Just like Vegas, math wins in the end
Kansas City Royals defying odds — in the short term
After the top of the eighth inning of the wild-card playoff game between Kansas City and Oakland on Sept. 30, the Royals trailed 7-3 and had a win expectancy of three per cent.
They haven’t lost a game since.
On Wednesday night in Chicago, on the much smaller stage of a regular-season NHL game, the Blackhawks outshot the Calgary Flames 50-18. And, the Blackhawks lost 2-1, on a Mikael Backlund goal in the final minute of overtime. On that same night, Boston outshot Detroit 39-20, not as outrageous but still quite decisive, and only managed to eke out a 3-2 win that was decided in a shootout.
Nothing ties these events together other than that they illustrate the triumph of randomness. But, coming as they do after a hockey off-season in which more attention than ever has been paid to the merits of math, games that turn probability on its ear are a useful reminder that arithmetic plays the long game.
After the second period of Flames-Blackhawks, with Chicago ahead 34-8 in shots but trailing 1-0, I checked Twitter, in the same way one glances sidelong at car crashes. Sure enough, there were enthusiastic claims that the game was sound evidence that the leaguewide push toward placing a higher value on the various flavours of shot differential was a fool’s errand.
Presumably, there was similar crowing about the tie game in Detroit despite Boston’s dramatic advantage in shots and, a night earlier, the fact that the Toronto Maple Leafs, with their new focus on advanced analytics, were losing to the Colorado Avalanche until late in the game while outshooting them handily.
This seems like the kind of thing that doesn’t require saying, but we’ll say it anyway: There are always statistical outliers. No one particular statistic, particularly in sports, where there always loads of variables at play, is going to work as a foolproof predictor of outcome. If it did, we could all go to Vegas and retire next week on the winnings. Over time, though, randomness will be filtered out, statistical probabilities can be gleaned, and you will be left with results that match what the underlying statistics tell you to expect.
This is pretty much the argument at the heart of the statistical revolution that has moved through professional sports, first with baseball and now with basketball and hockey: it says that teams will be successful if they prioritize those elements of their sport that have the strongest correlation to winning. Over time, that correlation will eventually win out, even if it clashes with traditional thinking. In hockey, that means an increased emphasis on players who control the puck and the play, and a lesser role for certain kinds of sandpaper guys who might as well be playing with a ringette stick given how little they touch the puck. In basketball, it means attempting more shots from beyond the three-point arc or from close to the basket, and eschewing the longrange two-point shot. And in baseball, it means favouring hitters who can get on base and hit for power, while putting decreased value on small-ball tactics like bunts and stolen bases.
In all cases, small sample sizes can kick out aberrations. Flames goalie Jonas Hiller made 49 saves on Wednesday, but it would still be overstating things to simply say Chicago ran into a hot goalie. They also ran into bad shooting; Dan Carcillo alone whiffed on two open nets. Various people have tried to figure out the point at which random events are smoothed out over the course of a season, with differing results, so there is little certainty in the matter. We do know that the Maple Leafs were able to fool science in 2013, because the lockoutshortened season ended before they were overcome by their tendency to be outshot by the bushel. They made it through half a season or so last year before those same poor statistical traits eventually caught up with them. But even a whole season doesn’t necessarily tell the tale: Colorado punched above its shot differential for all of last year, finishing second in the West, before losing in the playoffs. New Jersey, despite a good shot differential, couldn’t overcome a string of shootout losses and didn’t make the post-season.
Which brings us back to the Kansas City Royals. There will no doubt be much said in the coming days about how the American League champions have shown that the obituaries for smallball were premature. Stolen bases! Sacrifice bunts! Banjo hitters! And, oh, so very gritty. But, really, they are the Kansas City Randoms. Thanks in part to some baffling managerial decisions from Ned Yost in that wild-card game, they found themselves with a 3-in-100 shot, based on the probability of past events, of advancing beyond the first day of the playoffs. Even after they tied that game in the bottom of the ninth, they fell down again in the 12th and, at that point, had only an 11 per cent win expectancy. If things proceed as usual at either of those points, the Royals lose and we would conclude, as a footnote, that a team that didn’t have one player with more than 20 home runs in the regular season was never going to amount to much in the high-stakes playoffs.
Instead, statistical noise won out — as it does sometimes. You still want players who control the puck, and shoot threes, and hit home runs. These are the high-percentage plays. Math doesn’t care for your grit, or your traditions. Math, more often than not, will win in the end.