- NotGraphs Baseball - http://www.fangraphs.com/not -

A Cognitive Bias Revealed by WhatIfSports

Last Friday, in these electronic pages, I discussed the entirely fictional and computer-generated baseball league I’d recently joined — specifically, Aaron Gleeman‘s Hardball Dynasty league at WhatIfSports — and considered both (a) the significant pull of that fake baseball world on my powers of concentration and also (b) the subsequent emotional reaction (guilt, mostly) to that pull.

Four days later, the lure of the game hasn’t particularly waned. Were I to estimate how many hours I’d spent thinking about my entirely fictional team (the Burlington Aristocrats, they’re called) since last Thursday, I’d say four or five hours. I’d also be lying when I said that — like, by kinda a lot — lest my wife read this and inflict harm upon my person.

So far as the league itself is concerned, commissioner and real-live hermit Aaron Gleeman has managed to replace all the owners who departed after season 24, and now season 25 has officially begun. Among an owner’s obligations at this point are (a) the setting of the budget and (b) the re-hiring of the coaches. The first of these tasks isn’t unpleasant at all; the second is mildly tedious.

The budget for the author’s fictional baseball team.

Simultaneous to both of these events is the commencement of trade discussions among the league’s owners — and it’s to this point that I’d like to speak briefly.

It goes without saying — although I’ll say it, anyway — that, to begin the process of discussing potential trades, one must first understand the value both of his players and those of his various opponents’. Anyone who’s made the questionable decision of navigating his or her (although, most likely his) browser to this site probably has a pretty good sense of what basically every major leaguer is worth — like, what sort of contract he’d receive on the open market and what he’d fetch in a trade. Approximately, at least.

That isn’t the case in Hardball Dynasty, though. Individual owners have maybe developed methodologies for assessing value, but there is nothing so comprehensive as WAR available publicly. More than that, however, it isn’t even particularly obvious how a player’s various ratings correlate to his subsequent production. Part of playing the game, of course, and deriving joy from it, is discerning in which the manner the sim engine utilizes and weights the player ratings in question. It is a weird and giant logic problem, essentially, dressed in the trappings of baseball.

What one can’t do, however — but what the author finds himself constantly tempted to do — is assume that the various player ratings are worth in the HBD world what they are in the real one.

For example: it is overwhelmingly the case in major-league baseball that a pitcher with an excellent strikeout-to-walk differential (i.e. strikeout percentage minus walk rate) will be among the league’s best at preventing runs. This isn’t really the case in HBD. The best pitcher almost ever in Gleeman’s HBD world appears to be Ed Johnstone (pictured below). Johnstone, in his career, has pitched 2257.0 cyber innings and posted strikeout and walk rates of 6.6 and 2.6 per nine, respectively, while conceding 0.85 home runs per nine, as well — none of them dominant numbers. Yet, he possesses a 2.60 ERA for his career — lower than almost every other pitcher, and based very much on a .202 opponent batting average.

Ed Johnstone and Ed Johnstone’s computer-generated goatee.

While Johnstone doesn’t have an elite strikeout-to-walk differential, what he does have are platoon split ratings of 97 and 96 (out of 100) to left- and right-handed batters, specifically. Looking at other pitchers, it appears to be the case that these platoon ratings correlate quite highly to run prevention. A higher velocity rating does seem to produce more strikeouts and a pitcher with higher control rating will walk fewer batter. However, because the sim engine handles BABIP differently than in real life — appears, really, to treat BABIP suppression as a skill that becomes reliable rather quickly — and, because those platoon ratings seem to represent BABIP suppression at some level, they are very important.

So, to win the game, the player-owner must recognize the correlation of each rating category to run-scoring or -prevention (as the case may be). What he must also do, though — and this, I think, is more difficult — is abandon the very powerful notion that the laws of the sim world are the same as the laws of the real one. Instead of playing within a sort of shadow world of baseball, one is actually playing (in HBD and other, similar games) a very real type of other game — against both the programming of the engine and the other owners’ understanding of that programming.