Ben Weinrib is a junior at Vanderbilt University on the Fred Russell-Grantland Rice Scholarship for sports writing. He is the founder of the Knuckle Blog and writes for the Charlotte Observer, the Vanderbilt Hustler, Bobcats Baseline, the Dirty South Sports Report, and Rivals. Weinrib has also published seven annual MLB previews.
Between the 2012 and 2013 seasons, the Cincinnati Reds had essentially the same roster. Yes, oft-injured Scott Rolen retired and they swapped Drew Stubbs for Shin-Soo Choo, but the vast majority of the personnel was the same.
But despite the similarity in teams, the outcomes of the past two seasons were drastically different. While the Reds were second in baseball with 97 wins two seasons ago, they dropped to 90 wins—just third in the NL Central—the very next season. Those seven wins were the difference between winning their division and having to play in the wild card playoff, which they ended up losing.
So what was the difference between the two seasons if the players were effectively the same? One could even make the case that they had a more talented team in 2013 considering how much better Choo is than Stubbs. The answer lies in the fact that wins aren’t the end-all be-all for determining how good a team is.
In fact, they’re far from it.
When it comes to evaluating teams, most people boil down their analysis to a quote used by Hall of Fame football coach Bill Parcells: “You are what your record says you are.”
To be fair, there is some truth to that quote. The team that wins the championship is the team that plays the best, not the team that is the best, and that’s an important distinction. But when you’re looking to predict what will happen in the future, you should look at the past process, not the results (wins).
I’m here to tell you that wins aren’t the best way to evaluate a team. If the goal is to figure out which team is the best—that is, the team that produces at the highest level—then there are several better stats to look at.
The truth is that a team’s record is far from a perfect indicator of how good the team was. For a team with a given talent level, there is a wide bell curve of possibilities for how a season could unfold. For instance, many people picked the Blue Jays to win 90 games in 2013 because they were incredibly talented, but they only ended up winning 74 games because of injuries and players not playing up to their normal level.
Even in a single season, given the number of runs a team scores and allows, there is a bell curve of possibilities for how the season could unfold. The 2012 Orioles only scored seven more runs than they gave up, yet they were a shockingly good 93-65.
That’s why I’ve developed a system to determine what each team’s record should have been, which I have named RAWS for Retrospective Analysis of Wins System. (These sort of things always catch on better when they have an easily pronounceable abbreviation.)
RAWS has two main components it uses to determine what a team’s record should actually have been: the team’s total production and the sum of the players’ production. Those two parts take form through run differential and team WAR.
Bill James came up with the idea of Pythagorean Record many moons ago, and the idea is that there is a non-linear relationship between runs scored, runs allowed, and wins. The equation for Pythagorean expectation is as follows:
(runs scored2) / (runs scored2 + runs allowed2)
That formula has since been updated to match empirical results, and the current exponent used is 1.83.
Runs are the best way to measure how productive a team was because, well, runs win games. Discrepancies between actual records and Pythagorean records often come about because of unsustainably good play in close games, like when those 2012 Orioles went 16-2 in extra innings games and 29-9 in one-run games. Despite their 94-68 record, the O’s had an 82-80 Pythagorean record, and largely thanks to regression to the mean, Baltimore went just 85-77 the next season.
Another signal that a team is headed toward regression to the mean is with unsustainable play like the Cardinals had last year with runners in scoring position. They were head and shoulders above all other teams with a .370 wOBA and .377 BABIP when the next closest teams had a .344 wOBA and .321 BABIP. That type of outlier play isn’t factored into Pythagorean Wins, but it is factored into the second half of RAWS’ formula.
WAR is an incredibly useful stat in terms of describing the value of a single player. It’s also incredibly useful when it comes to projecting wins and losses.
Glenn DuPaul’s 2012 study showed that a team of all replacement-level players is projected to win 52 games with each additional team WAR supplying an additional win. Thus, we can project what a team’s record should be based off the sum contributions of their players.
By averaging the two winning percentages (the one from Pythagorean Record and the estimate through WAR) and adjusting it so that the league-wide average winning percentage is exactly .500 (snipping off about two wins over a full season), we arrive at the record that RAWS projects the team should have had considering its overall production.
When using RAWS, it’s crucial to remember what the record it spits out actually means. RAWS is not a predictive tool; the fact that it said the Red Sox should have had a 102.7-59.3 record in 2013 does not mean they are projected to win 103 games in 2014. The record RAWS gives us only tells us the record the team should have had last year.
But although RAWS itself does not predict future records, it can still be used as a baseline to help predict the future.
Take, for example, the 2014 Detroit Tigers. It’s almost impossible to deny that Detroit has less talent this year with Prince Fielder, Doug Fister, and Jhonny Peralta gone, Jose Iglesias out for most of the season, and Ian Kinsler and Joe Nathan the only notable additions. Since they won 93 games last year, one could logically assume that they may only be a 90-win team in 2014.
But if we use RAWS to evaluate how good the Tigers actually were last year instead of using their 2013 record, you could knock a few wins off their 100.8-win projection instead to logically assume that they’re only a 98 win team. RAWS isn’t a predictive system, but it can be used as a baseline for how good a team was before.
Still in search for a reason why the Reds dropped seven games in the standings with essentially the same roster, an answer appears if we look at RAWS. Despite their lofty record, the Reds only had the production of an 89.8-72.2 team in 2012. Their 7.2-win difference between actual and projected record was the largest gap in the league other than that of those lucky Orioles.
On the other side of the coin, we have the Cardinals. Although they only won 88 games to claim a Wild Card spot in 2012, they were the third-best producing team according to RAWS and had a projected record of 93.8-68.2. Of course, they would go on to win 97 games and the NL Central crown the following year.
The Reds and Cardinals are a perfect example of how to properly apply RAWS in analysis. Although their records would indicate that the Reds were a much better team, the underlying numbers hint that the Cardinals produced better. And since both teams effectively had the same rosters in both 2012 and 2013, it would have been logical to predict the Cardinals to have a better season than the Reds in 2013 by using RAWS.
However, RAWS is far from perfect when it comes to projecting ahead, mostly because it is not a predictive measure. It doesn’t factor in anything to do with the future like outlier player performances, injuries, or transactions. But what it does explain well is how teams actually performed in the past. Used within the context of what it actually represents, RAWS can be very helpful when it comes to making predictions because teams are often not as good (or bad) as their record says they are.