Wounded Animals
Hockey goaltenders remind me of wounded animals in the posture they adopt to cover up a puck and stop play, oblivious to the 200-pound behemoths whacking, shoving, falling, and crashing away above them. Almost look like oversized wolverines.
I've been wanting to write something about the NHL playoffs for quite some time now and with 5 of the 8 first-round series closed out, here I go. I've been reading a bit about stock markets lately -- and specifically the utility of simple statistics in the form of financial statements and information distilled from them (and stock prices) in predicting how a basket of stocks will perform on average in the future. I got to wondering whether I could conflate these two lines of thought: are available tangibles in the form of hockey stats enough to predict the outcome of the Stanley Cup playoffs?
For starters, let's look at the history of the President's Trophy (introduced in the 1985-1986 season by the NHL Board of Governors). It is presented annually to the team with the best regular-season points record. Notice that in the previous 20 seasons (including this one, since the Good Lord Stanley hath smitten the Detroit Red Wings from the first round by His agent ye mighty Oilers of Edmonton), the President's Trophy is a mediocre predictor of who will make the Stanley Cup finals, correctly predicting a finalist 8 of 20 times (40%) and a champion 6 of 20 times (30%). Not the stuff great averages are made of (though, on the bright side, when the President's Trophy winner makes the final it historically wins 75% of the time; that's nice).
| Year | President's Trophy | Stanley Cup | Finalist |
| 2006 | Detroit Red Wings | ? | ? |
| 2004 | Detroit Red Wings | Tampa Bay Lightning | Calgary Flames |
| 2003 | Ottawa Senators | New Jersey Devils | Mighty Ducks of Anaheim |
| 2002 (*) | Detroit Red Wings | Detroit Red Wings | Carolina Hurricanes |
| 2001 (*) | Colorado Avalanche | Colorado Avalanche | New Jersey Devils |
| 2000 | St. Louis Blues | New Jersey Devils | Dallas Stars |
| 1999 (*) | Dallas Stars | Dallas Stars | Buffalo Sabres |
| 1998 | Dallas Stars | Detroit Red Wings | Washington Capitals |
| 1997 | Colorado Avalanche | Detroit Red Wings | Philadelphia Flyers |
| 1996 | Detroit Red Wings | Colorado Avalanche | Florida Panthers |
| 1995 (*) | Detroit Red Wings | New Jersey Devils | Detroit Red Wings |
| 1994 (*) | New York Rangers | New York Rangers | Vancouver Canucks |
| 1993 | Pittsburgh Penguins | Montreal Canadiens | Los Angeles Kings |
| 1992 | New York Rangers | Pittsburgh Penguins | Chicago Blackhawks |
| 1991 | Chicago Blackhawks | Pittsburgh Penguins | Minnesota North Stars |
| 1990 (*) | Boston Bruins | Edmonton Oilers | Boston Bruins |
| 1989 (*) | Calgary Flames | Calgary Flames | Montreal Canadiens |
| 1988 | Calgary Flames | Edmonton Oilers | Boston Bruins |
| 1987 (*) | Edmonton Oilers | Edmonton Oilers | Philadelphia Flyers |
| 1986 | Edmonton Oilers | Montreal Canadiens | Calgary Flames |
So, why is Detroit out so early? Are they just a bunch of pansies? Could it be that, as has been suggested by many people including me, their impressive regular season points total was compiled by victimizing hapless bottom-feeders like Columbus, Chicago and St. Louis (all of which share a division with Detroit)? Could it be as well that the NHL's point system, which hands out points not merely for winning in regulation time, but for battling to a draw, for winning in 4-on-4 overtime (not real hockey) and for winning in shootouts (not even a spectre of real hockey), presents a distorted picture totally at odds with the reality of 5-on-5 battle-'til-you-die playoff hockey?
Well, here's a new system. Let's award points only for wins in regulation (5-on-5) and each team gets one point only for such a win. Below is a table juxtaposing the NHL's rankings with this new system:
| NHL rank | Team | NHL points | Rank0 | Team | Points |
| 1 | Detroit | 124 | 1 | Detroit | 51 |
| 2 | Ottawa | 113 | 2 | Ottawa | 48 |
| 3 | Carolina | 112 | 3 | Calgary (+4) | 42 |
| 4 | Dallas | 112 | 4 | Buffalo (+1) | 41 |
| 5 | Buffalo | 110 | 5 | Carolina (-1) | 40 |
| 6 | Nashville | 106 | 6 | Nashville | 40 |
| 7 | Calgary | 103 | 7 | Dallas (-3) | 38 |
| 8 | Philadelphia | 101 | 8 | Avalanche (+5) | 37 |
| 9 | New Jersey | 101 | 9 | Mighty Ducks (+3) | 37 |
| 10 | New York Rangers | 100 | 10 | San José (+1) | 34 |
| 11 | San José | 99 | 11 | Vancouver (+6) | 34 |
| 12 | Anaheim | 98 | 12 | Philadelphia (-4) | 34 |
| 13 | Colorado | 95 | 13 | Montreal (+2) | 33 |
| 14 | Edmonton | 95 | 14 | New Jersey (-5) | 33 |
| 15 | Montreal | 93 | 15 | New York Rangers (-5) | 33 |
| 16 | Tampa Bay | 92 | 16 | Minnesota (+6) | 32 |
| 17 | Vancouver | 92 | 17 | Los Angeles (+3) | 32 |
| 18 | Toronto | 90 | 18 | Toronto | 31 |
| 19 | Atlanta | 90 | 19 | Tampa Bay (-3) | 31 |
| 20 | Los Angeles | 89 | 20 | Atlanta (-1) | 31 |
| 21 | Florida | 85 | 21 | Edmonton (-7) | 28 |
| 22 | Minnesota | 84 | 22 | Phoenix (+1) | 28 |
| 23 | Phoenix | 81 | 23 | Florida (-2) | 25 |
| 24 | New York Islanders | 78 | 24 | New York Islanders | 24 |
| 25 | Boston | 74 | 25 | Boston | 23 |
| 26 | Columbus | 74 | 26 | Columbus | 21 |
| 27 | Washington | 70 | 27 | Washington | 20 |
| 28 | Chicago | 65 | 28 | Chicago | 17 |
| 29 | Pittsburgh | 58 | 29 | Pittsburgh | 17 |
| 30 | St. Louis | 57 | 30 | St. Louis | 14 |
Well, although it seems like the table is right about a few things, it seems even wronger about others. Namely, it exagerates the supposed disparity between Detroit and Edmonton to an even greater extent. What about this picking-on-the-weak theory? Let's re-rank the teams using our previous ranking system as a point of departure (called Rank0). We're going to define a new ranking, Rank1 based upon Rank0 and the supposed strength of each team beaten by a given team as follows: for every regulation victory over another team, the victor is awarded the Rank0 of the losing team. Hence, if Detroit beats St. Louis, it gets a measly 14 points whereas if it defeats a strong team like Calgary, we give it 42 points. Beating Calgary is worth 3x as much as beating St. Louis now!
| Place | Rank0 | Rank1 | Rank2 | Rank4 | Rank7 |
| 1 | Detroit | Ottawa (+1) | Ottawa | Detroit (+1) | Detroit |
| 2 | Ottawa | Detroit (-1) | Detroit | Ottawa (-1) | Ottawa |
| 3 | Calgary | Calgary | Calgary | Calgary | Calgary |
| 4 | Buffalo | Buffalo | Buffalo | Dallas (+1) | Dallas |
| 5 | Carolina | Carolina | Dallas (+1) | Buffalo | Buffalo |
| 6 | Nashville | Dallas (+1) | Carolina (-1) | Carolina | Carolina |
| 7 | Dallas | Colorado (+1) | Colorado | Colorado | Colorado |
| 8 | Colorado | Anaheim (+1) | Anaheim | Anaheim | Anaheim |
| 9 | Anaheim | Nashville (-3) | Nashville | Nashville | Nashville |
| 10 | San José | Vancouver (+1) | Vancouver | Vancouver | Vancouver |
| 11 | Vancouver | San José (-1) | San José | San José | San José |
| 12 | Philadelphia | Montreal (+1) | Montreal | Montreal | Minnesota (+1) |
| 13 | Montreal | Minnesota (+3) | Minnesota | Minnesota | Montreal (-1) |
| 14 | New Jersey | New Jersey | Los Angeles (+1) | Los Angeles | Los Angeles |
| 15 | New York Rangers | Los Angeles (+2) | New Jersey (-1) | New Jersey | New Jersey |
| 16 | Minnesota | New York Rangers (-1) | New York Rangers | New York Rangers | New York Rangers |
| 17 | Los Angeles | Toronto (+1) | Toronto | Toronto | Atlanta (+1) |
| 18 | Toronto | Atlanta (+2) | Atlanta | Atlanta | Toronto (-1) |
| 19 | Tampa Bay | Philadelphia (-7) | Philadelphia | Edmonton (+2) | Edmonton |
| 20 | Atlanta | Phoenix (+2) | Phoenix | Philadelphia (-1) | Phoenix (+1) |
| 21 | Edmonton | Edmonton | Edmonton | Phoenix (-1) | Philadelphia (-1) |
| 22 | Phoenix | Tampa Bay (-2) | Tampa Bay | Tampa Bay | Tampa Bay |
| 23 | Florida | New York Islanders (+1) | New York Islanders | New York Islanders | New York Islanders |
| 24 | New York Islanders | Florida (-1) | Florida | Florida | Florida |
| 25 | Boston | Boston | Boston | Boston | Boston |
| 26 | Columbus | Washington (+1) | Washington | Washington | Washington |
| 27 | Washington | Columbus (-1) | Columbus | Columbus | Columbus |
| 28 | Chicago | Pittsburgh (+1) | Pittsburgh | Pittsburgh | Chicago (+1) |
| 29 | Pittsburgh | Chicago (-1) | Chicago | Chicago | Pittsburgh (-1) |
| 30 | St. Louis | St. Louis | St. Louis | St. Louis | St. Louis |
What I've actually done is created a recursive function Rankj where Rankk+1 is defined in terms of Rankk and a team's regular-season regulation wins. It's clear that each new solution tends to be closer to the true result than the previous one (e.g. if Team A beat Team B a lot and Team B had a lot of points in Rank0 because it had a lot of wins against weak teams, we'd expect Team A to have too many points in Rank1 but this number would decrease in future iterations as Team B's numbers decreased). In any case, without proving it mathematically, it seems obvious that this function has a solution that can be approximated to a high degree of accuracy via iteration. So much for the Detroit as bully to Columbus theory.
The function seems to converge around the 10th iteration: more or less as shown in the rightmost column but with the Wild up to 12th, the Oil up at 17th, and the Flyers down to 21st. If this ranking system were "better" than the league's point system, in the East Tampa and Philly would be out of the playoffs in favour of Toronto and Atlanta; in the West, the Canucks would supplant the Oilers. This of course is utter nonsense: Edmonton took Detroit out in 6 games of the first round, home ice advantage and all.
The first problem with stats like this is that they look at averages across an entire season, which may not accurately reflect the state of things as the playoffs begin. Injuries occur. Trades are made. Players are sent down to the minors or called up. Coaches and general managers are fired. Players and personnel deal with personal problems and the capricious and unpredictable human emotional state. Some teams are young and full of energy. Others are older and perhaps tire as the 82-game regular season and subsequent playoff grind wears them down. Furthermore, teams that have to claw their way into the playoffs are essentially playing playoff hockey weeks before the playoffs start: phrases like "staving off elimination" are all too familiar to them... Could this have worked in Edmonton's favour? Finally, at some point we have to ackknowledge the power of things like momentum and plain old luck.
I know there are some people in the world who are actually quite good at betting on sports pools of one kind or another. By both the league's own metric and my attempt at designing a new one however, Stanley Cup playoff series are too unpredictable to bother betting on. If someone has a better way of valuing teams relative to each other, I'd love to hear about it. As for the relation to the stock exchange, there really isn't one. If you could look at a season's stats for a "portfolio" of 30 years and attempt to predict series winners, finalists, and Stanley Cup champions in each of those years, you might come away not doing too shabbily. Unfortunately, the only place to get those 30 years of stats is in the past, where the winners are already known. I suppose it's not beyond the realm of possibility that a detailed study could be undertaken to attempt to discover and validate a method that could be used successfully on average going into the future. But it's hardly worth the while.
Having no idea whatsoever who's going to win, I say "Go Ottawa!!".
posted by Ghang! @ 11:31
0 Comments:
Post a Comment
<< Home