Monday, May 31, 2010

I've Got Blisters on my Fisters:
Ignore Context at Your Own Risk  

Veni, Vidi, Praetermissi (I came, I saw, I missed the point)
-- Calvin Coolidge Julius Caesar Tuskahoma McLish

One of the most endemic problems in American business is management taking numbers they haven't considered in full context and, because the numbers support a pre-held belief or wish, relying on them for decisions.

I'm not talking about lies, white or otherwise, such as BP's knowingly shrinking the numbers on the daily ooze out of the Gulf well hole or Enron's or AIG's three-card monte accounting numbers. I'm talking about real numbers that, when taken without adjusting for context, tell a different, less certain story than wishful thinking would lead a manager to otherwise.

Take the Seattle Mariners' back of the rotation success story, starting pitcher Doug Fister. He's having a very cool April and May, perhaps the highlight of an otherwise grim-looking-to-date 19-30 (last place, 7 games out of their division lead, 9-1/2 games out of the wild card) season for the team.

Fister's 2010 record through May 30:

2010 SEA 3 2 .600 2.03 9 62.0 49 15 2 10 26 2 241 202 0.95 7.1 0.3 1.5 3.8 2.6
2009 SEA 3 4 .429 4.13 11 61.0 63 29 11 15 36 2 256 105 1.28 9.3 1.6 2.2 5.3 2.4
2 Seasons 6 6 .500 3.07 20 123.0 112 44 13 25 62 4 497 138 1.114 8.2 1.0 1.8 4.5 2.48
162 Game Avg. 10 10 .500 3.07 35 214 195 77 23 44 108 7 867 138 1.114 8.2 1.0 1.8 4.5 2.48
Provided by Baseball-Reference.com: View Original Table Generated 5/31/2010.

The cool synchronicity is that as of today, he's pitched roughly the same number of innings pitched as he did in his 2009 rookie season, so we can look at his counting stats (such as Homers surrendered or Hits and Walks allowed) to compare his outcomes year to year. His homers-surrendered is much better (league-leading, in fact), both his Walks/9 innings and his Strikeouts/9 innings are down, but the ratio has held and is slightly improved. Most significantly, his hits-allowed figure has dived.

Just looking at these numbers, there's no doubt that his outcomes are improved. Curious minds attacked the numbers to try to de-code why this not-famous newcomer is putting up Ace starter numbers. Sadly, the results were Veni, Vidi, Praetermissi (I came, I saw, I missed the point).

David Golebiewski over at Fan Graphs noted many important numbers underpinning Fister's success, most notably the M's team defensive skill and some luck factors (including his ratio of homers to total fly balls allowed). My acquaintance, the clever Matthew Carruth chimed in at the same site a few weeks later, also noting some luck-not-likely-to-continue (including the percentage of runners left on base when he left the game that relievers allowed to score). And there's this opinion from Seattle Sports Insider that it's Fister's superior rotations-per-minute on his 2-seam fastball (possible, but I have to suspect this is a conclusion made because data was available, not a conclusion confirmed/denied by data).

And even a clever article with the word "context" in the title, this one from Brooks Baseball doesn't find the major reason Fister is successful this year.

HE'S PITCHING MOSTLY AGAINST WEAK OFFENSES. AND IN PITCHERS' PARKS. As Don Malcolm pointed out, in the nine games Fister started to date, six of the nine offenses he's faced are below league average. And in all nine of his starts, he's pitched in stadia that are playing to date this year as pitchers' parks.

Glance at the following table. The columns labeled H/R are stating the offensive numbers for Fister's opponents Home or Road, depending on where they faced him. In the Park Factor column, compare the fraction of runs scored in that park compared to the MLB average ballpark this season (and to the AL median average of .986); so games played this year to date at Seattle's home park where Fister has pitched six of his nine starts have lowered runs totals, .945 to a norm of 1.000, or about 5.5% fewer runs per game. The Offense Factor number represents the normalized (1.000 equals the league norm) of the team facing Fister in the context of where they are playing. Some teams play better at home; Texas & Baltimore, for example, look like good offensive teams if you look at their team batting stats for the season, but both teams faced Fister in Seattle, and both teams have much weaker offensive results away from their home stadia. The numbers listed here are their road numbers, just as Chicago's numbers here are for their home games (because that's where they faced Fister.

Opp Site   Park Factor Offense
DET SEA .945 1.037 .273 .340 .424 .764
TBR TBR .923 1.037 .260 .337 .427 .764
TOR SEA .945 1.012 .232 .303 .443 .746
CHW CHW .920 .991 .237 .322 .409 .730
OAK OAK .818 .966 .268 .333 .378 .712
LAA SEA .945 .929 .240 .302 .383 .685
OAK SEA .945 .901 .239 .308 .356 .664
BAL SEA .945 .894 .247 .301 .357 .659
TEX SEA .945 .864 .239 .299 .338 .637
AL Average .986 .258 .331 .406 .737

I included the AL median average for park factor and the mean average performance for AL batting average, on-base, slugging and OPS. The Park Factors come from ESPN. The Opponent Home/Road batting stats come from Baseball-Reference.

TRICK, TRASH OR TRIVIA? In business, as in baseball analysis, it's emotionally easiest to presume that something that happened happened because the actor/department/company we are looking at and associate with the result is the cause of the outcome. But even when the numbers we're looking at are legit, not BP or Merrill Lynch-quality scam, analysts can overlook the obvious.

This doesn't mean Fister is doomed to collapse. Each of the analysts posed reasons to believe his numbers might not hold up, but it's critical to note none of these clever gents caught the frozen Wooly Mammoth in the room, factors external to "luck" or Fister's own possible changes -- the immutable fact that all 6'8" of Merced's finest hurler has been facing twice as many below-average lineups and that every single appearance he's made has been in a ballpark that has dampened offense this year. Apparently everyone has been so focused on their fancy (and admittedly very interesting) third-order minutiæ that the sixteen-wheeler careering down upon them is overlooked.

Context is recognizing the differences between the norm that will even out over time, and the exceptions to the norm. Good management in baseball and in business can take advantage of the differences they recognize. Has the Mariner management team intentionally protected Fister, or is the nine-game run for facing weaker-than-average opponents, or pitching only in offense-dampening parks (or both at the same time, as has happened in six of nine cases) just the way the schedule has lined up?

If this was an American business, I'd be agnostic. Because it's Baseball, where the practice of management is significantly superior, my thinking is leaning towards the possibility that it's intentional in part or full. And there's a possible side-effect that if Fister builds up his confidence and experience with success, the emotional challenges he faces when (almost inevitably) he has to face tougher teams in tougher parks for pitchers, his abilities may enable him to produce outcomes closer to his 2010 performances to date than anyone one of us would have expected from his history and scouting reports.

Pay attention to context. Don't allow your deep collection of numbers and your deep knowledge of numeric analysis fool you into overlooking the obvious.

This page is powered by Blogger. Isn't yours?

free website counter