June 12, 2009
Numbers On Ice
League Equivalencies and GVT

by Tom Awad

One of the strengths of GVT is that it is an automatic normalizer. Since GVT naturally accounts for schedule length, rates of goals per game, and average save percentages, it can be applied to the statistics of any league, not just the NHL, and produce meaningful results. Here, I will attempt to apply Gabriel Desjardins’ league equivalency principles using GVT to see if we obtain useful insight. If so, it would give us yet another tool to predict how players coming from European leagues or the AHL will perform in the NHL.

How To Compare League Strength

Gabriel Desjardins has done the most exhaustive analysis of league equivalencies that I know of, so I will resume his method here. He has compared points-per-game rates among players who play in one league one year and in the NHL the next (or vice-versa); since their skill level should stay reasonable similar from one year to the next, the difference in PPG rates should be entirely due to league difficulty. Using this methodology, he has arrived at league difficulty rates of roughly 0.45 for the AHL and 0.55 to 0.6 for the European Elite Leagues. Upon seeing this analysis, I had two questions: was the simple PPG conversion valid, and how would GVT lend itself to this analysis? First, a caveat: when comparing European leagues, one has to beware of what my colleague Iain Fyffe has dubbed the “Crossing the Atlantic” effect. Namely, European players who come over to the NHL have to adjust to a different style of play, a different culture, and many other factors which prevent them from immediately stepping up to their maximal level of play. European hockey may also lend itself to different skills: as anecdotal evidence, Kevin Dallman, a defenseman who never wowed in 3 seasons in the NHL with Boston, St. Louis and Los Angeles, led KHL defensemen in scoring last year, and had more points than either Jaromir Jagr, Alexander Radulov or Alexei Yashin.

Luckily, the NHL provided us with a fantastic “natural experiment” during the 2004-05 season by shutting its doors. While we rarely get to see top NHL talent in the AHL or the Swedish Elitserien, we did during that year as NHL players either returned to their home country or sought employment where it could be found. While this slightly skewed the strength of these leagues during that year, the number of NHL players was small enough that it did not significantly change things. Indeed, the statistics of players already in the leagues were hardly affected.

Here are the results for PPG for the AHL, Swedish Elitserien (SEL), Finnish SM-Liiga (FNL) and Czech Extraliga (CRL) for the years 2004-06:

PPG conversion rates, 2004-2006

	Slope	Intercept
AHL	0.50	-0.02
SEL	0.81	-0.02
FNL	0.64	-0.07
CRL	0.70	 0.01

Math note: for those unfamiliar with the slope/intercept concept, the equation is: y = slope * x + intercept. So if a player scored 0.9 PPG in the FNL and you want to convert this to an expected NHL rate, you calculate 0.9 * 0.64 – 0.07 = 0.51.

What does this mean? To convert a player’s PPG rate from the AHL to the NHL, multiply it by 0.50 and then subtract 0.02. This shows that higher PPG rates are more likely to be maintained. This makes sense: while a player who scores 0.3 PPG in the AHL is unlikely to be able to achieve even half that in the NHL, a player who scores 1.2 PPG may well achieve 0.6 in the NHL. So far, I’ve mostly only reproduced Gabriel’s work.

What About GVT?

I was curious to see how GVT would apply itself to this analysis. Luckily, since all these leagues provide +/- and SOG data, we can at least perform a decent calculation of GVT data on all these players. I chose to calculate GVT/G instead of GVT/ice time, for two reasons. First of all, my estimates of ice time for the non-NHL leagues are very coarse, so I would be worried that they are skewing the results. Secondly, my colleague Rob Vollman pointed out to me that if you are a first-liner in your original league, but only good enough for a third-line shift in the NHL, that’s telling you something as well. Therefore I have stuck to GVT/G for the time being.

I must admit that the results surprised me:

GVT/G conversion rates, 2004-2006

	Slope	Intercept
AHL	0.53	-0.04
SEL	0.59	 0.01
FNL	0.49	-0.05
CRL	0.50	-0.01

I was expecting GVT to be better conserved than PPG, but this turns out not to be the case. Interestingly, the leagues yielded a much closer result of GVT conversion, all with slopes between 0.49 and 0.59. Once again, the intercepts are negative, as expected: a replacement level player in the Finnish SM-Liiga should be a below-replacement level player in the NHL.

These average slopes hide a wide range of results. As an example, here is a graph of every AHL / NHL player during this period, with their AHL GVT/G on the X axis, NHL GVT/G on the Y asix, and predicted NHL GVT/G:

Thought Experiment

What these results tell us are that a 2nd line AHL player, with a GVT of about 10 over the course of a season, will be a replacement level player in the NHL, with a GVT of approximately 1, while a top scorer in the league, with a GVT of 20, will on average achieve a GVT of 5 or 6. This seems to mesh with everyday experience, where top-line AHL call-ups do manage to hold their own in the NHL but rarely dominate.

This got me thinking: how would an AHL team fare in the NHL? Conversely, how badly would an NHL team cream the competition if it was “demoted” to the AHL? The math seems to imply a replacement level of about 80-100 NHL goals, or conversely 160-200 AHL goals. Which one is it? History, mathematics and instinct tell me that the defense would underperform more than the offense. Thus, an AHL team in the NHL would probably have an offense that is below average by 40 to 50 goals, while its defense would be would give up between 80 and 100 goals more than average, for a total goal differential of -120 to -150 and approximately 30 to 40 points in the standings. While this doesn’t sounds that bad, it would be worse than all but a dozen or so teams in league history, some of which, like the Washington Capitals of the 70s, were stocked with talent that would be pretty marginal by the standards of today’s AHL. Players like Justin Abdelkader and Darren Helm, both of whom have impressed in the Stanley Cup Finals, were less than point-per-game players in the AHL.

Conclusion

The benefit of having a conversion method that goes beyond points-per-game is that it allows us to make more enlightened cross-league projections. When a player has little or no track record in the NHL, it’s nice to be able to get an idea of how he will perform in the league. Cross-league GVT still won’t help you pick out the next Teemu Selanne, but it’s a start.

Tom Awad is an author of Hockey Prospectus. You can contact Tom by clicking here or click here to see Tom's other articles.