The Most Perfect Non-Perfect Game

by
June 19, 2014

Because Hanley Ramirez sucks at playing defense, baseball will not officially recognize Clayton Kershaw’s effort tonight as a “perfect game”. But I would like to submit that if this doesn’t qualify as a perfect game, nothing should.

28 batters came to the plate; 15 of them struck out. Of the 13 who managed to put the ball in play, nine of them hit the ball on the ground. One of the four balls hit in the air didn’t leave the infield. His FIP for the game was -0.24, because the model isn’t designed to handle dominance at this level. His xFIP was 0.19.

Here is the full list of nine inning outings with a Game Score of 102 or better, since 1914.

Kerry Wood: 105 (9 IP, 1 H, 0 BB, 20 K)
Clayton Kershaw: 102 (9 IP, 0 H, 0 BB, 15 K)

That’s it. That’s the entire list.

Clayton Kershaw did not retire every single batter he faced tonight, so technically, he wasn’t perfect. Screw technicalities, though; what Clayton Kershaw just did was far more impressive than going 27-up, 27-down and relying on your defense in order to do it. Clayton Kershaw just threw one of the most dominant performances in the history of baseball.

It might not have been perfect. It was better.





 

Dave is the Managing Editor of FanGraphs.

111 Comments
Oldest
Newest Most Voted
Inline Feedbacks
View all comments
Pearberr
9 years ago

I’m fairly new to this… but FIP is supposed to “Mimic” ERA while providing a better evaluation that is context-neutral, correct?

So how exactly does one have a negative FIP?

0
Dave Cameronmember
9 years ago
Reply to  Pearberr

Models designed to work in normal environments break at extremes. It’s pretty common, not just with baseball modeling, but with basically any model you create. You set parameters that mimic real life, and then when something insane happens, your model says “haha what” and falls apart.

116
Bryz
9 years ago
Reply to  Dave Cameron

I gave the literal answer below, but yeah, I know exactly what you mean. In a college class, we had to devise a model to determine the optimum number of tollbooths based on rate of cars approaching and number of lanes before approaching the toll plaza. I can’t remember the exact criteria (something like a 6 lane highway with a very low rate of traffic), but there was one situation where the model told us we needed a negative number of tollbooths, while the other ~11 situations all worked.

2
Pearberr
9 years ago
Reply to  Bryz

Part of why I love Sabermetrics so much is because I’m an Economics major starting some higher level courses and this science (Can it be called that?) combines two of my favorite passions. I’ve never really seen a model break completely like this one though, haha.

At risk of sounding pathetic, I’m starting my thesis this fall (And want to be well on my way by the end of summer) and would love to do it on the impact of taking pitches separate from the increased walk rates and would love some help from one of you guys!

1
atoms
9 years ago
Reply to  Bryz

You can only call it a science if you use the word “dismal” in front of it. It’s in the econ bylaws.

8
kdm628496
9 years ago
Reply to  Dave Cameron

hi dave, could you explain why FIP doesn’t have an artificial floor at 0, the way LOB% has an artificial ceiling at 100%? thanks.

0
a eskpert
9 years ago
Reply to  kdm628496

It’s a straight up linear formula. If there are too many strikeouts relative to walks and home runs, it will become negative.

0
Bipmember
9 years ago
Reply to  kdm628496

The ceiling of LOB% is not artificial. LOB% is not a stat, it’s a record that measures what percentage of allowed baserunners don’t end up scoring. There is no way for more than 100% of baserunners to be prevented from scoring.

As for why FIP doesn’t have an artificial floor, it’s because it’s not meant to be used over a single-game sample. Over a sample the size of a half-season to a season, there is no way for it to be negative. No pitcher could possibly manage that.

1
kdm628496
9 years ago
Reply to  kdm628496

@Bip
according the fangraphs library, LOB% = (H+BB+HBP-R)/(H+BB+HBP-(1.4*HR)). if you take a look a koji uehara’s 2014 season so far, you’ll see it shows his LOB% to be 100%. he’s given up 18 H, 4 BB, 0 HBP, 3 R, and 3 HR. plugging these into the formula gives you 106.7%.

obviously 106.7% LOB doesn’t make sense, so it’s capped at 100%. similarly negative FIP doesn’t make sense, so it should be floored at 0.

2
Bipmember
9 years ago
Reply to  kdm628496

That’s odd, didn’t know that.

But anyway, another reason that FIP wouldn’t be artificially floored at 0 is that it’s not accurate at that level anyway, so there’s no point. Obviously a pitcher with a true-talent 56.3% K rate, 0% BB rate, and 0% HR rate would give up almost no runs, but he would give up some runs, so even an artificially floored FIP of 0 would be too low.

It would be more worthwhile to adjust the FIP formula to work better at extremes than to just floor it.

1
Bryz
9 years ago
Reply to  Pearberr

Tons of strikeouts with very few/no home runs and walks. Since the formula adds HR and BB while subtracting K, if you have the right mix you can override the constant added to the end, causing a negative FIP.

Like Dave said, FIP isn’t meant to handle pitchers that pitch this well.

2
Pearberr
9 years ago
Reply to  Bryz

OH! I see, the presence of the constant makes this possible.

Makes me curious if anything like this has happened before, haha.

0
tz
9 years ago
Reply to  Pearberr

On a single game scale, it happens all the time with dominant relievers. Craig Kimbrel has already had 102 games with a negative FIP in his young career.

Maybe we should have a new stat for any game with a negative FIP: a “Kimbrel”.

28
Mr. Jones
9 years ago
Reply to  Pearberr

Look at Eric Gagne’s splits from his prime. I think I remember reading that he had a negative FIP vs. RHH for an entire season, or something crazy like that.

10
tz
9 years ago
Reply to  Pearberr

You’re right. Gagne in 2003 vs RH batters, over 151 PAs:

http://www.fangraphs.com/statsplits.aspx?playerid=650&position=P&season=2003

His 2003 season was the only time in MLB history that a reliever had 100 more strikeouts than hits allowed (137 vs. 37)

14
John C
9 years ago
Reply to  Pearberr

Forget all of these relievers doing it in pitchers’ parks. Koji Uehara last year, between Aug. 21 and Sept. 13: 11 games, 12 innings pitched. 36 up and 36 down, with 17 K’s. Five of those games in Fenway and one in Yankee.

-1
MattR
9 years ago
Reply to  Pearberr

It didn’t matter which park Gagne was pitching in since no one could hit him.

3