October 1, 2015

Prospectus Feature

The Story of the Rest

by Henry Druschel

Ask a bunch of analysts or writers what the next frontier in baseball analysis will be, and some large percentage are going to say “health”—particularly if that bunch includes Billy Beane, who last week called injuries “the aspect of the sport that is most in need of better understanding.” Pitcher health (or the lack thereof) gets the headlines and the book deals, but the day-to-day ability of a position player to stay on the field is increasingly viewed as a skill. Health seems to present a giftwrapped opportunity for some clever team to gain a huge advantage.

For a while, there’s been a sense that the Pirates might be that team—if not that they’ve achieved perfect health, at least that they’ve gone ahead of their 29 competitors. Pittsburgh Tribune beat writer Travis Sawchik wrote about the Pirates’ use of the Zephyr BioHarness last March, a device that measures players’ caloric intake, heart rates, and other biological factors to understand their workload and fatigue. This spring, Jayson Stark of ESPN wrote that the Pirates were trying to emulate the NBA’s Golden State Warriors, who rested their top players more than their competitors and appeared to see better overall performance as a result. Another concept in that vein is “bundling,” introduced in this article by Sawchik. The idea is to cluster starting players’ days off, supposedly allowing them to better rest and recuperate, and to concentrate backup players’ starts, letting them get into more of a rhythm.

It all sounds great—and the Pirates’ season has, overall, thus far been great—but the extent and impact of all those strategies have been unclear. The Pirates have ranked fairly well in surface-level health measures, both in 2014 and 2015. That said, Ben Lindbergh, writing about Pittsburgh in Grantland last September, found that current “injury performance” was generally not a good predictor of future injury performance, making it difficult to conclude the Pirates have demonstrated a skill, or that they will continue to in the future.

All this leaves a few pressing questions. Have the Pirates actually deployed their players differently this year? And if they have, what impact has it actually had? Health is a tricky thing to quantify, so some creativity is in order.

To shed some light on the first question, I looked at the position player starters for each game in the 2015 season, beginning on May 5^th (the 31^st day of the season) and continuing through September 6^th. If the Pirates are actually “bundling,” this data should show fewer isolated starts for their players, and if they are resting more intentionally, it should also show fewer long stretches without a break, as compared to the rest of the league.

This following chart shows each start in the dataset made by a non-regular—a player who started fewer than 15 games in the previous 30 days—categorized by how many starts the player had made in the previous three days. If the Pirates are bundling and the rest of the league is not, their players (gold) should have fewer starts with zero games started in the last three than the rest of the league (grey).

That distribution shows the opposite of what we were expecting: The Pirates' reserves were less likely to have had their games bundled. Bundling, then, seems to be a dud. But that’s not to say the Pirates aren’t doing something different with their playing time patterns. In the next chart, we see each start (now including all players, not just reserves) in the dataset categorized by the number of games started in the previous 30 days by the player making the start. Pirates players are in gold and all other players are in grey.

Indeed, it looks here like the Pirates are doing something different. Starts by Pirates regulars are concentrated in what makes up the middle range of the rest of the league. If all starts league-wide (minus Pittsburgh) were laid out, beginning with those by players with zero starts in the last month and going up to those with 29, the starts in the 17-to-25 range would make up 54 percent of the total. Pirates starts in the same range cover 77 percent of their starts. Just one-seventh of Pirates starts make up the left end of their distribution, and less than a twelfth cover the right end—the ironmen side of the graph. The team with the next highest share of starts in the 17-25 range is the Royals, at only 70 percent. Plainly: Far fewer Pirates starters, by a large margin, have spent their last 30 days starting a very high or very low number of games.

So it seems like the Pirates do indeed have an observable strategy that is noticeably different from what the rest of the league is doing. It may not be bundling as we initially described it, but their top players are resting more frequently and their designated reserves are playing somewhat more often. Great. Easy. Much harder is capturing the effect. Per the website ManGamesLost.com, which tracks time lost to injury for a variety of different sports, Pirates players spent 580 total days on the DL through September 8^th, good for a middle-of-the-pack 19^th-best in the majors. The vast majority of even that slight advantage, however, comes from pitchers; Pittsburgh’s position players have spent 272 days on the DL, just 15^th. They’re average.

But even if we’re generous and conclude the Pirates have done well to keep their hitters off the DL, that measure alone doesn’t really communicate the health of a team in a significant way. First (and most obviously), not every injured player ends up on the DL. Taking a step back, though: Injuries hurt a team by keeping its best players from playing, and forcing it to turn to inferior backups. If the average starter spends 10 days on the DL per year, and Pirates starters (hypothetically) spent five days on the DL per year, that would look like a huge advantage! But DL days don’t matter; playing time matters. If the Pirates (hypothetically) got that extra health by resting their players an extra five days per year, any advantage from playing their best players more often would be cancelled out.

This concept is broader than just health, but as distribution of playing time among regulars and reserves is the main way health affects team performance, it’s worth examining. The first step is to distinguish between those two categories of players—regulars and reserves. I used PECOTA’s preseason projections to identify the position players projected for at least 1.5 WARP per 600 PAs, and called them “regulars,” making all other players “reserves.” There are 324 regulars, or slightly more than 10 per team, so this is a reasonable if inclusive definition. I then calculated the number of regulars at the beginning of the season for each team, and the rate of starts by those players over the course of the season.

There are varying numbers of good players on each team, so precise comparisons are difficult. There are eight teams that began the season with between 12 and 14 players projected over the 1.5 WARP threshold, including the Pirates at 13.

Of the eight, the Pirates have ended up the second-lowest rate among their peers, with only 62 percent of their starts going to players with regular-quality preseason projections. If the number of regulars each team started with are ignored, and we compare the Pirates to the entire league, they’re just 17^th, despite starting with more “regulars” on their roster than the bulk of the league’s teams.

Again, this is measuring more than merely health, but health itself isn’t important; performance is. This is one of the important things that health impacts—the ability of a team to put its best players on the field—and it’s evidently not an area the Pirates have obviously excelled in.

This isn’t the only impact of health, however. Perhaps the biggest differentiating feature between baseball and all other American professional sports is what Russell Carleton has called the Grind, the relentless wearing-down that afflicts players over the course of a long season that runs through the heart of the summer. In that article, Russell found that hitters’ plate discipline erodes over the course of the season, presumably from the effects of fatigue. This is the less visible impact of good health on a player, but it’s just as important: Staying on the field doesn’t do a team any good if the player can’t perform well once he’s there.

There are a couple ways to see if the Pirates players are healthier than their counterparts, and estimate if and by how much that’s improving their performance. Another article by Russell found that players who had played five games in the last seven days performed better than their counterparts who had played in seven of the last seven days, by about three points in OBP. Pirates starters have played five or fewer games in the last week 65 percent of the time, versus 59 percent of the time for non-Pirates players, consistent with the narrative that they’re being more careful and deliberate with rest. Using Russell’s figure and some back-of-the-envelope calculation, and assuming those players had 4 PA each game and a baseline OBP of .320, Pirates starters will reach base an additional… one time over the course of a season. Not one time each, but one time as a group. That’s a rough estimate, and presumably Russell didn’t find the only impact of fatigue, but the magnitude is small enough to send a clear message. The difference between the Pirates’ rest patterns and those of the rest of the league would need to be a lot more dramatic to make a meaningful difference in this fashion. And more meaningful rest would mean more games going to non-regulars, perhaps undoing the benefit.

There’s one more way to search for beneficial impacts of extra rest on the Pirates: comparing projected and actual performance. PECOTA projects how a player will perform by looking at what comparable players have done in the past. Any given player’s historical peers presumably rested in the traditional fashion, so their levels of fatigue increased as the year went on, and that was reflected in their performance.

I took every position player’s performance through the end of August, and calculated their WARP/600 PAs and compared it to their preseason PECOTA projection, then calculated team figures as the weighted total of their individual players. I left out traded players, both for simplicity’s sake and because it’s not clear if Aramis Ramirez, for example, would benefit from changing his rest patterns in July.

The Pirates do rank better on this measure than most teams, but not in a very convincing fashion. League-wide, the average difference in actual WARP/600 minus projected WARP/600 has been just under 0.1, so, unsurprisingly, PECOTA does an excellent job in the aggregate. The Pirates’ figure is 0.4, good for 11^th, but behind several teams with much more extreme figures, including the Giants (1.3), the Diamondbacks (1.1), and the Blue Jays (1.0). And, of course, that 0.4 could be explained by any of dozens—hundreds? Thousands?—of other variables.

Unfortunately, if there is a real difference here, there’s too much noise to see it. If the Pirates consistently beat their projections over several decades, that would probably be enough data to draw a conclusion. In a single year, however, the effect would have to be enormous to be clearly visible through all the randomness inherent in performance. And we wouldn’t expect the effect to be enormous in the first place. That might make the previous few paragraphs seem like a waste, but I think this shows something important. Even if the Pirates have figured something out, it’s most likely the sort of thing that might be the difference between a postseason berth and a miss every hundred seasons or so.

The Pirates are definitely doing something different than the rest of the league. They appear to be resting their players more often. They are also one of the league’s best teams, and will go to the playoffs for the third straight year. But if the rest patterns deserve credit, it’s not enough to show up in several different measures of performance. Maybe they’ve figured something out, but the dramatic breakthrough in understanding health hasn’t arrived yet.

Henry Druschel writes for Beyond the Box Score. You can find him on Twitter @henrydruschel.