Sunday, October 26, 2014


To wage WAR, to zip up with FIP, or to quest down the road not taken with QMAX? The fix is in over at all the "advanced metrics" sites, where products that purport to combine predictiveness with probability prove only that they have more quirks than cut-to-the-chase insight.

Yet another case in point is the AL Cy Young choice in 2014, where WAR and FIP tell us that Corey Kluber is the consensus pick. It's interesting to note that these are the only two measures in which Kluber is the #1 choice, as demonstrated in the diagram at left.

Now don't take this as a slap against Corey. He had a fine year, becoming a much-needed ace for the Indians. He's just not quite on the same level (at least not yet...) with "FH."

AKA "King Felix" Hernandez.

WAR and FIP are measures that want to sweep up a lot of information and render them in a sequence of overly reductive equations. Please note that we are not dismissing them out of hand by saying this: we only want to issue a strong reminder that each stat has its limitations and caveats.

Neither of them can take what happened on the field and compare it to a more global probability of "what should have happened" using what actually happened. WAR doesn't even try to do this. FIP claims to do so, but makes a translation based on an equation-based summarization of "what should have happened."

QMAX uses what happens and translates it/compares it with a series of interlocking global probabilities, as represented in the forty-nine squares encompassing its grid category. (Remember that QMAX is an acronym for "Quality Matrix.")

It creates a series of stats that capture both value and shape. As such it is unique amongst all of its fellow measures. When we look at the matrix charts for Felix and Corey, we can actually see something different in their performance from the shape/pattern of the data.

Keeping in mind that the best games for a starting pitcher in QMAX are in the upper left and descend in quality toward the lower right, we can see right away (without any numerical support) that Felix was much better at avoiding games where he was "hit hard" (the region in orange that covers rows 6-7). Corey has nearly three times as many of these starts (8) than Felix (3).

And at the upper left, in the green region known as the "Elite Square," it's clear that Felix is ahead here as well (though Kluber's twelve ES games are nothing to sneeze at).

Each square in the matrix has an expected WPCT based on actual results (usually we use three years' worth of data to establish these--we call them QWVs (pronounced "qwivs"), for "QMAX win values."

When we add all of that up, we arrive at a winning percentage for the starter based on what his actual performance across all the squares in the matrix should produce if everything evens out. That's what WAR and FIP insinuate is the case for their measures, but really isn't.

The numbers for the AL starters show that Kluber had a fine year, but not as good as three other AL starters in 2014:

Two pitchers who missed stretches of the 2014 season due to injury--Garrett Richards and Chris Sale--were more effective than Corey when they were in there. WAR, which is a counting stat, penalizes them for that--which makes a certain kind of sense so long as you don't think of WAR as measuring quality (it's really measuring value).

The QMAX "range data" numbers at the right help to contextualize the results. It turns out that Felix reaches the "elite square" almost half the time--that's equivalent to Clayton Kershaw territory. Corey's 35% is down in the next tier with Sale, Jeff Samardzija, and Felix's teammate Hisashi Iwakuma.

Indeed, one of the reasons why Kluber doesn't rank higher is that his "Hit Hard" percentage (which we eyeballed above in the QMAX chart--orange region, remember?) is just too high to produce a dominant season. Being hit hard in one out of every four starts will knock you and your team out of some games. On the list above, Kluber is tied for tenth in terms of HH%.

The range data shows that Corey is pretty much middle of the pack amongst the top pitchers in the 2014 AL.

It would be ironic if Felix lost the Cy Young voting to Kluber because WAR and FIP actually aligned with starting pitcher win totals (Corey was tied for the most wins in the AL with 18, while Felix, who had seven no-decisions in games where he allowed one run or less, wound up with only 15).

That would demonstrate exactly what we've been asserting for quite some time now: that relying on any single measure to determine quality/value is risky at best and foolish at worst.

The same folks who voted for Felix back in 2010 when he was the best pitcher according to most of the measures (and is the case this year, as we've seen...) really ought to be doing the same in 2014.

Let's look at a few of the QMAX charts for the other notable AL starters in 2014. Garrett Richards was the only pitcher doing "a reverse" ("C" score higher than the "S" score) in the AL. Chris Sale is quite simply a helluva pitcher, and scary as all get-out to watch on the mound, either in the stands or in the batter's box. Max Scherzer was well off his 2013 form, and teams should be cautious about giving him big bucks (though someone will undoubtedly do so).

Finally, Phil Hughes is the new Tommy John. We haven't found anyone with a higher percentage of games in the "TJ" region (lower left, where many hits but no walks can still be successful...) in our database. As the table above notes, that's 59% of his starts.

Can he keep it up? Stay tuned...