Advanced stat models really seem to like the Flyers' chances to beat the Capitals, and here's why
The national media are universally predicting a Capitals series victory, but some of the leading minds in the hockey statistical analysis world aren't so convinced. Why?
On the strength of a 17-6-4 record over the season's final 27 games, the Philadelphia Flyers found a way to battle their way into the Eastern Conference's final wild card position. Their reward? A matchup against the team with the best record in the National Hockey League this season, the Washington Capitals.
Unlike the Flyers, who are in the process of a retooling (or a rebuild, depending upon who you ask) and should view just making the playoffs as a major achievement, the Capitals' expectations are far more ambitious. Alexander Ovechkin has yet to even see the Eastern Conference Finals in his NHL career, and now backed by a strong defense and one of the league's better goaltenders, his team's championship window appears wide open.
As a result, it's not surprising to see the vast majority of national media pundits choosing the Capitals to roll over the Flyers like a speed bump. In fact, not one ESPN hockey prognosticator picked Philadelphia to engineer the upset. Only two of them think the Flyers will even push the series to the distance.
Again, this pick seems to make sense at first glance. The Capitals won a league-leading 56 games this season; the Flyers won just as many as they lost (41). Washington outscored its opponents by 59 total goals, while Philadelphia was actually outscored by four. The Capitals have the league's fifth-most efficient scoring power play, and second-best penalty kill. And that's before we even mention that the NHL leader in goals is their captain and star.
But if all that remains true, then why are some of the best minds in hockey statistical analysis giving the Flyers a real chance to steal the series?
Micah Blake McCurdy, owner and operator of HockeyViz.com, has built a model with the goal of predicting team playoff success. McCurdy's model accounts for a team's even strength shot creation and suppression, shooting percentage, power play shot generation, penalty kill shot suppression, and career performance of the starting goaltender. Contrary to ESPN, his model sees the Flyers-Capitals series as a true toss-up.
Post-season chances, with slight updates due to shifting to 100% one-goalie deployments. pic.twitter.com/Dr8DXpI3nz— Micah Blake McCurdy (@IneffectiveMath) April 12, 2016
Another model built upon similar principles developed by Matt Cane of PuckPlusPlus.com and Hockey-Graphs.com actually gives Philadelphia a substantial edge over the heavily-favored Capitals, with a win probability of 59.92%. Per his work, the Flyers are second only to the rival Pittsburgh Penguins in their chances to represent the Eastern Conference in the Stanley Cup Final.
Initial playoff predictions. PIT, PHI favourites in East, L.A, STL tops in West. Pens over Kings for the Cup. pic.twitter.com/HlR4Udahb6— Matt Cane (@Cane_Matt) April 12, 2016
So what is causing the discrepancy between the general consensus and the stat-based predictions? To understand that, it's best to understand the specifics of McCurdy and Cane's models.
Both models are based around the idea that the final 25 games of the regular season are most applicable to future playoff performance. Gabriel Desjardins analyzed this trend back in 2011, and McCurdy also found when developing his 'Oscar' prediction model that predictability was maximized when limiting the analysis to the preceding 25 games for each team.
In training, this was measured as the score-and-venue-adjusted unblocked shots per sixty minutes of 5v5 play during the previous 25 games, or as many games as were available early in the season. Considering larger samples of games weakened the fit and predictivity of the model, as did adjusting the shot numbers for strength-of-schedule.
The theory passes the smell test. After all, a team's roster and lineup over the final 25 games of the regular season is very close to what their roster will look like on day one of the playoffs. Over the course of a season, players get injured, trades happen, and young players fight their way up the lineup, all changing the complexion of teams. By limiting the sample to the final 25 games of the season (since about late February), the models cut through the noise of roster adjustments and zero in on predicting the future performance of the team as it is at the start of the playoffs.
More than anything else, this is why the Flyers grade out so well by McCurdy and Cane's statistical models. As TSN's Travis Yost pointed out yesterday, Philadelphia is right there with the Capitals in every aspect over the past 25 games.
Suddenly, the series doesn't look like a mismatch at all. It becomes a battle between two teams executing at a high level across the board. It doesn't penalize the Flyers for their terrible play in late October and early November, nor does it credit the Capitals for running over the NHL through the season's first half.
Now, these metrics are far from perfect. Last season, the top four teams in the NHL by score-adjusted Fenwick (unblocked shot attempts for) over the final 25 games were the Pittsburgh Penguins, Winnipeg Jets, St. Louis Blues and Nashville Predators. All four squads lost in the first round of the playoffs. Still, it's fair to say that the 25-game theory has passed through vigorous testing and still holds water as the best way over a large sample to predict playoff performance.
The main argument from those who expect the Capitals to breeze to a Round One victory is that the last 25 games are not an accurate barometer of the true talent level of Washington. After all, the Capitals clinched a playoff spot nearly a month before the regular season ended, and were never in realistic danger of losing the Eastern Conference top seed either. As a result, the theory goes, by looking only at the final 25 games of the regular season, the models are putting too much weight on a period of time when the Capitals were not properly motivated to play their best hockey.
Sometimes this theory plays out. For example, the Chicago Blackhawks sagged in the early months of 2015 from a shot attempt differential standpoint, only to roll to a championship once the playoffs started. Then again, the 2010-11 Philadelphia Flyers dropped off in March after dominating the Eastern Conference through the majority of the season, and the result was a tough seven-game first round series against a flawed Buffalo Sabres team and then a four-game sweep at the hands of the Boston Bruins.
The Capitals did close their season with some strong performances, specifically a 5-1 thrashing of the Cup-contending St. Louis Blues in their second-to-last game of the year. If you believe that Washington is going to cruise to the second round this year en route to a long playoff run, you probably think that the Caps are peaking yet again and are now ready to regain the heights of their early-season play.
But if you believe the models, the Flyers will be right there with them, using their stellar performance over the year's final 25 games to hang right with one of the favorites. In a little over 24 hours, we'll find out.