What's going on with the Yankees second-order winning percentage?

Joe Girardi is currently in his tenth season as the manager of the New York Yankees. Without getting into how fast time seems to go by, it should be noted that Girardi has proven himself to be one of the best managers in baseball by this point.

For one, only Bruce Bochy (in his 11th season with the Giants) and Mike Scioscia (incredibly in his 18th season with the Angels) have been in their current managerial positions for longer. The Yankees have also finished above .500 in each of Girardi’s nine full seasons, and with just four more wins in their final 19 games, they will accomplish that feat once again in 2017. The franchise has Girardi to thank, in part, for a 2009 World Series ring, but what Girardi has done since 2009 has almost been more impressive. The 2017 season has shown a few signs of a young, future core, but from 2013-2016, Girardi took the bloated corpse of that 2009 roster and continued to at least threaten playoff baseball in New York. As noted previously, the Yankees were over .500 in each of those four seasons, however, whether or not they deserved to be a winning team is debatable. In three of those four seasons, the club finished with a Pythagorean win-loss percentage under .500.

The Yankees should scare you

Photo by Jim McIsaac/Getty Images

2017 was supposed to be a rebuilding year in the Bronx. But their surprising success this season means they’re a threat both now and in the future.

For those unfamiliar with Pythagorean win-loss percentage, the concept is straightforward. Over a 162-game season, a team’s winning percentage should basically align with their total amount of runs scored and allowed. Teams are bound to win a few blowouts, but they’re also bound to be on the flip side of a few of those. By boiling a team’s winning percentage down to their run differential, analytical fans can get a better vibe for the true(r) talent of a team. Pythagorean win-loss percentage has a better season-to-season correlation than actual winning percentage and quoting a team’s run differential is quite in vogue these days.

Therefore, a team beating their Pythagorean win-loss percentage for three out of four seasons isn’t always that common. Especially given that that fourth season saw the two percentages basically even. In fact, here are the Yankees actual win totals versus Pythagorean win totals under Joe Girardi, prior to this season:

The Yankees have outperformed their pythagorean win total in six of Girardi’s nine full seasons, collecting a total of 21 “extra” wins since Girardi’s hiring in October of 2007. In the past four seasons, they have won an average of 4.25 more games/season than their run differential would have suggested they should have.

All of which brings us to 2017.

The Yankees are in the midst of their strongest season since 2012. Their current winning percentage (.545) would be their highest since reaching their 2012 ALCS campaign, and they are currently leading the AL wild card race. They would be favored in that wild card game, and if they won, it would be their first ALDS since, again, 2012.

However, it could be even better.

In addition to Pythagorean win-loss percentage (1st Order Win Pct), Baseball Prospectus gives baseball die hards an even deeper (and theoretically truer) value of what teams have done in a given season. Second order winning percentage also tosses “projected” runs into the mix, trying to remove the luck that some teams have with runners on base to bore down even deeper to the true talent of a team. Third order winning percentage leaves all previous factors in the mix and adds in strength of opponents. For the nerdy sect of baseball fans, these are the purest of standings.

By second and third order winning percentage, only the Cleveland Professional Baseball Team and the Los Angeles Dodgers have been better than the Yankees in 2017. By first order winning percentage (Pythagorean win-loss percentage), the Yanks are fourth, with only the Washington Nationals moving ahead of them. They hold an easy lead on the real-life division-leading Boston Red Sox by any of the first, second, or third order winning percentages, with the gap getting as wide as 14 games by second order winning percentage. The Yankees’ -14.4 and -13.8-win differences between their actual win total versus their second and third order winning percentages are easily the largest gap (negative or positive) in baseball this season, with Cleveland’s -10.8-win gap in their second order winning percentage the only other double-digit gap. In the eight seasons for which Baseball Prospectus has tracked second and third order winning percentage, the Yankees, if they hold this pace, will be the “unluckiest” team ever. No team has ever underperformed their second and third order winning percentage by as large a margin as the 2017 Yankees.

So how is it that a team that has overperformed their run differential - and what they “should” have done - the past decade is now historically underperforming by those same measures?

Bullpen woes

Although the first half of this article heavily implied that Girardi might be the cause of the Yankees overperforming their underlying win-loss metrics over the past four years, the Yankees have also had a stud bullpen over those four seasons. From 2013-2016, the Yankees tied with the Royals for the highest fWAR among bullpens in baseball. Only the Rangers and Padres had fewer blown saves over those four seasons.

In 2017, the story has been a bit different. Somewhat surprisingly, the Yankees pen still leads all of baseball in fWAR (thanks in part to having such a strikeout-heavy bullpen), but they have, at the same time, blown 23 games, a total that ranks third in baseball behind only the Seattle Mariners and Toronto Blue Jays.

Aroldis Chapman has gone from Bullpen Superman to Rather Large Question Mark, as his 2017 ERA (3.83) is higher than the combined totals of his 2015 and 2016 ERAs . He lost the closer’s role, and it was only recently announced that he will be back in that role moving forward. For most of the past four seasons, there haven’t been any question marks at the back end of the Yankees bullpen, and as Royal fans know, bullpens appear to be one area that can really throw off baseball projection systems and run differential totals.

Different roster construction

The Yankees, as a unit, look a lot different in 2017 than they have in recent seasons. In many ways, that is for the better. As noted above, the 2017 team is finally showing signs of a future core, with Aaron Judge and Gary Sanchez putting together their first full seasons, while Luis Severino is finishing his first season entirely in the rotation. Judge is 25, Sanchez is 24, and Severino is 23.

That’s a vastly different age group than the one that led the Yankees from 2013-2016, when Mark Teixeira, Carlos Beltran, Alex Rodriguez and several other tricenarians were among the leading contributors to the Yankees win total. There are still a stable of veterans in New York (Brett Gardner, C.C. Sabathia, and Jacoby Ellsbury most notably), but the average age of the leaders on this team is noticeably lower than in seasons past.

Although that fact is certainly a positive for the future of the pinstripes, it may help to explain the difference in the Yankees actual winning percentage versus their second and third order winning percentages. Older players can have their flaws, but the experience that comes with 15+ years in the league almost certainly can permeate a clubhouse and have an effect that hasn’t been accounted for by metrics like second and third order winning percentage just yet.

Simple regression

While some folks may look at the Yankees recent success compared to their run differential as a repeatable pattern, many others would look at it and say they’re due for some heavy regression. In all likelihood, both would be wrong, as no one has been able to find a team that can reliably top their second and third order winning percentage over a truly significant stretch of time, and the latter half of the above sentence would be falling into the Gambler’s Fallacy. The Gambler’s Fallacy states that just because a coin has come up heads five times in a row, there’s not a statistically better chance that the coin will be tails on the next flip.

However, play enough seasons, and a team is bound to have a few seasons with a worse actual win-loss percentage than their Pythagorean win-loss percentage (the field of statistics is weird that way). In a vacuum, the 2017 season for the Yankees wasn’t more likely to be one in which they were unlucky by Pythag win-loss, but it (probably) also wasn’t more likely to be one in which they were once again lucky. (This is a bit of an over-simplification, but for this section, it will do for now.)

One-run games

If all the previous factors were the clues leading us to the actual perpetrator, one-run games are the actual murder weapon. Check out this rundown of the Yankees sharp turn in one-run games over the past five seasons:

There you go. If the bullpen gets a bit shakier, the players get a bit less experienced, and the team is dancing with the regression devil, this is what can happen. The difference is stark, and it certainly explains why the team has the fourth-best run differential in baseball but the eighth-best winning percentage.

The Yankees have spent the better half of the past decade trying to prove that the outcome in one-run games is sustainable for a certain team if the correct factors are in place (namely a strong manager and bullpen). Now they’ll have to hope that the outcome in one-run games isn’t actually telling over a 162-game stretch, or they may be in a lot of trouble when the tight matchups of October baseball swing around.

Jim Turvey is a baseball diehard who also writes for DRays Bay. You can follow him on Twitter @BaseballTurv.