July 18 2012 01:40PM
By: Stephen Cooper
Ever since Gabe Desjardins calculated the first NHL scoring equivalencies from NHL feeder leagues, the use of NHLE has been standard for “fancystat” analysis of prospect players. The original system has been refined by many, including Bruce Peter of Puck Worlds and Rob Vollman of Puck Prospectus. But the basic system of looking at only goals and points on a per game basis has remained essentially the same since the venerable and terrifying Desjardins made his first foray on the subject.
But any follower of hockey advanced stats worth their salt will tell you that a lot of useful information can be derived from shot based metrics which have the advantage of significantly larger sample size per game to make evaluations from and are more constant year to year than goal based measurements. Bringing the statistical strength of shot based metrics to league translations would significantly enhance the amount of data that hockey statisticians can bring to bear to making prospect statistical evaluations. To that end, I have examined the relationship between the available shot based measurements across translations from the best understood NHLE league, the AHL and the NHL.
Player Shots per Game Translation
First, we want to examine the relationship between shots per game at the AHL vs the NHL level. A dataset was assembled by selecting forwards who had played consecutive seasons in the AHL and NHL after 2004 (the earliest point individual shot data was collected in the AHL) and also belonging to the post-lockout NHL era. As criteria for selection, players must have played more than 25 games in both leagues post-lockout. The resultant dataset included 113 players to study.
For guys who simply played in the AHL one year and the NHL the next, the shot rate is compared from both seasons. For skaters who played in both leagues in one season, these "transition" seasons under 20 games are ignored and two seasons compared as if they played in one league one year and the other the next. For the more complicated players that had multiple consecutive seasons of playing significant (10+) games in both leagues, the results from each season in the separate leagues are added together as if they were one season each and compared against each other. The data graphed results in the following chart:
The graph demonstrates that AHL shot rates corresponds with shot rates in the NHL, albeit with high degree of variance. This is to be expected given the vastly different ice-time and linemate situations that NHL rookies would find themselves in; from buried on the 4th line to riding shotgun with much better players on an offensively oriented unit. The average conversion rate works out to .5607 or about 56% of shooting rate is maintained. From our R squared value of .36 we can see that AHL shot rate is a significant predictor of NHL shot rate, but far from being the sole, dominant variable.
Player Shooting Percentage Translation
Using the same player dataset, we can look to see how the other component of a player’s goal scoring, shooting percentage, translates from one league to the next.
As we can see, the co-relationship is far weaker for shooting percentage, much like how it is subject to severe year-to-year fluctuations in the NHL. The average forward keeps .79 of his shooting percentage but with so little correlation that a season’s worth of AHL shots is not very predictive of a season’s worth of NHL shots. It does however neatly multiply with the .56 of the shot rate to produce .44 or the NHLE for points for the AHL (.79 x .56 = .44). This serves as good independent confirmation that no gross mathematical errors were made.
Current Prospect Projections
What does this mean for next year’s potential graduates from the AHL to the NHL? Here is a table of drafted forwards playing in the AHL last season and their projected goal and shot rates for the upcoming season. A hypothetical shooting neutral goal scoring projection was done by projecting a 11% shooting percentage (fairly typical for a scoring forward) for every player.
|Prospect||Shots||Games||AHL Shot Rate||Projected NHL Shot Rate||AHL GPG||NHLE GPG||Shoot% Neutral NHL GPG|
Palmeiri for Anaheim looks to be a very strong bet to be a good NHL goal scorer. Calder watchers should keep an eye on him for a potential break out rookie year in 2012-13 and the only player that easily projects to be a 20 goal man in 82 games (.29 x 82 = 23.8).
AHL to NHL Save Percentage
To finish up my look at shot based AHL to NHL translations, I looked to see how a goaltender’s save percentage translates from one league to the other. Much like the previous data set, I collected info from currently active NHL goaltenders that had at least one consecutive season in the AHL and NHL of 20 games or more since 2001. 39 goaltenders met these criteria. Two seasons of AHL and NHL save percentage was collected from these goaltenders and the average for each population compiled.
|Year 1||Year 2||Average|
This was repeated with the total save aggregate save percentage from each relative year calculated but it did not result in a significant difference. As you can see there is about a .7% loss of goaltender save efficiency between the two leagues that appears stable for this sample of goaltenders. This calculates to an 8.4% increase in shooter efficiency between the NHL and AHL. The average results of two season’s worth of save percentage form each goaltender was compared graphically:
By request, a paired T-test was performed for the two-year average save%.
- Standard def: .0081
- Standard error: 0.0013
- Standard def: .0090
- Standard error: 0.0014
t = 3.2264 df = 38 standard error of difference = 0.002 Two tailed P value: 0.0026
Very significant difference. Using a two sample T-test reveals essentially the same result. Probability of AHL SV% being greater than NHL is 99.89%.
- The principle predictable difference between AHL and NHL goal scoring comes from opportunies to score generated by the player, although an increased difficulty to convert a shot into a goal against an NHL defense and goaltender appears to be a significant, but much less predictable factor.
- Shots per game rates in the AHL are a significant but not dominant factor in determining a player’s future NHL shots per game performance, which is obviously a factor in future NHL goal scoring ability
- The average NHL shot is at least ~8.4% harder to stop than the average AHL shot. However, the effects of defenses and goaltender improvement which may correlate with promotion to the NHL are unknown and may mask a greater degree of shot difficulty between the two leagues.
- Predicting NHL results from even two seasons worth of AHL save percentage data appears to be very unreliable. Goaltender save percentage appears to be too volatile to use to make those sorts of predictions, although there appears to be at least some minor loss in save percentage when a goalie jumps between the AHL and NHL.