 6 IP every game while giving up exactly 3 ER every game ends the season with a 4.50 ERA, which would have been around 10% worse than the league average last year.

So the ROOT Sports broadcast tightened the quality start criteria, awarding "ultra quality starts" for starters who go at least 7 IP and give up at most 2 ER, and "mega quality starts" for going at least 8 IP and giving up at most 1 ER. This table shows how those classes of starts have compared from 2010 to 2013. We see that mega quality starts are much more likely to result in a win for both the team and the starting pitcher; a starting pitcher who turns in a mega quality start has a winning percentage of nearly .950.

No. SP Win Pct Team Win Pct Game Score ERA
QS 10226 0.744 0.667 63.1 1.95
Ultra QS 4840 0.852 0.757 70.0 1.18
Mega QS 1222 0.948 0.887 78.5 0.50

If we look at the leader board over those four years, we see why the Mariners' broadcasters brought up the concept. Hernandez leads the league in both ultra and mega quality starts. Joe Posnanski calculated in 2011 that the average MLB starter with at least 100 starts earned a quality start 55 percent of the time; over the past four years, Hernandez has earned an ultra quality start at that same rate.

Name QS QS % Ultra QS UQS % Mega QS MQS %
Felix Hernandez 95 72.5% 73 55.7% 38 29.0%
Cliff Lee 88 72.7% 64 52.9% 31 25.6%
Clayton Kershaw 100 76.3% 70 53.4% 29 22.1%
James Shields 89 66.9% 56 42.1% 21 15.8%
Cole Hamels 93 72.7% 59 46.1% 21 16.4%
Roy Halladay 71 68.9% 47 45.6% 20 19.4%
Justin Verlander 97 72.4% 60 44.8% 20 14.9%
Jered Weaver 93 76.9% 56 46.3% 18 14.9%
Adam Wainwright 69 69.7% 42 42.4% 18 18.2%
Matt Cain 92 71.9% 55 43.0% 17 13.3%

Impressive as all of this is, the most interesting part to me was the pitchers who earned a mega quality start -- that's 8+ IP and one or no earned runs allowed, remember -- but were still saddled with a loss. There were 55 such starts over the past four years; let's look at that leaderboard.

Name QS QS % Mega QS Mega QS Losses
A.J. Burnett 62 49.2% 9 3
Cliff Lee 88 72.7% 31 2
Roy Halladay 71 68.9% 20 2
James Shields 89 66.9% 21 2
Chris Sale 42 71.2% 8 2
Matt Cain 92 71.9% 17 2
Matt Garza 60 57.1% 11 2
Jake Peavy 52 57.8% 9 2
Jordan Zimmermann 63 64.9% 7 2

To see Cliff Lee or James Shields or Roy Halladay near the top isn't that surprising: they only have two losses, after all, and they have the highest number of these games. But poor A.J. Burnett threw nine mega quality starts over the past four seasons, and picked up three losses! Here are Burnett's three losses:

  • Aug. 15, 2010: Royals 1, Yankees 0 - Bryan Bullington had two quality starts in his career, and this was one of them. The 2010 Royals lost 95 games, but they held the Yankees to two hits. Willie Bloomquist scored the only run in the bottom of the first after a single, a stolen base with an errant throw, and another single.
  • Apr. 25, 2011: White Sox 2, Yankees 0 - Burnett was 3-0 coming into this game, and picked up his second loss as Philip Humber one-hit the Yankees over seven innings. Humber threw a perfect game in 2012 against the Mariners, and is now at Triple-A Sacramento.
  • Sept. 28, 2012: Reds 1, Pirates 0 - This one's my favorite: Homer Bailey no-hit the Pirates and struck out ten (including Burnett once). Andrew McCutchen walked in the seventh, stole second, but was caught trying to steal third.

From the Comments

I want to close with a quick response to something from the comment section on last week's article. Vivafakethirdtofirst asked whether pitchers should prefer solo home runs, writing,

It occurred to me that giving up a solo homer has an advantage in the pitch count of starter over a run scored otherwise. Has anyone tried to calculate this? Because it seems it would be a factor inside of the home run rate. He is able to stay in the game longer and get more outs. Or maybe this is too situational to have a widespread notation.

As I said in my response, I think this is an interesting concept if you can control for confounding factors like times through the order and pitcher/hitter quality. But for now, let's try a thought experiment.

To start, let's consider the current run environment. Here is the run expectancy matrix over the last four seasons, when the average team scored around 4.3 runs per game.

Run Expectancy Matrix, 2010-2013

BASES 0 OUTS 1 OUT 2 OUTS
___ 0.4827 0.2571 0.0994
1__ 0.8477 0.5078 0.2236
_2_ 1.0938 0.6583 0.3188
12_ 1.4119 0.8725 0.4275
__3 1.3709 0.9401 0.3337
1_3 1.7759 1.1254 0.4788
_23 1.9788 1.3734 0.5627
123 2.2352 1.5402 0.7523

Let's consider two scenarios. In the first, the leadoff batter hits a solo home run. After this home run, the bases are empty and there are no outs, so we expect an additional half a run to score in that inning.

But in the second scenario, the leadoff batter gets on base somehow, and the next batter drives him in somehow. This is the "run scored otherwise" scenario. There are four possible base/out states we can be in: bases empty, one out (0.25 additional runs expected); runner on first, no outs (0.85); runner on second, no outs (1.09); and runner on third, no outs (1.37). In three of the four situations, we expect to score more runs in the average inning than after a solo home run. So we would expect a starter who gave up a solo home run to give up fewer runs than a comparable starter who gave up two straight doubles, for example.

Obviously, this isn't rigorous. This experiment assumes an average offense and pitcher, but the real talent levels may vary. We should also check the relative frequency of each of the four run-scoring states, so we can be sure that we have a higher run expectancy when we average the states together. And using two batters to score a run is less efficient than a solo home run, and having a better hitter up after the solo home run may offset some of the difference.

But overall, I don't know that you'd be able to see any effects on the home run rate: pitchers prone to solo home runs are probably more likely to give up home runs with runners on base, and those will definitely limit how long you stay in the game.

. . .

All statistics courtesy of FanGraphs, Baseball-Reference, and Retrosheet.

Bryan Cole is a featured writer for Beyond the Box Score, and can't believe this is Felix Hernandez's tenth season. You can follow him on Twitter at @Doctor_Bryan.