What Has Worked in the Postseason by Eno Sarris November 4, 2014 Billy Beane once famously said that his poop doesn’t work in the postseason, and ostensibly he wasn’t talking about his digestive system. The Athletics will spend another offseason wondering about it. The rest of baseball’s fandom usually thinks about the victors as models. Do the Giants and Royals represent some sort of sea change, do they represent a way to succeed in the postseason? I thought I’d run some numbers to see what I could find. What I found is — it’s very difficult to study the postseason as an entity. The first idea was to compare post-season and regular season correlations between certain statistics and winning percentage. The strength of the correlation would only be important as it pertained to the other statistics, in other words. I’m less interested in how strong the relationship between walks and wins is, as I am interested in how much more or less important strikeouts are to wins. A ranking of team skills, in essence. For the regular season, using records back to 1974, that was easy. All of the p-values were tiny (less than .00001), and all of the relationships existed on a continuum that lines up with how many of us think about baseball. I indexed the rate stats to the league rate stats in every year — the average strikeout rate has undergone many changes in the last thirty-plus years. Stat Correlation to Win% K%+ 0.044 BSR 0.061 BB%+ 0.151 ISO+ 0.217 wRC+ 0.453 oWAR 0.571 By leading the league in strikeout rate and stolen bases, the Royals made it to the postseason using excellence in the two stats that correlate very weakly to win percentage. It was an improbable run, though, and we knew that. Did it become more probable because of those same correlations in the postseason? We can’t know using this type of analysis. The strongest correlation between regular season team stats and postseason winning percentage belonged to offensive WAR, and that only described 1.3% of the variance in the sample. But low r-squared numbers aside, the real problem is the size of the sample itself. 246 points of data are just not enough — the lowest p-value in the et was .081 (for WAR). The other p values just suggested that the sample was random. Which it is. Even for the championship winners in the dual wild-card era, we’re talking about 15 games or so. Remember when the Athletics started the season 11-4? That would have won them the World Series in October. Then a slightly different version of the same team went 3-12 in late August and early September. That almost cost them a chance at the postseason. Perhaps that’s it, full stop. Maybe I should stop there. But I spent a lot of time collating the postseason data, so I didn’t stop there. I took the top 28 of the 246 postseason teams — the teams that won 2/3 of their games, most of them were champions — and compared their regular season work to the work done by the entire sample of postseason teams. BSR wRC+ WAR K%+ BB%+ ISO+ top 28 ave 1.92 104.06 27.12 96.52 105.90 105.90 sample ave 1.78 103.48 25.95 97.48 105.75 106.63 For the most part, this statistically unreliable approach just gives us the no-duh answer. Better teams are better. They have better weighted offensive numbers, they walk more, they strikeout less. Sure. But there are two things that stood out to me. Probably just confirmation bias, but look at that last column. Postseason winners did not out-slug their opponents once they got to the postseason. Also, even though their regular season walk rates were a little better than other postseason teams, the gap has been very narrow. In fact, the champions set was — relative to other postseason teams — much better at making contact compared to their league. We know it can’t mean much, but it also does seem worth further thought. Is the postseason any different from the regular season? If not, then power and patience still rule the roost. If it is, though, perhaps collecting the lottery ticket that is the ball in play becomes a more viable strategy once you make it to October.