So some colleagues were discussing PEDs and the HOF, a topic which really doesn’t interest me much at all. (I say let ‘em in.) The inevitable amphetamine argument was broached, you know, Mantle used greenies, what’s the difference? This got one of my more quantitative colleagues to try and figure out if one could estimate whether or not amphetamines had an effect on batting stats. He formed a hypothesis which, for all its obvious infirmities, at least should point in the right direction.
He opined that (1) the effect of amphetamines should be most pronounced on day games after night games; (2) batting should be more affected than pitching, on the grounds that even Whitey Ford probably drank a little less the day before he was supposed to pitch; and (3) the effect should be bigger in the 1970-74 period than in the 2006-2010 period, since in the latter period there was testing; indeed, in the prior period I don’t even think it was illegal.
He then said: “Hey, JonathanF: you’ve put together a database of every baseball game ever played. Can you try it?” So I did. I really didn’t expect to see anything, so I was a little surprised.
Taking every game from those two periods, I compiled a simple TeamOPS number for every game (H+W)/(AB+W) + (H+D+2*T+3*H)/AB. I then compared the average teamOPS (simple averages here – you don’t do anything fancy when you don’t think you’re going to get anything) separately for day games after night games (which I called greendays) and all other games.
I did this year by year and got the following results. The “difference” column measures the team OPS difference between greendays and non-greendays. The bold results are statistically significant.
About 15 percent of the early games are greendays, and about 22 percent of the later games.
In all but one of the early years, TeamOPS was higher on greendays than non-greendays, and very nearly the same in 1973. Now that doesn’t indicate anything, since it is probably easier to hit in day games, and greendays are all day games while only some of the non-greendays are. But here’s the interesting part: in the later period, greendays were worse in three years out of five, and very, very close in the two other years, much closer than they ever were in the earlier period. Over both periods, there is almost no difference. Standard significance tests confirm that this result is unlikely to be due to chance.
Now that doesn’t mean it’s due to amphetamines, either. Off the top of my head I can think of quite a few possible reasons for this. But when someone makes a hypothesis, and you don’t expect to find anything, and you do find something the first place you look, it bears some thought. Mantle and Ford! Kick ‘em out! (Not really.)