The Rays Are Proving The Randomness Of 1-Run Games
Tampa Bay is only 11-10 in games decided by the slimmest margin.
(Note: This post was updated to reflect results from May 31.)
Yes, I’ve already written quite a bit this season about the dominance of the Tampa Bay Rays. But the team continues to have MLB’s best record despite a recent slowdown, and the Rays have pretty clearly been the best team in baseball according to the advanced metrics. Tampa Bay’s BaseRuns-predicted winning percentage (.682) is 14 points clear of the No. 2 Texas Rangers and 78 points ahead of the No. 3 Atlanta Braves, while they’re tracking for 10.4 more Wins Above Replacement than any other ballclub. (Texas is No. 2 at 57.6 WAR/162.) All signs point to this Rays run remaining as legit as ever was.
Well, almost all signs. I mentioned that recent slowdown — through Wednesday, Tampa was 6-4 in its past 10 games and 11-9 in its past 20. But within that 11-9 record since May 11, the Rays are 4-7 in contests decided by 1 run and 7-2 in multi-run games. That’s part of a theme for Tampa Bay this season: While the Rays had an MLB-best 29-8 record in multi-run games through Wednesday (good for a .784 WPct), and in fact were the only team in baseball with single-digit losses in those games, they were merely 11-10 (.524, 14th in MLB) in 1-run games.
Tampa Bay should be fine either way — an average of the projections at FanGraphs and Baseball-Reference sees the Rays winning 99 games with a 98% playoff probability and a 15% chance to win the World Series — but they’d be up by a lot more than just 4 games in the AL East if they had better luck in those close games. And by and large, it is luck.
We can see this when we test what is more predictive of a team’s performance: its record in 1-run games, or multi-run games? Looking at all full seasons since the 1994 strike, I sliced up each team’s record within several different splits — 1-run versus multi-run games, and even-numbered versus odd-numbered games. By doing this, we can take a team’s record by margin type within one half of the season (even or odd) and use it to predict how the team did in the opposite half of the season.
For instance, during odd-numbered contests, the 1996 Atlanta Braves were 14-16 (.467) in 1-run games but 33-18 (.647) in multi-run games. So which was more predictive of their record during the other half of the season? They went 49-32 (.605) in even-numbered games, so obviously the multi-run performances told us a lot more about their true quality than the 1-run performances.
That’s not just a cherry-picked example. (The ‘96 Braves literally were the first team that came up alphabetically in my dataset.) In our entire sample, the correlation between a team’s winning percentage in multi-run games within one half of a season and its overall winning percentage in the opposite half is 0.565; the same cross-season correlation for 1-run games is just 0.217. If we wanted to predict a team’s overall opposite-half record, we’d have to give 5.7 times as much weight to its multi-run record as we gave to its 1-run record.
In other words, although 1-run games don’t exactly carry zero information, you can mostly ignore them if you want to safely predict how a team will do going forward.
That’s for all teams in a general sense. Naturally, there are some structural reasons why the Rays in particular have struggled in close games. Their OPS is 7% lower than usual in late and close situations, while their bullpen is third-to-last in WAR and eighth-to-last in Win Probability Added. Those factors probably make them slightly more inclined to underperform when games come down to the wire.
But mostly, it’s just luck that explains why a team that goes 29-8 in decisive outcomes is only 11-10 in close contests. Baseball is a funny sport sometimes, but there are worse places to be than having luck be the only thing holding you back.
Filed under: Baseball