The currency of baseball is wins. The ultimate goal is to win enough games to make the postseason and then win enough games in the postseason to win a World Series. For that reason, we care a lot about what leads to wins and losses, and outscoring your opponent is the only path to victory. This is all pretty obvious, but if we unpack it we stumble on to some pretty important realizations.
Before we go anything further, this post stays at 30,000 and serves as an introduction to Pythagorean Record and Base Runs. I won’t be going into the details of the exact formulas, but rather why these statistics are useful when looking at the team level. If you’re already well-versed in the various expected records, there probably isn’t a lot of new information below.
The problem with wins and losses
After 50 games, if a team is 30-20, are they a good team? They’ve won 60% of their games, so if they keep doing that, they’re going to have a great season. But are they a good team? Can you tell if they’re good only from their record?
The answer, most of the time, is no. This is especially true in small samples. For example, if a team goes 2-8 in their first 10 games, they probably aren’t a true talent 32- 130 team. If a team starts 10-20, are they a true talent 54-108 team? Almost certainly not.
Anyone who has ever watched baseball knows that a team’s record early in the season isn’t a clear indication of their true talent. Maybe you think a 2-8 team isn’t capable of winning 100 games, but you know they’re going to win more than 32. From here, you can infer that wins and losses through a given date are not necessarily reflective of a team’s true talent or even their true performance. This is true for many reasons, such as variable scheduling, luck, sequencing, and all sorts of other issues. A good team can lose a bunch and a bad team can win a bunch. The goal is to win games, but winning games through a certain date doesn’t tell you the whole story about a team. Is there something that offers more?
You may not know Pythagorean Record by that name, but you’ve likely heard about run differential. Run differential allows you to calculate a team’s Pythagorean Record. The idea here is that how many runs a team scores and allows is a better reflection of their ability and performance than their W/L record.
Imagine two teams. Team A has played five games with the following outcomes: W (10-0), W (5-0), W (6-2), L (3-4), L (3-5). They are 3-2 while scoring 27 runs and allowing 11. Imagine Team B: W (2-1), W (3-2), W (4-3), L (1-8), L, 2-7). Team B is 3-2 while scoring 12 runs and allowing 21. Are they equal? By W/L record, they appear the same, but Team A has won by a lot of runs and narrowly lost while Team B has narrowly won and lost by a bunch.
I’ve used a five game example for clarity, but this can happen over 50 or 100 games as well. And going forward, teams are more likely to play like their run differential than their W/L record because those runs scored and allowed are less noisy than the individual games. But they aren’t completely without noise. If you score a lot of runs and don’t allow very many, you’re going to win, but that doesn’t mean run differential is perfect either.
Enter Base Runs. It’s relatively easy to convince someone that a team winning by large margins and losing by small margins is better than a team with the same record who wins by small margins and loses by large ones, but it’s harder to convince someone that those runs themselves are a somewhat imperfect measure of performance as well. If a team wins 5-1, is it really possible that the team that scored one run actually played better?
Consider this inning: Single, walk, home run, ground ball, strikeout, fly out. In that order, the team scores (or allows) three runs. Consider this inning: home run, single, walk, ground ball strikeout, fly out. The team scores (or allows) one run. Those are the exact same plate appearances just in a different order, yet the the run output is totally different.
Expand that over a full game, and you could have a team that gets five doubles in one inning (probably at least four runs) or scatters them over nine innings (maybe zero runs). The key here is that there’s no evidence that teams can control when they get their hits and walks. A good team gets on base more than a bad team, but clustering their hits together for six weeks doesn’t mean they’ll do it for the next five months. It’s one thing to say you don’t believe a team will keep producing a .350 wOBA, but it’s another to think a .320 wOBA team is better just because they happened to have scored a few more runs over 40 games.
Base Runs helps here because it takes into account a team’s performance without considering the sequencing to calculated expected runs scored and runs allowed, and then takes those numbers to generate expected wins and expected losses. There are a lot of ways to get to 4-1 in the standings, but the team’s expected run differential is going to be a better predictor of future success than that 4-1 record going forward.
Putting it all together
Wins and losses are driven by runs scored and runs allowed within individual games, but runs can be clustered unevenly throughout season, making runs scored and runs allowed on their own a better representation of true performance than wins and losses. But we also know that runs scored and runs allowed can be misleading because hits and walks can be clustered unevenly within a game. You can look at the actual records, the records based on runs scored and allowed, or the records based on expected runs scored and runs allowed. We have a handy page at the site that lets you put all three side by side.
The various records tell you different things. Obviously, the wins and losses are the ones that count, but if you want to get a better sense of how well the teams have actually played, Base Runs is probably the best bet. Comparing those two measures can be very interesting, because it tells you which teams have gathered extra wins and losses based on sequencing.
You’ll see our writers talking about Base Runs quite a bit. No one would every argue it’s a perfect estimator, only that it’s usually a better tool than Pythagorean Record and almost certainly better than the W/L record if you’re interested in underlying performance. The key, however, is that all of these records are based on the actual results on the field to date. Base Runs is not a projection. It does not take into account other information about the players or teams, it merely considers how a team would typically perform if their hits, walks, extra base hits, etc were normally distributed.
In the future, we’ll get into exactly how all of these are calculated, but hopefully this gives you a better understanding of exactly what it means when we talk about a team’s record compared to their Base Runs record to date.
Neil Weinberg is the Site Educator at FanGraphs and can be found writing enthusiastically about the Detroit Tigers at New English D. Follow and interact with him on Twitter @NeilWeinberg44.