Looking Back at the 2017 Postseason

The 2017 season is over, capped by a postseason that had a little something for everybody. It had home runs galore, including two players, José Altuve and Enrique Hernández, hitting three in a single game. It had every kind of pitching performance, from the New York Yankees bullpen throwing 8.2 innings to win an elimination game to Justin Verlander pitching conceivably the last postseason complete game we will ever see. We even had a World Series game come in under two and a half hours. That also might be the last we ever see.

In the end, it had another long championship drought breaking. At 56 years, the Houston Astros’ wait didn’t measure up to the Cubs’ 108 that ended last November. It had, though, been the third-longest extant drought, behind the Indians’ 69 seasons and the Washington Senators/Texas Rangers’ 57 years. The Astros paid a hefty price in long rebuilding years with terrible records and local TV ratings of 0.0, but they got the payoff, and at a time when the city of Houston needed it most.

I’ll be looking back at the postseason today, just as the title promises. As has become my annual custom, I will be gauging the excitement of the games rather than the ups and downs of player and team narratives. For this, I bring back out of cold storage the WPS Index.

Back in 2012 and ’13, I did daily reports on postseason games supported by a metric I call the Win Percentage Sum, or WPS. When THT changed its format in early 2014, I went to a post-postseason wrap-up. This is the fourth of those.

The next several paragraphs give a summary of WPS methodology, though you can instead refer to my original articles from August 2012 for a fuller explanation. As always, I will first credit THT’s Editor Emeritus Dave Studeman for creation of the original game excitement metric. When I conceived my system, I did not know of his similar work, published in the 2007 Hardball Times Annual. It is thanks to his forbearance that I continue to use WPS, while giving credit where it is due.

WPS is calculated with a base score that adds together the change in game Win Expectancy created by every play, meaning every continuous action that changes the base-out state of the game. For example, if Mike Scioscia leads off the bottom of the ninth with a popout that lowers his team’s chance of winning by eight percentage points, that’s a score of eight. A few batters later, when Kirk Gibson hits a two-run home run that takes his team from a 13 percent chance of winning to an actual victory, that’s 87 points added to the WPS score.

To the base score I make two additions, meant to reflect fans’ subjective enjoyment of the high points of a game. I add the three plays with the highest Win Expectancy changes to the base, then add the value of the game’s final play. These reflect, respectively, the highlights of the game and how much the game was in doubt at the very end.

If a walk-off play is one of the top three, that means it counts three times: in the base, in the highlights, and as the final play. This is intentional. We fans love walk-off wins, especially those going from defeat to victory in one play, so the WPS system favors them.

The median WPS score of a game is roughly 300, with wide variations running from below 100 to over 1,000. Under 200 denotes a snoozer, while 500 points is my criterion for a “great” game. The mean of WPS scores is somewhat higher than the median, because there is more numerical room for a game to be great than to be boring.

The system has blind spots. Outstanding individual performances or milestones make no impression except by how they affect the game’s competitiveness. No-hitters and perfect games are hardest-hit: As they usually lack the see-saw movement between offense and defense, or between the competing teams, their scores are generally flat. Best-play and final-play bonuses can capture some subjective reactions, but others are beyond a system with any pretension to simplicity.

Another, more ambiguous, feature of the system is that it believes more baseball is always better. WPS points can only accumulate, not be subtracted. Thus, extra-inning games have a structural advantage, growing greater the longer the game goes. This blends into the question of how a game’s time duration affects fans’ enjoyment, a matter which once again turned out to be highly germane this postseason.

This year, I have made an alteration to my ratings system. It acknowledges the rule change that allows a player to be intentionally walked by a mere signal without the throwing of any pitches. While this action does change the base-out state, there is no actual gameplay that produces the change. I have therefore excluded intentional walks from the ratings. Any such free pass gets a score of zero, no matter how it alters win probabilities.

For the record, there were 24 intentional walks in the 2017 postseason. They added up to 0.137 Win Percentage Added, and thus 13.7 WPS points. As the whole postseason compiled a total (before intentional walks were excluded) of 11,761.1 WPS points, leaving out the free passes removed 0.116 percent of the score. My spurt of tidiness did almost nothing to change the scores. Whether automating the intentional walk did more than that to alter the length of games, I will not proselytize about right now.

There is actually a second change I have made, but I will postpone rolling it out for the proper moment.

A quick note before I start delving into WPS. I used FanGraphs’ WPA numbers to work out WPS game scores this postseason, but for games before 2012 (which arise in historical comparisons), I used Baseball-Reference data. This will produce small differences in scores, bigger than the intentional walk tweak, but not enough to turn a good game bad or vice versa.

Postseason, By the Numbers

The Wild Card round gave us two wild games–the four starting pitchers combining for just 7.1 innings pitched–that produced surprisingly average scores. The Twins and Yankees provided great back-and-forth action in the early going, but New York locked in three- and four-run leads from the fourth inning onward, and it is high leverage in late innings that produces the best scores. WPS came in with a verdict of 295.1 points, a hair below average.

Colorado and Arizona did better, but not by that much. The D-Backs ran out to a 6-0 lead after three innings that could have killed suspense, but the Rockies got our attention back with a four-run fourth inning to climb back within range. Every half-inning from the seventh on saw runs score, and while Colorado never managed to pull even, their threats breathed fresh life into the game. The contest scored 319.3 WPS points, just above average and pretty good for a game in which one team led from the first inning to the end.

The divisional round had a much broader range of results.

On balance, it was a good year for the LDS. Out of 96 League Division Series ever played (from 1995 on, plus the 1981 strike-year round), Chicago-Washington had the 16th-best WPS score, New York-Cleveland the 21st-best, Boston-Houston the 52nd, and Arizona-Los Angeles the 91st.

The Dodgers and D-Backs did well in putting people to sleep. Game One between them, at 169 points, was the weakest of the divisional round, and none of their games reached the average of 300. On top of the figurative soporific effect was a literal one. All three games lasted over three and a half hours, the only one of the seven playoff series to achieve that distinction. Add the West Coast starting times that began games past 10 p.m. in the East and ended them close to 2 a.m., and not many people stuck around to see the final pitch. It was a mercy to them that this was the most missable series in 2017.

Boston-Houston rode out an early pair of 8-2 Astros wins to turn more exciting. Game Three, an even more lopsided 10-3 win for the Red Sox, was still a one-run game at the seventh-inning stretch, letting it compile a moderately good WPS score. Game Four did much better. It never had more than a one-run lead until the Astros’ ninth, and Boston pulled it back within one on Rafael Devers’ leadoff inside-the-park home run in their half of the ninth. The WPS score just edged into “great game” territory, though taking four hours and seven minutes to get there.

The biggest scores, as we’d expect, came from matchups that went the limit. The Yankees and Indians finished slightly lower, despite a wild 13-inning Game Two that posted the second-highest WPS score of the 2017 postseason. (WPS is not capable of measuring the drama of Joe Girardi’s non-challenge of the Lonnie Chisenhall hit-by-pitch, but it got Francisco Lindor’s subsequent grand slam just fine.) A pair of games below 200 WPS points, along with two others hovering near average, kept this series from soaring.

That put the Cubs and Nationals on top for the round. Their Game Two was heading for a somewhat mild finish until the five-run Nats rally in the home eighth. Bryce Harper’s bomb turned a 3-1 deficit into a tie; Ryan Zimmerman’s made it a 6-3 lead. That half-inning raised Washington’s Win Expectancy by 79 percent and tallied 95.0 WPS points, making it the most exciting half-inning of the whole postseason.

The rest of the series passed without extraordinary highs or lows–until Game Five. That game became statistical legend for the top of the fifth when the Cubs, in the midst of a four-run rally with two outs, had a string of: intentional walk, dropped third strike, catcher’s interference, and hit-by-pitch. Those four unusual results had never before occurred consecutively in any recorded major-league game.

(For those who lament the automatic intentional walk, imagine the glorious chaos had Max Scherzer been required to throw four wide to Jason Heyward and uncorked a wild pitch in the process. The slim chance of such a mistake is one of the leading arguments against the automatic pass–and it is precisely the kind of thing that would have happened in this bizarre sequence.)

For all that, the top of the fifth wasn’t the best WPS half-inning of the game! It was the bottom of the eighth, when the Nationals, down 9-7, put the tying runs aboard, lost two baserunners to a double play, pushed a tally across to make it 9-8, got the tying and go-ahead runners on base, then had said go-ahead run ruled out on a pickoff after a video review. Whiplash like that is good for WPS scores, and the Nats’ eighth earned 71.2 points versus 63.6 for the Cubs’ fifth.

The game in total came to 489.4, just under the “great game” threshold. That may be apt, as the game was more memorable for its stranger aspects than for truly great gameplay. Nationals fans will not mind the downgrade, as for them the game ended up yet another playoff disappointment, and the last hurrah of Dusty Baker’s managerial career.

The next round represents a step downward, in a way that fans may well wish to dispute.

Of 64 best-of-seven League Championship Series played, this year’s NLCS comes in at 59th by WPS. This extends a streak of terrible LCS to three years. In 2015, the Mets and Cubs played the lowest-scoring LCS in the seven-game era. Last year, the Indians and Blue Jays played the third-worst, which was also the worst ever that was not a sweep. This year’s Dodgers-Cubs matchup was the second-worst non-sweep in the best-of-seven era.

This came despite a quite good Game Two. That game was tied for 48 of its 53 outs, and the only lead greater than one run arrived on the final pitch of the game. Justin Turner’s walk-off three-run home run raised his Dodgers’ Win Expectancy 39.4 percentage points (from 60.6 percent to victory) and provided the largest final-play score of the postseason.

That was the highlight of the series: Cynics might say the only one. The series was filled out with three pedestrian games and a hopeless blowout in Game Five, which at a WPS score of 134.1 was the most boring game of the playoffs. Enrique Hernández’s three-homer performance in that game gave us non-algorithms something extra to enjoy, but I’ve explained that blind spot in the system already.

The Astros and Yankees did better, but surprisingly not by very much. The ALCS came in at only 37th out of 64, in the bottom half of best-of-seven Championship Series. Of the 16 LCS that have gone the full seven games, it registers as the second-least exciting.

The problem was that, while the game-by-game balance of the series carried excitement, the games themselves had far less. Five of the seven contests scored below 250 WPS points, two of them below the 200 line that distinguishes the true snoozers. The other two games, driven by late rallies, were merely above average, neither even approaching 400 points.

Part of the trouble was that the pitchers’ duels of Games One and Two were just too tight. WPS likes back-and-forth action, and offenses limited to 30 combined baserunners in two full games do not provide that. Game Two did end with a good score, thanks partly to Jose Altuvé’s mad dash home to win it, but the style of the games was not calculated to get WPS’s needle moving. Game Four was just as bad until late bursts of runs, Houston’s and then New York’s, jolted the game to life.

Add to those games four rather lopsided contests, none closer than four runs at the end, and the raw WPS scores come out quite bad for a series that went to a winner-take-all game. If you believe a series going the limit should receive added consideration for that, I will soon take your objections under advisement.

First, however, we will look at the 2017 World Series, lauded by some in the aftermath as the greatest of all time.

WPS’s answer to that enthusiastic assertion is a calm “Not quite.” The system ranks 2017 as the seventh-best World Series out of the 113 played. It falls over 400 points short of the 1912 Series, whose berth as best ever bears the asterisk of a tied game called for darkness and an eighth game deciding the best-of-seven. Exclude that match, and 2017 still comes almost 400 behind the fabled 1975 Series, and trails the Fall Classics of, respectively, 1924, 1991, 2011, and 1972 as well.

Granted, after five games (and definitely right after Game Five), the idea that this was going to be the best World Series ever was quite plausible. One more extra-inning classic like Game Two or Game Five, especially if it came as Game Seven, could well have pushed it over the top. Instead, we got a decent Game Six and a Game Seven that was all but over after just eight outs. Few could dispute that it was an anticlimax.

The letdown, though, came from a great height. Two games broke the 600-point mark, a feat that only two previous World Series (1924 and 2001) had accomplished. Game Five, at 764.7 points, had the fifth-highest score ever for a World Series game, just edging out the epochal Game Six in 1975–and in two fewer innings to boot. WPS is not so precise that you couldn’t make a case for the Houston/LA Game Five being as high as third all time or as low at 10th, but out of 660 World Series games ever played, that is still lofty territory.

Those two classics unsurprisingly dominated in WPS single-inning scores. The 10th inning of Game Two was the most exciting inning of the postseason, coming in at an even 140 points. (That still left it well short of last year’s best, the 175.2 eighth inning of Game One of the Cubs-Dodgers NLCS.) L.A. scored twice in the top of the frame, but George Springer’s homer in the bottom matched them and laid the cornerstone of his World Series MVP campaign. The circumstances echoed Game Six in 2011: The Rangers scoring twice to open extra innings, and the Cardinals scrambling to regain the tie. When talking about exciting games, that game is a perfect model to emulate.

Game Five could not match that one-inning score, but it did produce the rarity of two 100-plus innings. The seventh, in which Cody Bellinger tripled the Dodgers ahead but the Astros clouted Brandon Morrow for four runs, rated 119.5 points, second-best for the postseason. Then came the ninth, where L.A. made up a three-run deficit and then stranded Yulieski Gurriel’s winning run at second to force extras, which tallied 100.3 points. On top of that, the game had three separate plays producing Win Expectancy swings of over 30 percent, an excellent result. (Game Six in ’11 had four, but falling a little shy of that unbelievable contest is no shame.)

Outside those two games, though, the series faltered. Two games were roughly average, two others moderately below average, and the final game was a sub-200 fizzle, unworthy of what had come before. That is how the series fell shy of the absolute first rank, even in a system like WPS that does not take the order of games, the value of a classic contest coming at the most pivotal moment, into account.

Postseason, By Some Other Numbers

That leaves the question: What if the system did take that into account?

There’s a way to do this, one that I anticipated when I first designed the system. To quote myself at the time, “Something akin to Sky Andrecheck’s Championship Leverage Index could be added to the WPS to produce this modified measurement.” I have not done this before, but the way the World Series ended has persuaded me to try it this year.

Championship Leverage Index measures each postseason game by how much it moves its teams toward, or away from, a World Series victory. In this system, Game Seven of the World Series is worth 1.0 Championships Added (ChampAdded), for the obvious reason. Game Six of the Series is worth 0.5 ChampAdded, as is Game Seven in a League Championship Series. A full accounting of ChampAdded values for all postseason games is in the chart below.

The “X-Y” figures show progress in the playoff series. A tighter series makes for higher leverage in a game, just as a closer game creates higher leverage for a plate appearance.

I used this lens to take a different look at postseason history. I took the WPS scores for every postseason game and multiplied them by the Championship Leverage Indexes of those games. I’ll call the resultant ratings the Championship Percentage Sum (CPS).

This gives us a new way of comparing World Series, or LCS or LDS, to each other, either by game or by series. Comparing games or series across playoff levels would not really work; the inherent bias in favor of later series would be just too great. One needs also to be aware of another bias in CPS: a very strong affinity for series that go longer. Adding a Game Seven to a World Series increases the potential points not by a sixth, but by at least 42 percent and sometimes more.

Those cautions given, let’s look anew at historical rankings for the playoff series of 2017. I will dispense with actual raw figures, as they are just a flow of numbers without context. I will instead give that context by ranking each series among its kind.

Three of the four League Division Series lose ground, but the Cubs and Nationals end up looking like a classic, leapfrogging 11 spots. Two factors cause this move. The high WPS score of Game Five, combined with its high Championship Leverage Index, boosted the CPS. Also, just reaching Game Five provided the series a big leg up. In WPS, the top three LDS are five-game series. In CPS, it’s the top 13.

This is also how the Astros-Yankees ALCS bounded from a mediocre 37th to a good 21st. If you measure that series just against seven-game LCS, by WPS it comes in as the second-worst ever, 15th out of 16. By CPS, it’s 14th out of 16. A seventh game, even one that lost its juice after five innings, was enough to lift the series in the eyes of CPS. The Dodgers-Cubs NLCS effectively treaded water: going from WPS to CPS, it remained the second-worst (best-of-seven) LCS that wasn’t a sweep.

The World Series had that critical seventh game, so one might expect it to gain in the CPS rankings. The trouble is, all classic World Series have gone the limit. By WPS, the top 10 Series all ended in a winner-take-all game. By CPS, it’s the top 28. (There have now been 39 such World Series.) Reductionist as this appears, it matches well with the fan experience I am trying to quantify. Think about it: The World Series that you consider the greatest ever either went the distance or involved your favorite team winning it all. I am confident this is true.

A World Series needs something more than seven games to awe the CPS system, something like an amazing Game Seven. The 2017 Series had the opposite, and so it slips from seventh in the WPS rankings to 11th by CPS.

I would be interested to learn which system THT readers think better matches their own judgments about what makes a playoff series great or lousy. Depending on the answers, you may see more of CPS in the future.

The Game Clock

Once again, I have things to say about the length of postseason games, because once again game duration became a postseason subplot, and not in a good way.

A pair of Wild Card games that approached four hours in length without extra innings, or even a bottom of the ninth, opened the proceedings. The divisional rounds produced four games of longer than four hours, only one an extra-inning affair, and was capped by the 4:37 of the Cubs-Nationals clincher. Game duration shortened in the LCS round, with no four-hour endurance tests, but the postseason remained on pace to be the first ever to have no games that finished in fewer than three hours.

Game One of the World Series upended that narrative, hard. The 2:28 game was a huge outlier, a throwback fueled by a mere 211 combined pitches thrown. It was not repeated. The 11-inning Game Two cracked four hours, and Game Five was the longest of the postseason, five hours and 17 minutes despite going just 10 frames.

The trend in game length, both before and during October, has been strongly upward for years. There was a reversal in 2015, when pace-of-play rules such as batters having to stay in the box during a plate appearance were enforced for a while. That time is long past.

2017 did not have the longest postseason game average, but 2014’s “win” came on a technicality. That year had six extra-inning postseason games, totaling 18 extra frames, half of them coming in one epic Giants/Nationals game. 2017 had three instances of free baseball, adding just seven innings to regulation. This is why 2014 held the lead for total time. Going by time per inning, 2017 was well ahead, by almost a full minute per inning.

This reflects the regular-season numbers. After a two-year reprieve, game pace, measured by game and by inning, once again set new all-time records in 2017. The pacing gains made in 2015 are now all gone.

Commentators raised the chorus, again, of East Coast children not being able to watch more than a few innings of World Series games before having to go to bed. The problem is not limited by age. Several times during the postseason, I bailed out on games that were going very long, catching the rest of them via my MLB.tv subscription the following morning, if at all. That’s how I watched Game Five of the World Series: five innings at night, five innings the next day. While watching two great halves of a game is still fun, it loses something in the sundering.

It’s pretty clear Commissioner Manfred will take more steps this offseason, adding to the change in the intentional walk. The advent of the pitch clock, which appeared likely after last year’s lengthening postseason, seems all but certain this year. I commented last year on the disruption this will bring to the game, and my opinion remains the same.

Instead of the light touch that improved matters in 2015 before being reversed, a heavy hand is going to be brought down. This will not raise quite the opposition it should, because things grew so bad this year that something clearly must be done. A crude remedy will be offered instead of the deft one from two years ago, and we will accept it because the alternative is no remedy at all. I am tempted to say Rob Manfred has a bright future in politics. He is probably practical enough, however, not to want to take the pay cut.

With that act of minefield tip-toeing complete, I will sign off. WPS will very likely return next year, and depending on your reactions, it may have company in CPS.

References and Resources

• FanGraphs for 2017 WPA data
• Baseball-Reference for historical WPA and game time data
• “The One About the 2013 Postseason”, Brad Johnson, The Hardball Times Baseball Annual 2014, for the postseason Championship Value table