There is one very important lesson that all baseball fans need to understand about the game regardless of their level of fandom and feelings about analytics: Don’t overreact to small sample sizes because variance swamps everything.
It’s tough for people to accept that. It’s human nature to want an explanation for everything, even when the most likely explanation is randomness. After all, we have evolved due to our excellence in recognizing and assessing patterns.
The media doesn’t help with this cause because professional writers can’t exactly take the early season off and write "randomness"; some of them will produce content that draws conclusions from small sample sizes. Popular topics along these lines are teams that are early surprises or early disappointments. However, more often than not, true talent rises above the statistical noise over the course of the season. This is better known as regression to the mean.
There are quite a few teams that have gone in different directions after a hot or cold start to the season, but in this piece, I’ll focus on the Phillies and the Astros due to their predictably divergent paths.
At the beginning of the season, nobody was exactly expecting the Phillies to light the world on fire in 2016. We all knew they were rebuilding and were likely to be one of the worst teams in baseball. FanGraphs had them projected as a 64-win team, the worst projected record in baseball. However, ignoring the first couple of weeks in April, the Phillies didn’t fall under .500 until June 1st. They were a half-game out of first place as late as May 18th! There were even rumors that the Phillies could be buyers at the trade deadline!
Since June rolled around they’ve gone 22-33. So what happened? Nothing happened. They’re the same team. Just regression to the mean hit. They were 14-5 in one-run games, and that was without a good bullpen. They are 6-9 in one-run games since then. If one were to do a deep dive into the Phillies first two months, we’d likely find some combination of BABIP luck and players with unsustainably good performances in high leverage situations; some writers have gone into this in detail so we don't need to rehash it here.
It’s hard to find somebody who really believed in the Phillies in May. Writers and fans are getting smarter. For the most part, people realized the team was over-performing and that the Phillies’ true talent was far below what their small sample size record indicated. I’m not so sure that something similar could be said about the Astros.
My perception was that Astros fans were panicking in May. Some writers were quick to criticize the team’s flaws and failures. Of course, that isn’t necessarily unfair. It only becomes unfair once anybody implies that a team is completely different than what we thought they’d be based on one to two months of data, as if the players’ track records before then mean nothing. It’s odd how people are more willing to trust a bad team’s true talent but not that of a good team’s.
The Astros were in last place as late as May 25th, and were as high as ten games out only a few days before. Since then, the Astros have gone 36-21 and are in second place in the AL West. Though they still remain six games back of the Rangers and two games out of the Wild Card spot, they've overcome most of the deficit they were in just a couple months ago. According to FanGraphs, they have a 44.6 percent chance of making the playoffs. That number was at 22.9 percent on May 25th. It’s almost as if the Astros were a good team all along!
The Astros are no more a .631 win percentage team since May 25th than they were a .404 win percentage team on May 25th. What they are is the .537 win percentage team that FanGraphs have them projected for the rest of the season. By the way, the win percentage that was projected for the Astros before the season started is .541. Imagine that.
I won’t cover the Astros’ struggles in detail, but it’s well known that their starting pitching was the main culprit. Yes, the bullpen did have a terrible April with a 5.01 RA9 and 4.36 FIP. The runs allowed were mostly the result of a flukishly high 22.1 percent HR/FB ratio. But the regression monster gives as much as he takes away! Since the calendar turned to May, the Astros bullpen has had an outstanding 2.76 RA9 and 2.58 FIP. Their 3.54 DRA for the season is second only to the Yankees, who are likely to drop in that ranking after trading away Aroldis Chapman.
The biggest cause of the Astros initial struggles was that nobody in the starting rotation could throw hard or strike anybody out. Recalling Lance McCullers helped to remedy that, but so did Dallas Keuchel starting to pitch more like his 2015 self. He did get lit up in his last start, but he is still pitching better in general.
During the team’s struggles, Keuchel had a horrendous 5.92 RA9. He did have a 3.99 FIP, though FIP has to be taken with a grain of salt when a pitcher is allowing so many runs. Even his high .351 BABIP doesn’t completely explain an RA9 that high, especially when the team’s defense is actually very good. His 14.6 percent HR/FB ratio isn’t significant either because he has always had a high HR/FB ratio.
Keuchel’s 90 MPH fastball finally caught up with him. A pitcher can absolutely get by with a mediocre fastball. It just means he has a smaller margin for error. Well, Keuchel made a lot of errors. According to Brooks Baseball, this is how his fourseamer and sinker fared during the first two months of the season.
Here’s how those pitches fared in 2015.
The sinker is especially noteworthy because that’s the pitch he uses the most. He threw it 50 percent of the time last year and 45 percent this year. Interestingly enough, his sinker is still getting knocked around. Hitters are hitting .319 against it with a .504 SLG. Regardless, Keuchel has a 3.92 RA9 and 3.43 FIP since the start of June, despite continuing to be homer-prone. For the season, he has a 3.90 DRA and 88 cFIP. That’s good, but it’s a far cry from his 2.43 DRA and 68 cFIP last year. Keuchel is just a weird pitcher.
Despite their unexpected starts, the Phillies and Astros are right where they should be based on their true team talent. It’s important to always remember that two months offer incomplete pictures of teams and players because of the random fluctuations that are inherent to small sample sizes. If your team is under-performing, don’t panic! Talent doesn’t reset at the start of a new season. This is part of the reason why projections are important. They are good measures of a player’s and a team’s true talent. Now that I’ve said this, I’m sure nobody will ever overreact to performances through May ever again!
. . .
Luis Torres is a Contributing Writer at Beyond the Box Score. You can follow him on Twitter at @Chemtorres21