What if the Playoffs Made Sense?: A Postseason Preview

If you’ve paid any attention to baseball over the past decade or so, you know that playoff results are governed by something other than logic. There is no formula that can identify a great playoff team. October baseball is driven by some combination of heart, guts, grit, and randomness- a little heavier on the last one.

But what if the playoffs made sense? What if Major League Baseball Players were of such varied skill levels that better players beat lesser players every time? Little League is a little like this, as is the NBA. Baseball might have been more like this a century ago, when only white players from the northeastern United States played the game, “sports medicine” wasn’t yet a thing, and most players had other jobs in the winter to make ends meet. A team with Babe Ruth and Lou Gehrig and Tony Lazzeri and Earle Combs was just better than a team without those guys, so of course they won. (editor’s note: sometimes they lost)

In an effort to understand which teams are best positioned for the playoffs- in other words, which teams would win if we could somehow set aside all the randomness in baseball- I created a model that assigned run values to each pitcher, hitter, and fielder based on how many opportunities he’s likely to have to impact a game in a best-of-one, -five, or -seven game series. Why not just add up each team’s WAR and call it a day? Because the Blue Jays don’t care how many games their bullpen blew early in the season when the personnel was entirely different. Because the Dodgers don’t care how their fifth starter fared this year. Because the Mets are a different team with Yoenis Cespedes and a healthy David Wright.

I started building the model by establishing a prototypical 25-man roster, consisting of four starting pitchers, a long reliever, a closer, lefty and righty setup men, four additional relievers, one regular starter at each position including DH (or primary PH in the NL), a backup catcher, a utility infielder, a fourth outfielder, and a baserunning specialist. I understand that some teams won’t construct their rosters this way, but I had to standardize the field to measure teams consistently.

The model makes a few core assumptions. First, teams will lean heavily on the pitchers at the top of their rotation and a few key relievers. In a one-game playoff, a team will only use its ace (assuming he’s available) and a few core relievers. In a best-of-five, at least three starters will appear, but the number one guy is likely to start twice. In a best-of-seven, most teams will use four starters, a long reliever is likely to make an appearance or two, and the team will probably dig deeper into its bullpen, pulling out a second lefty specialist or a starter who didn’t make the playoff rotation. I based my innings pitched projections on some facts- starters averaged 5.8 innings per start in the majors this year and teams used 3.09 relievers per game- and some guesses- in a four-game sweep, a team is equally likely to use its ace twice than to use its fourth starter.

Next, a team’s best hitters will come to bat more than its worst, but that impact is less severe than with pitchers. Again, plate appearance assumptions were based on some facts- teams averaged 37.8 plate appearances per game this season- and some guesses- players who typically hit in the bottom third of a team’s lineup are more likely to be pinch-hit for or given a day off in the playoffs. In all, I assigned 69% of all plate appearances to guys who usually bat in the top six spots, 25% to the bottom three regulars, and 6% to the subs.

With the framework of the model built, I assigned players from each of the ten playoff teams to the 25 roster spots on each team. While Paulo Orlando might get more plate appearances as Kansas City’s fourth outfielder than Travis Ishikawa gets as Pittsburgh’s, they slot into similar enough roles that the run values will be directionally correct.

For each player, I looked at his full-season stats (courtesy of fangraphs), even if he changed teams or leagues during the year. For pitchers, I pulled their Runs Above Replacement per inning pitched and multiplied it by the number of innings their roster spot is likely to pitch in each scenario (1, 5, or 7 games) to arrive at an estimate of how many runs they’re likely to save in a series. For hitters, I grabbed offensive runs above average per plate appearance and multiplied by their lineup spot’s projected plate appearances. Then I took defensive runs above replacement per plate appearance and multiplied by projected plate appearances for a separate fielding runs estimate (I used a different PA estimate for fielding to more closely resemble innings fielded; using innings fielded as the denominator is complicated by the negative adjustment given to designated hitters in Defensive Runs Above Average). The finishing touch was a look at baserunning runs above average per game played, multiplied by the assumption that each team’s baserunning specialist will get one opportunity per game to pinch run.

Adding these values for all 25 roster spots, I came up with a score for each team in a best-of-one, -five, or -seven game series. While the baseline is replacement level for pitchers and average for position players, all figures are measured in runs, so we’ll call the result Playoff Runs. Some of the results were surprising.

Rather than simply ranking the teams based on their total scores, let’s mock the playoffs by assuming the team with the better score for the appropriate length series wins every time. Then we can watch the playoffs and wonder why half of these predictions are dead wrong as always.

American League Wild Card Game
Astros (2.37) over Yankees (1.65)

Why the Astros Win This Simulation
By any measure, Dallas Keuchel had a better season than Masahiro Tanaka. By Playoff Runs, the Astros also score better than the Yankees offensively, where the Astros have home run power all over the lineup, and in the field, where the Yankees are by far the worst team in the playoffs.

What the Model Doesn’t Consider
Keuchel pitched far better at home than on the road this season, and will pitch on three days’ rest for the first time in the big leagues. Tanaka, meanwhile, gets an extra day of rest and the two-headed monster of Andrew Miller and Dellin Betances with extra rest behind him.

National League Wild Card Game
Cubs (2.91) over Pirates (1.83)

Why the Cubs Win This Simulation
It’s not just Jake Arrieta. Sure, Arrieta’s 6.5 projected innings are worth 1.76 runs alone, but Gerrit Cole counters with 1.49 of his own. Pittsburgh has a slight edge in the bullpen, but that’s about all they can hang their hats on. The Cubs’ defense, anchored by Addison Russell and Miguel Montero, grades out far better, and their bats pick up another third of a run, thanks to lots of projected plate appearances for Kris Bryant, Anthony Rizzo, and Kyle Schwarber.

What the Model Doesn’t Consider
These baby Cubs (I guess that’s redundant) were called up piecemeal this season and took little time adjusting to the big leagues. As such, we’re dealing with small samples when we say that Schwarber and Jorge Soler are elite hitters and Russell has a great glove. One defensive miscue or rookie mistake on the basepaths could swing this game, so perhaps we shouldn’t put too much stock in a few months’ worth of solid numbers. The Pirates won 98 games and are hosting this game for the third straight year, so they may be better than their Playoff Runs. Then again, observing Jake Arrieta over the last few months gives us reason to believe there may not be many innings left in this Pirates season.

American League Division Series
Blue Jays (10.08) over Rangers (5.64)

Why the Blue Jays Win this Simulation
This may be the only postseason series that feels like it has an obvious favorite, and Playoff Runs back that up, ranking the Blue Jays first among AL playoff teams and the Rangers last. The Blue Jays field better and hit better, and both their rotation and bullpen are superior.

What the Model Doesn’t Consider
There are actually some similarities between these two teams. Both were stuck around .500 at the trade deadline, but decided to make bold moves that paid off. Each team added its current ace (David Price and Cole Hamels) and enough bullpen pieces to turn a weakness into a strength. Hamels and the bullpen were particularly good down the stretch for Texas, and with Martin Perez and Derek Holland healthy, they’ve got the pitching depth to try to hang with the Blue Jays’ dangerous lineup.

Astros (8.32) over Royals (7.87)

Why the Astros Win this Simulation
First off, they barely do. This was the closest of all the matchups. Houston’s pitching grades out better, both in the rotation and, perhaps surprisingly, in the bullpen, where only the Yankees and Blue Jays are better. Relievers Luke Gregerson, Tony Sipp, and Josh Fields were immensely valuable this year, and having Lance McCullers (or Mike Fiers; the model guesses it’s McCullers) as a potential longman out of the pen is a bonus. Kansas City has far better gloves, but the model likes the Astros’ swing-for-the-fences approach slightly more than Kansas City’s put-everything-in-play gameplan, by a count of 2.09 Playoff Runs to 1.7.

What the Model Doesn’t Consider
Kansas City’s bullpen is good too, and Ned Yost can lean heavily on Wade Davis, the game’s best reliever in 2015, if needed. This team seemed to add up to more than the sum of its parts all year, with the great outfield defense picking up the mediocre rotation and the hitters scraping just enough runs across with aggressive bats and legs. Furthermore, switching Keuchel to “SP3” and shifting Scott Kazmir and Collin McHugh up drops Houston’s Playoff Runs to 7.95, a virtual tie with the Royals. This one’s a toss-up.

National League Division Series
Mets (11.13) over Dodgers (9.62)

Why the Mets Win this Simulation
Here’s the biggest surprise of the simulation. The Dodgers have the best pitching, whether we’re looking at one game or a series of any length. The model sees Kershaw and Greinke pitching enough innings, particularly in a short series like this, to bully any opposing staff. But the next best pitching staff is in New York, with Jacob deGrom, Noah Syndergaard, and what’s left of Matt Harvey leading the way. Whether Bartolo Colon or Steven Matz gets the fourth start doesn’t make much difference, as a long man in relief could throw more innings in a short series than the fourth starter anyway, particularly if it’s a sweep. The Mets dive ahead here in both defense and offense. Yoenis Cespedes and Travis d’Arnaud get big points on both sides of the ball. The offensive numbers love Curtis Granderson, David Wright, and Lucas Duda at the top of the order and Juan Lagares and Wilmer Flores at key defensive positions. These Mets are good.

What the Model Doesn’t Consider
These Mets are good, but are they great? Four of the five starting pitchers have thrown more big-league innings this year than in any other season, and the fifth is 42 years old. Are there enough quality innings on this staff to offset Kershaw and Greinke? Furthermore, some of the Mets’ biggest per-at-bat numbers are based on small sample sizes. David Wright returned from injury to post .041 offensive runs per PA (better than Alex Rodriguez an similar to Lorenzo Cain), while Michael Conforto’s .029 defensive runs are based on less than 400 innings in the field. The Dodgers’ second-half slump might make their 2.84 offensive Playoff Runs (third best in the playoffs) look a little aggressive, but would anyone have guessed that the Mets would score higher than anyone, including Toronto? Numbers don’t lie, but they can deceive.

Cubs (10.01) over Cardinals (6.53)

Why the Cubs Win this Simulation
The Cardinals won 100 games based on the consistent excellence of their starting pitching and not much else. In October, they’ll only occasionally have the better starting pitcher, and with Yadier Molina out (the model considers him the backup catcher), there’s not much to be afraid of in terms of offense or defense. In contrast, the Cubs started relatively slowly and never really contended for the division title, but as their rookies matured, the team started to gel, and Jake Arrieta established himself as perhaps the best righty in the game. With the regular season behind us and 25-man rosters set, there’s not much to suggest that the Cards are better. In fact, the Cubs have big edges in pitching, fielding, and hitting over the Cardinals.

What the Model Doesn’t Consider
Cardinals devil magic? The starting pitchers will keep St. Louis in every game, and in recent Octobers, it seems like any game in which the Cards keep it close end up going their way. Past playoff performance doesn’t mean much, though, and these Cubs are just better. Even if we change Arrieta to the third starter, which is where he’d likely slot in after pitching the Wild Card game, the Cubs still have the advantage in every category.

American League Championship Series
Blue Jays (14.06) over Astros (8.66)

Why the Blue Jays Win this Simulation
Batting. But also starting pitching. This is where we finally get into best-of-seven series, where depth matters a little more, but the Blue Jays are still winning based on the heart of the order, where Donaldson, Bautista, and Encarnacion alone hold a 2.2-run advantage over the Astros’ 2 through 4 hitters- and the rotation, where David Price, Marcus Stroman, RA Dickey, and Marco Estrada were far more effective this year than the Astros’ front four.

What the Model Doesn’t Consider
The Blue Jays’ opponent here might be the Royals or Yankees, rather than the Astros. Of course, runs above average prefer the Blue Jays to any of them by a good distance, but no opponent is a pushover in October, and any of these teams could be the one to turn the Blue Jays back into the pumpkin that was struggling to stay relevant at the trade deadline.

National League Championship Series
Mets (15.37) over Cubs (13.91)

Why the Mets Win this Simulation
Regardless of the length of the series, the model likes the Mets more than any NL team. While the Cubs get more Playoff Runs for pitching and fielding, the Mets’ offense is a significant advantage. While Schwarber, Rizzo, and Bryant outscore their lineup counterparts in the middle, the Mets hit better at the top and bottom of the lineup.

What the Model Doesn’t Consider
At this point in the playoffs, we don’t really know which teams will be able to line up their rotations optimally. The Mets have one of the deeper rotations, so they’re more flexible in terms of using someone other than their ace in game one. On the offensive side, though, the Mets were no-hit by Max Scherzer on the second-to-last day of the regular season, then they get dates with Kershaw and Greinke and possible multiple meetings with Arrieta and Lester. Their offense grades out well in Playoff Runs, but they feasted on Phillies, Braves, and Marlins pitching in compiling those numbers. This slate could prove to be their undoing.

World Series
Mets (15.37) over Blue Jays (14.06)
Why the Mets Win this Simulation
These two teams are almost identical in pitching Playoff Runs, and Mets field better than the Blue Jays, but the difference here comes on the offensive side, where New York’s bats are actually better too. Toronto has the edge over anyone in the heart of the order, but extend that “heart” to Granderson/Wright/Cespedes/Murphy/Duda/d’Arnaud/Conforto, with Cuddyer DHing in Toronto, and the Mets look really good. If the three young guns are still throwing gas in late October, this could be the team that finally neutralizes the murderers’ row in Canada.

What the Model Doesn’t Consider
Fangraphs’ WAR is adjusted for league quality, but the run components are not. The National League was top-heavy this year, with seven good teams and eight really bad ones, while the AL was solid from top to bottom. This may be part of the reason why National League teams rank first, third, fourth, seventh, and eighth in total Playoff Runs. The Blue Jays played better teams all year and didn’t have the same opportunities to put up garish run/PA numbers, but they’re clearly stacked offensively and defensively. It’s hard to imagine the Mets being considered favorites if this World Series actually came to fruition.

So there you have. Mets over Blue Jays. I’m not sure that “makes sense”, but it’s at least based in quantitative logic. If (recent) past performance were a strong indicator of playoff success, the Mets would be the most formidable team this October, though their first playoff opponent wouldn’t be far behind. You’ll note that this model favors the newcomers to the playoff party over the establishment. While I believe playoff experience is overrated, we’ve seen the Cardinals and Yankees win enough in October to believe that the Mets and Blue Jays have uphill climbs despite their loaded rosters.

Let’s close with the ten playoff teams ranked by their aggregate Playoff Runs over a best-of-seven series:

1. Mets (4th in pitching, 3rd in fielding, 1st in hitting)
2. Blue Jays (3rd in pitching, 5th in fielding, 2nd in hitting)
3. Cubs (2nd in pitching, 2nd in fielding, 4th in hitting)
4. Dodgers (1st in pitching, 8th in fielding, 3rd in hitting)
5. Astros (5th in pitching, 7th in fielding, 5th in hitting)
6. Royals (9th in pitching, 1st in fielding, 6th in hitting)
7. Cardinals (8th in pitching, 4th in fielding, 9th in hitting)
8. Pirates (6th in pitching, 9th in fielding, 8th in hitting)
9. Yankees (7th in pitching, 10th in fielding, 7th in hitting)
10. Rangers (10th in pitching, 6th in fielding, 10th in hitting)

For player-by-player stats, check out the companion piece here.