Review
The inset graphic for standing changes illustrates how little changed over the regular season's final weekend. Only four programs either moved up or down relative to all other programs in the conference. The standing of so many programs remained constant that for the first week, this writer decided to remove those that remained stagnant and provide representation of movement for only those programs that did move.
Four programs moved. Eight programs held constant. All shuffling and apparent shifts had everything to do with tie-breaking procedures and nearly nothing to do with last-ditch efforts in final pushes to the playoffs this season. That is except for Clarkson. The Golden Knights made good on a big weekend. They registered the largest single-weekend movement for any program in the ultimate week of the regular season. Clarkson climbed three spots over two games. More on that later.
Harvard, Quinnipiac, St. Lawrence, and Yale rest this coming weekend. Two of those programs with coveted home ice heated over the course of February while the other half chilled. It is worth noting that dominance-penalized Quinnipiac by the end of the month managed to return its rate of earning conference points to the level that it opened February. The Bobcats are not as hot as they were in November as Rand Pecknold even concedes. They are far from dead yet. They slumber this week.
Now, on to the teams that hope to become that one-in-four team that on average makes championship weekend, the teams that play this weekend in the first episodes of the best hockey of the season. Three of the four hosts of first-round series chilled over the course of February. The single exception? Clarkson.
Now, before this writer drags you through the process of predicting from this model which hosts should be put on upset alert and which are safe, let's take a look at how well this model would have predicted last season.
Expectations
Which team was hottest in February relative to its performances from November through January last season? Another way, which teams would have been predicted to win each playoff contest if this heating-and-chilling model were predictive?
The answer would not have been Harvard.
The team heating most when the regular season met its end last season was Brown. The Providential bruins improved over the course of the last month of last regular season by an astounding 584% (no, that is not a typo). Brendan Whittet's team earned points in February's contests at nearly six times the rate that it was earlier in the season.
Harvard? Well, Harvard was third-worst in the month of February in terms of its chilling. Only the engineering duo of Clarkson and RPI were feeling more left out in the cold than was the ultimately victorious Crimson. The tournament is played in discrete episodes, not holistically. Using the lens of which games occurred, how accurately would comparing heating and chilling figures of opponents predict which team emerged victorious?
Hockey lore says that those teams playing their best hockey at the right time will emerge victorious in the postseason when legacy is on the line. This aphorism gave rise to this very model. A team that was heating in February, or at least warmer, should unseat a team that was chilling in February, or at least colder, if the two meet in the ECAC Hockey tournament.
This heating-and-chilling model would have predicted 36.4% of the 11 eliminations in the 2015 Whitelaw-Cup playoffs. The team warming or warmer in February defeated the team chilling or colder in February in only about four of ten rounds. A closer look at the data yields a more promising and sensical understanding.
The ability of this model to predict results over the entire tournament is clearly limited. Reader, consider that the proverbial coin-flip with its even odds provides better prognostication in choosing winners. However, if one looks at first-round match-ups alone, the games played without a weekend of rest between them and the regular season, one observes a marked improvement in the model's ability. The heating-and-chilling paradigm is 27.2% more accurate in predicting first-round results.
No single upset happened in the first round for which the model would not have put the home team on upset notice with the home team's facing a warming or warmer opponent. In fact, in an entire month of the tournament, the only upsets that occurred for which the model would not have put the higher seed on notice involved Harvard. The model anticipates which series likely may yield an upset even if it cannot predict faithfully those upsets's occurence. This is not entirely devoid of value even if it is not as probative as myth may imply.
One other interesting truism emerges from last season's tournament. A chilling team, one whose play in February produced a rate of earning points less than 100% of its rate in the first three months of the season, lost every contest to a warming team, teams on the other side of the 100% demarcator, except for one circumstance. That circumstance was Harvard. The Crimson in its four-round run thrice beat teams on the opposite side of the warming/chilling divide. No other team did.
The heightened applicability of the heating-and-chilling model in the playoff's earliest rounds makes sense. Teams whose play directly abuts the month of February without a hiatus are those most likely to begin playing in the playoffs in ways most similar to their end-of-the-regular season form. The further a team's play gets from February by either a deeper playoff run or rest, the less able the model is at predicting results. Two trends guide any below predictions.
Colder teams's hosting warmer teams should place the former on upset notice. Warming teams beat chilling teams even if warmer does not necessarily beat colder. What of the Harvard exception? Perhaps, history matters. The general conclusion is that the cliché about best hockey-and-right time may be hackneyed or in need of retooling. This season will provide new data.
Predictions
Clarkson made huge moves, a three-slat climb, in the last weekend of the regular season. Additionally, the North Countrymen spent each week of February as one of ECAC Hockey's warming teams. Trust me, this writer has tremendous respect for the way that Ron Fogarty's teams play, but out of deference to the model, one conclusion is unavoidable. Anyone who pencils in a Princeton upset either is playing the odds or knows something that mere statistics cannot bear. Clarkson wins.
The other match-ups all place the home teams on upset notice to varying degrees. The greatest disadvantageous disparity of the three is between Brown and RPI. The Bears are 105 percent-degrees warmer than are the Engineers. RPI had twice as many wins as did Brown in February. The Bears avoided defeat in one more contest over the same span. Seth Appert in four attempts never has escaped Houston Field House. Considerable weight is on the side of an upset in this series.
Colgate accumulated two-thirds of its February wins in the last weekend of the regular season. Dartmouth ended the regular season getting swept. The model is no kinder to the Big Green than was the regular season's last weekend. A 95.2 percent-degrees differential that crosses the warming/chilling line means that in accord with this model, Colgate upsets.
The Cornell-Union series presents interesting quandaries. Cornell ended February earning points at exactly the same rate that it had during the regular season's first three months. This is largely a product of the law of averages's dragging down the Big Red's phenomenal start with a zero-conference win January. However, only Quinnipiac and Yale avoided defeat better than did Cornell in the warm-up season's last month. Cornell may not have won like it, but it did not lose like a top-four team.
Avoiding defeat and not losing are insufficient in the playoffs when there must be a winner. The carnelian-garnet clash is the only first-round series in which teams on the same side-ish of the warming/chilling measure meet. Cornell is reported as cold in the above graphic because of its trajectory, a one-point weekend at Lynah Rink, during the regular season's last weekend. It is more accurately room temperature. What remains to be seen is whether its temperature resembles that of a corpse or a slightly cooled ready-to-consume feast.
40 percent-degrees separate Cornell and Union with the latter's warming being greater. The other three first-round series average a percent-degrees spread of 104%. The series's parity is obvious. Upset is likely. As opposed to the series at Houston Field House and Thompson Arena, the model gives Cornell a fighting chance. One who uses last season as a guide may predict a Cornell victory. Home-ice advantage held sway each time a warming team hosted a warming team.
This writer is not as optimistic. Hopeful? Yes. Certain? No. Last season gave this Lynah Faithful a crisis of playoff faith.
In the spirit of the heating-and-chilling model, assume that Cornell advances. The whole point of this is to predict what the field will look like as the East's best programs race toward picturesque Lake Placid. Clarkson's next round would not put it far on the road as the Golden Knights would head down Route 11 to the barn of the Saints. Colgate would try to find a cavity in the defense of Alex Lyon and Keith Allain at Ingalls Rink. Hamden would host a reenactment of the 2013 ECAC Hockey Semifinal between Brown and Quinnipiac. Well, the heart wants what the heart wants.
Cornell and Harvard would have their first postseason meeting in what seems like an eternity (just four seasons). The Red would challenge the Crimson at Bright-Landry Hockey Center in both teams's times of year for the first time in 22 years. Before the Faithful can worry about return trips to Lynah East, the carnelian and white need to blot out garnet.
Does this writer think this model is predictive? Probably not. Its purpose was to test a truism. The value of that truism will be adjudged over the next three weeks of great hockey. History (Harvard's eight Whitelaw Cups before last season) and time (the tournament may grow unpredictable as time passes because of a mix of playoff experience and rest) confound using play in February to quantify the best hockey-right time metric used in this model.
Will rest save a chilling Quinnipiac team? Can bulletin-board material and history propel Cornell to an encore victory from the first round that rivals Harvard's Whitelaw-Cup win last season? What this model may make clear over the course of the season's best month is the tautology that it will not be apparent which team is playing its best hockey until the first playoff pucks bounce off the East's frozen battlefields.