The Fundamental Flaw in Regular Season NHL Analytics When Applied to the Postseason.
The Colorado Avalanche were laptop darlings who put up the best five-on-five regular season analytics since that advanced data started being tracked closely in 2007-08. Despite being significant favorites to win the West Division Finals, they were upset by the Vegas Golden Knights in six games. Vegas was then upset in the Stanley Cup Semifinals against a Montreal Canadiens team that many mathematical projections considered the worst team in the playoffs. The Golden Knights were -5000 money line favorites to advance to the Stanley Cup Finals. The New York Islanders overperformed relative to the projections from the analytics’ models for the third straight season under head coach Barry Trotz by taking the Tampa Bay Lightning to seven games in their Stanley Cup Semifinals series. What gives? Are these outlier results? Or is there a fundamental flaw in the assumptions of the mathematical projections used to make predictions in the Stanley Cup playoffs? I think that is the case.Frankly, most of the methods I use to handicap the NHL in the regular season I throw out when handicapping in the playoffs. The goals and dynamic of the regular season seem too different than the competitive experience of the Stanley Cup playoffs. This is not just my observation from handicapping the NHL for 25 years — this is what NHL players and coaches say. The regular season is a grind with teams playing three or four times a week. Often the zeal from the players is on scoring goals and padding statistics since stats help with new contracts. Seeding for the playoffs is not nearly the priority that it is in other sports since home-ice advantage is less of a factor. The coaching edge in making the final shift change at home is important, but the roar of the crowd has less impact on the game, generally, because the game is so fast. Just getting into the playoffs healthy and rested is more important than seizing a higher seed. The Los Angeles Kings won the Stanley Cup in 2012 despite being seeded eighth in the Western Conference. None of the top seeds in the four divisions this year advanced to the Semifinals. Once the playoffs start, the zeal shifts from scoring goals to stopping goals. Players are much more willing to sacrifice their bodies to block shots. Star players take longer shifts. Benches are shortened. And the nature of a seven-game series completely changes the game-to-game dynamic. Players and coaches get deeper into their planning and preparation in stopping their opponents. Speed advantages begin to get neutralized. In the battle between offensive technique and defensive tactics, the defense tends to get the upper hand. The higher stakes of the playoffs create a new sense of urgency not felt in the regular season. Game management mistakes are more often game-changing plays. Finally, the referees are more likely to swallow their whistles and let the action continue. I have seen many NHL analytics experts complain loudly about the lack of penalties in this postseason. They might as well complain about the weather. It has been that way for decades. Every team that has made the playoffs in the last 50 years has sob stories. Keep crying (and losing), or adapt. There is a long list of NHL teams that were dynamic offensive teams in the regular season who then folded in the postseason. Tampa Bay had this problem before adjusting their style of play and making some subtle changes to their roster after getting swept in the first round of the 2019 playoffs to a defensive-minded Columbus team. That Lightning team also needed to learn to be better game-managers. Colorado made some critical mistakes in their series with Vegas. Rather than making the high-risk pass in the third period that often netted an additional goal, the Avalanche need to learn to not risk the costly turnover that might give their opponents a breakaway advantage. Scoring more is not as valuable as not risking getting exposed. When Tampa Bay finally learned that lesson, they won the Stanley Cup last year — and they appear to be on the verge of repeating as champions.The analytics folks are in a pickle as to how to adapt. Their modus operandi depends on the predictability of their data. It is difficult for them to concede the limitations of their regular season numbers, even if that happens to be the case. By the time the sample sizes become actionable in the postseason, the playoffs are almost over. Handicappers (like me) that also incorporate qualitative analysis can still find success (never more effective than in the 2017 playoffs where we ended on a 20-3 sides run with Pittsburgh’s repeat as Stanley Cup champions). Teams with veteran players with playoff experience that play defensive-first physical hockey with counter-attacking offensive tactics and elite goal-tending fit a profile that tends to perform better in the playoffs. That description seems to apply to the Montreal Canadiens and New York Islanders this postseason. Perhaps the NHL analytics community would find fertile ground in identifying team profiles that have success in the playoffs — and then make appropriate comparisons to new playoff teams. This was the interesting approach that Jordan Brenner and Peter Keating endeavored in their seminal 2015 article on Giant Killers in ESPN The Magazine that profiled the different templates of teams that tend to pull upsets in college basketball’s March Madness.Best of luck — Frank.
Read more
