Expected Run Differentials 2.0

Over the first couple of months of the season, I’ve done a couple of end-of-the-month posts on Expected Run Differentials. While pythagorean expected record — the number of wins and losses a team would be expected to have based on their runs scored and allowed — has become nearly a mainstream concept, I’ve never been a huge fan of using runs to determine how well a team has played thus far.

After all, the entire point of looking at run differential instead of actual wins and losses is because we’re acknowledging that wins and losses are affected by the timing of when runs are scored or allowed, and history has shown that run sequencing is mostly just randomness. So, developing an expected win-loss metric that removes the affects of sequencing is a good idea, but pythagorean record only goes halfway to that goal. It removes the timing aspects of converting runs into wins, but ignores the timing aspects of converting baserunners into runs. Evaluating a team by its run differential removes some of the sequencing effects of wins and losses, but leaves plenty of other parts, with no real reason why we should arbitrarily include some sequencing while taking other parts out.

That’s why I’ve always preferred to look at a team’s performance based on expected runs scored and allowed, rather than actual runs scored and allowed; this gives us the most context-neutral evaluation of team performance to date. In the two preceding posts, I walked through the creation of expected runs scored and allowed totals based on each team’s wOBA and wOBA allowed, adjusted for baserunning and fielding values. As a linear weights based metric, wOBA is a very good context-neutral evaluator of individual events.

However, as Jesse Wolfersberger eloquently illustrated at The Hardball Times last week, run scoring at the team level isn’t really linear.

The exponential nature of offense means a good hitter in a good lineup is worth more than that same hitter in a bad lineup. On a good offense, that hitter is more likely to come to the plate with more runners on, more likely to get driven in once he’s on base. And, the lineup turns over more often, meaning he gets more plate appearances. Not only is he more valuable to a good lineup, but he’s even more valuable to a better one – the effect builds on itself.

While the wOBA-to-runs conversion works well in most cases, it does begin to break down a bit at the extremes, where the non-linear effects of team strength can come into play. And while these extreme examples still don’t change the conversion much, they can add up over a full season.

A team with a .365 wOBA would be predicted to score about 5.85 runs per game, but will actually score about 5.93 runs per game. On the low end, a team with a .285 wOBA would be predicted to score about 3.24 runs per game, but would instead score about 3.33 runs per game. Those are small differences, but remember that baseball has the longest regular season of all major sports. A difference of .09 runs per game equals about 14.6 runs per season, or about one-and-a-half wins.

Given that our sister site just published an article explaining how wOBA can break down a little bit at the team level, I figured that continuing to use wOBA to create the Expected Run Differentials for the monthly posts was probably not the best idea, especially if there was a better alternative. And there is.

The model is called BaseRuns. It’s significantly more complex than a linear weights model like wOBA, but that complexity leads to estimates that fit each team’s own run environment; if a team has a very good offense, the extra value will be captured in BaseRuns when it is not in wOBA. If you’re particularly interested in how BaseRuns works, this is a good primer. If you don’t care about the how and just want to know that it does work, however, research supports the idea that BaseRuns is probably the most accurate run estimator in the public domain.

So, rather than continue to create good-but-imperfect run estimations based on a linear weights model, I prodded our Dark Overlord and said “hey, we have BaseRuns in the database; can I have them please?” And being the benevolent overlord that he is, he did me one better; not only did he give them to me, he’s given them to us all.

On our updated 2014 BaseRuns Standings page, you will now find three columns of year to date data: Actual win/loss and runs scored/allowed totals, the pythagorean expected record based on those runs scored and allowed totals, and finally, the expected runs scored and allowed (and the corresponding win/loss totals) calculation based on each team’s BaseRuns estimate. Essentially, this standings page could be read from left-to-right in descending order of context.

The left-hand side includes all events and the sequencing of those events, giving us the totals that actually count in the Major League standings. The middle columns give you a reduced-context win estimate based on actual runs scored and allowed, retaining the sequencing that turns baserunners into runs. The right-most part of the table is the metric that is as context-neutral as you want to get at the team level, accounting for the non-linear nature of run scoring without giving teams extra credit for bunching their hits together above a reasonably normal expectation.

From here, you can check in every day and see the best estimate of how many teams your team should have won based on their context-neutral performance without having to wait for my end-of-the-month post to update the leaders. And because all the columns are sortable, you can easily see which team has had the best offense or defense, or the combination of both. Here are the BaseRuns numbers, as of this morning.

(Note: +/- is the number of wins a team has accrued relative to their expected wins by BaseRuns. A team with a +5 in that column has won five more games than expected, for example.)

Not surprisingly, the best team in baseball this year has been the Oakland A’s; it was that way on both of the previous two Expected Run Differential posts as well. The A’s are just trouncing their opponents, and no team in baseball has played better. But the A’s are also a good reminder of why BaseRuns is more useful to look at than pythagorean record, because their run differential suggests that they’ve actually underperformed this year. By pythag, they’ve played like one of the best teams in baseball, and have gotten “unlucky” to only be 48-30.

But that only tells half the story; they haven’t been great at converting their runs into wins, but they’ve been amazing at converting their baserunners into runs. They’ve been “lucky” one way and “unlucky” the other, and only looking at their run differential overstates how well they’ve played by including the “unlucky” part of sequencing while ignore the “lucky” part.

Perhaps a graph will be more helpful than a big table at illustrating these differences. Below, I’ve plotted every teams actual winning percentage, pythag winning percentage, and BaseRuns expected winning percentage on a marked line, so you can see where the variations are for each team. When the blue point is above the green line, you have a team that has won more games than expected; when the blue point is below the green line, that team has won fewer games than expected.

The big overachievers? The Brewers, Royals, and Yankees, who have each clutched their way into better records than they have earned based on the underlying hits, walks, and other ways of reaching base or advancing runners. On the other end of the spectrum, the Rockies, Cubs, and Rays have all played better than their records would suggest.

But perhaps maybe the most interesting data point to me? Look at the Rangers. They’re a well-known disappointment, but by BaseRuns, they’ve been the worst team in baseball this year. Worse than the Padres, who just fired their GM, and worse than the D’Backs, who hired an overseer to transition the organization down another path. The Rangers, who spent $140 million to sign Shin-Soo Choo and $136 million to trade for Prince Fielder, have been worse than their win-loss record and even their uninspiring pythagorean record. The 2014 Rangers have been atrocious.

Of course, a lot of that can be chalked up to injuries, and we’re still just dealing with a half-season of performance data. While BaseRuns is a very good run estimator, single season inputs still shouldn’t be taken as measures of true talent, and if the Rangers played the Astros in a seven game series next week, I’d probably still take the Rangers. But it’s not the slam dunk choice you might think, and an argument could be constructed that the Astros are not the worst team in the AL West right now.

The good news is that you don’t have to wait a month for the next Expected Run Differential update. With the data now right on our Standings page, you can check it each morning, and see where your team’s to-date performance stacks up against other contenders. When projecting future performance, you still want to account for more than just season-to-date numbers, including past track record for each player, roster changes, and future schedules, which is where the projections on our Playoff Odds page come in handy. But if you’re just wondering how well your team has played this year, and what kind of record they should have, you’re not going to do better than the BaseRuns expected record.