Thursday, March 6, 2008
Taking the Hockey Media to Task: Cox and Proteau
The world of journalism is rarely deficient when it comes to favourably quantifying ones argument. In different terms, writers spin and manipulate numbers to support their case in the same manner other professionals do. Lawyers, research scientists, doctors, salesmen all fudge a number or two every now and then. The simple explanation for this phenomena: because they can.
Which brings me to this piece by Damien Cox in the Toronto Star (and subsequently, Adam Proteau of THN):
“The most compelling stat to emerge from almost two full seasons of mandatory visor use in the AHL has been a more than 40 per cent drop in major stick penalties and a more than 15 per cent drop in stick minors. ”
Ohh?? The use of the word almost is very sneaky. In this case, almost equals ¾ of the season. To get around correctly assessing the figures he presented (shown below), Cox also sneakily denotes the 2007-08 stats as “(to date)” to further substantiate his otherwise weak data. Here is the raw data (courtesy of the AHL), but reproduced in Cox's column:
2005-06 (pre visor rule): 757
2006-07 (post visor rule): 632
2007-08 (to date): 553
2005-2006: 8.95 min./game
2006-2007: 8.22 min./game
2007-2008: 9.02 min./game
As was mentioned, the number from this season requires adjustment by a factor of 1.25, as 25% of the games are yet to be played. (They play an 80 game season in the AHL, and to date, most have played 60. 60/80= ¾, or 75%). In order to correctly assess these data, we'll project these stick infractions to seasons end.
Projected after 80 games (or a full AHL season):
High-sticking Minors: 691
Fighting: 9.02 min/game (unchanged as it is already averaged on a per game basis)
Doesn't that change the landscape significantly? We'll start with high-sticking minors. When you correctly adjust the numbers for valid comparison, you see that the value lies almost exactly between that of the previous year (06-07) and the last year of voluntary visor usage (05-06). While three years is a CRAPPY data sample under any circumstance, it would seem that (contrary to the assesment of Cox and Proteau) the 06-07 season was the anomaly and that things are beginning to normalize.
Moving unto high-sticking majors, which are notably subject to judgment from the referee. Infact, Rule 61 of the NHL Rulebook surprisingly offers no direct characterization of a 'high-sticking major'. This in and of itself should alert one to the potential for inconsistencies in the data. Anything of this nature leaves itself vulnerable to a substantial degree of variability, as is demonstrated by the above figures. Consider how many refs will call a given high-sticking infraction a double-minor in lieu of a 5min major, or vice versa? Thus, the discretion awarded to the official limits any definitive statistical conclusions. That being said, much like the high-sticking minors addressed above, the adjusted figure for this seasons high-sticking majors (71.5) lies significantly above last seasons tally, and a mere 34% below that of the number provided for the pre-mandatory visor year(109 v. 71.5).
Fighting is up? Not really. It would appear to be leveling off (from last years dip) and returning to levels observed in the 05-06 season. Since fighting is up marginally over the previous season, but nearly equal to that two seasons prior, the numbers do not allow us to conclude that visors have ANY effect on the amount of fisticuffs -- a factor that Cox intentionally overlooked. Fighting is a multi-variable event that is dependent on a dozen or more factors unrelated to visors in any way. What is the score of the game? Are they rivals, or in a playoff hunt? Do the coaches dislike each other? Have the refs lost control of the game, or was their a dirty play that sparked a line brawl, or two? The intangibles that go into a fight kicking off far outweigh the fact that all players are wearing faceshields.
I suppose the moral of the story is best summarized from a quote via Leonard H. Courtney: “There are three kinds of lies: lies, damned lies, and statistics.” Whether you want to believe my numerical manipulation(s) or Cox/Proteau's is up to you, but my point about statistics has been well illustrated -- beware.