Does Terry Collins rest his players efficiently?

Since becoming the Mets’ manager, Terry Collins has used his bench in a very interesting way. Specifically, he tends to rest his starting position players in bunches. This year, in seven of the 11 games in which Collins started his backup catcher against a righty, he also started a backup third baseman.

In last Wednesday’s 1-0 loss to the White Sox, for example, Collins not only started Rene Rivera behind the plate over Kevin Plawecki, but also Ty Kelly at third over Wilmer Flores, and Juan Lagares in center over Yoenis Cespedes. Is Collins right to start multiple bench players at a time, or should his starters’ rest be more spread out? A quick look at win probability metrics suggests that it actually doesn’t matter all that much.

Win Probability Added (WPA) measures how much a player’s actions increase his team’s chances of winning a game. If a David Wright home run increases the Mets’ chances of winning from 50.0% to 60.0%, Wright is credited with +0.10 WPA. WPA results are not predictive year-to-year because much of the stat is dependent on context (e.g., the score of the game, the runners on base, and how many "big" opportunities a hitter gets). That said, it’s clear that good players tend to have high WPAs, while weak players do not.

By using WPA, we can estimate that every time the Mets start a bench player over one of their everyday starters, they reduce their chances of winning by about 2.0%. Last year, for example, Wright had a 0.28 WPA in 38 games, good for a 0.007 WPA per game. In other words, the Mets had about a 0.7% better chance of winning a game when Wright played than they would have had an average big leaguer started in his place. Flores, meanwhile, had a -0.40 WPA in 137 games, costing his team about 0.3% of winning each game. Therefore, assuming that the separation between Wright and Flores is about what it was last year on a per-game basis, starting Flores over Wright reduces the Mets’ chances of winning a game by about 1.0%.

Travis d’Arnaud, meanwhile, increased the Mets’ chances of winning each game by about 2.0%, whereas Kevin Plawecki reduced those chances by about 1.1%. Therefore, starting Plawecki over d’Arnaud reduces the Mets’ odds of winning a game by around 3.1%. As far as the Mets’ current situation behind the plate goes, Plawecki (-1.1% per game) has about a 1.2% win probability advantage over Rene Rivera (-2.3% per game).

It seems reasonable to put the difference between a starter and a backup somewhere in the middle of these figures, at around 2.0%. This is line with the separation between the Mets’ starting and backup outfielders. Last year, Michael Conforto, Yoenis Cespedes, and Curtis Granderson increased the Mets’ chances of winning each game by about 0.5%, 1.5%, and 2.5%, respectively. Juan Lagares reduced the Mets’ chances by 0.4%, while Alejandro De Aza cost his teams 0.2% per game. Again, a 2.0% drop-off in WPA from a starter to a bench player seems to be a good ballpark estimate.

Given that a team decreases its chances of winning by about 2.0% every time it rests a starter, let’s see how the timing of that rest impacts a team’s chances of winning. The 2015 Mets finished 90-72, good for a .556 winning percentage. The 2016 team has a slightly better .558 winning percentage based on its 29-23 record. To make things easy, let’s assume that this year’s Mets will win about 90 games, meaning that they should have about a 55.6% chance of winning any given game against an average, standard opponent.

Now let’s say that the Mets enter a three-game series against an average opponent, and have three players they’d like to rest. Remember, every time they rest a player, they reduce their chances of winning by 2.0%. If the Mets rest one player in each game of the series, they enter each game with a 53.6% of victory, meaning that they are expected to win 1.61 games in the series (.536+.536+.536).

If, however, they rest all three players in one game, the Mets reduce their odds of winning that game to just 49.6%. Their odds of winning the other two games are 55.6% each, meaning that they are expected to win an identical 1.61 games (.496+.556+.556).

Intuitively, this makes perfect sense. To take an extreme example, if a team has a 99.0% chance of winning two games and a 1.0% chance of winning a third game, it would have the exact same win expectancy (1.99 wins over three games) as it would if it had a 66.3% chance of winning all three games.

The implications are pretty clear. First, while Collins is unusually aggressive in using multiple bench players at a time, his strategy does not affect the Mets’ chances of winning in any significant way. That said, Collins—and Mets fans in general—have to accept the fact that this strategy essentially involves "giving up" games. By starting three bench players in a single game, as he did on Wednesday, Collins can’t be surprised when his team scores just one run in 13 innings.

In other words, the sequence and timing of rest comes down to a matter of preference. Would you rather be competitive in all three games, or very competitive in two and less competitive in one? Collins has clearly made his choice. It’s not necessarily a bad choice, but an interesting one to point out.

So far, we’ve only considered Collins’s use of his bench against right-handers, in order to avoid the complications that arise with platoons. But his strategy against lefties has been similar. Collins has worked Wilmer Flores (when healthy) and Juan Lagares into the lineup virtually every time the Mets have faced a left-handed starter, which makes perfect sense from a platoon standpoint.

He also uses those opportunities, however, to start players like Eric Campbell, Ty Kelly, Rene Rivera, and Plawecki (when d’Arnaud was healthy), none of whom offers any platoon advantage over the players they replaced. What that shows is that, when Collins goes to his bench, for whatever reason, he likes to do it en masse. Again, that’s not necessarily a bad thing, but a unique aspect of Collins’s managerial style that doesn't get much attention.