In my well-worn copy of Bill James' 1985 Baseball Abstract, I remember first reading about Major League Equivalencies (MLE), which were designed to show how to use a player's minor league data to reliably predict their major league statisticsor at least as reliably as if it were previous major league season data.
Then in 2004, Gabriel Desjardins first introduced the idea to hockey. (What took him so long?) You essentially divide a player's scoring in the NHL by the player's scoring the season previous in the target league, and average that for all the players who made that move.
We have expanded on those ideas on this site, and in the pages of Hockey Prospectus 2010-11, and with the new season up on us, it would be helpful to revisit some of the questions that were either left behind, or that have bubbled up since.
Is it linear?
Though we've greatly refined NHL league translations over the years, most notably by splitting up goals and assists and breaking it down by age, we've never broken them down by scoring level. We've been treating them as linear, but are they really? Is it reasonable to use the same translation factors for a third-line checker that you would for a top star, or to apply the same translation for a high-scoring winger as you do for a stay-at-home defenseman?
A picture can say a thousand words. Since the Lockout, there have been roughly 500 players that have played at least 20 games in the AHL the season before playing at least 20 games in the NHL. Here's the distribution, and remember that the long-established AHL-to-NHL translation factor is roughly 0.45 for both goals and assists.
Though it's a little hard to tell from the point graph, the NHL points per game appear to have a strong tendency to stay on the south side of the 0.4 line regardless of how high scoring was in the AHL. That has the net effect of arcing the curve down the further you advance, yielding far different translation factors when you chop up the player by their AHL scoring levels. Here's how noticeable it looks at a high level, when you split the players up and calculate translation factors separately:
AHL to NHL translation factors, at different scoring levels
Points/game Count Translation Factor
1.00+ 52 0.34
0.75 100 0.35
0.50 158 0.47
0.25 129 0.53
0.00 57 0.73
The translation factor changes depending on scoring levels. Of the 52 players who scored at least a point per game in the AHL, only 12 kept the expected 45% of their scoring in the NHL. Compare that with the more average 0.50-0.75 points per game group, where over half (85 of the 158) exceeded that same expected 45%.
It's clear that using 0.45 as your blanket AHL translation is bound to overestimate the production of higher-end players, and underestimate that of the lower scorers. The AHL is especially beloved because the huge volumes of data leave very few blanks, but the trend is likely consistent across other leagues.
Take the KHL for instance, one of the NHL's closest peers, with translation factors far higher than the AHL's. Two years ago, Jiri Hudler was a point-per-game KHL player, while Zherdev was close behind at 0.75, but each of them barely kept half of their scoring in the NHL last season. On the other hand, lower-scoring KHL players like Linus Omark, Anton Babchuk, and Nikita Nikitin all kept at least three quarters of their scoring when making their move to the NHL. Keep a close eye on Jaromir Jagr this season, because based on the AHL study and this anecdotal evidence, Jagr may finish closer to 40 points than 70 points.
Another interesting future side study would be to look at the AHL players who exceeded expectations when they came to the NHL, to look for commonalities that might help us identify others that may do the same. The best anecdotal examples of players who exceeded their AHL-based projections include:
1. Kyle Quincey scored 20 points in 66 games for the Grand Rapids Griffins, typical of his three-year, highly-penalized stint in the AHL. When he jumped to the NHL at age 23, he scored 38 points in 72 games for the Los Angeles Kings. He has since scored 30 points in 100 games with the Avalanche.
2. T.J. Galiardi scored 27 points in 66 games for the Lake Erie Monsters at age 20, and then jumped to 39 points in 70 games on the miracle 2009-10 Colorado Avalanche.
3. Anthony Stewart scored 31 points in 77 games for the Chicago Wolves at age 25, very typical of his performance at that level. Despite just 12 points in 105 previous NHL games, the next year he scored 39 points in 80 games for the Atlanta Thrashers.
4. Bryan Bickell scored 31 points in 65 games for the Rockford IceHogs at age 23again it was not an off-yearand then jumped to 37 points in 78 games for the Chicago Blackhawks last year in the NHL.
Personally, I'm inclined to believe that these players simply had unexpectedly strong seasons, which just happened to coincide with their moves to the NHL. Indeed, in a random sample of 500 NHLers, it wouldn't be unusual to find a handful of players with unexpected jumpsand it remains to see if they're sustained.
The group who missed expectations by a wide margin is equally interesting. Picking another four standouts:
1. Jeff Tambellini was a consistent AHL point-per-game player when he earned a trip to the NHL with 76 points in 58 games for the Bridgeport Sound Tigers at age 23. Unfortunately, he managed just 15 points in 65 NHL games for the New York Islanders in 2008-09, a career high he has topped by only two points since.
2. Zenon Konopka was a consistent 20-goal, 55-point AHL performer for years, but earned just five points in 74 games for the Lightning in 2009-10 and nine points in 82 New York Islander games last season in primarily an enforcer's role.
3. Denis Hamel has 605 points in his 776 AHL games, including 56 goals and 91 points for the Binghamton Senators in 2005-06. His 2006-07 NHL season? Just eight points in 53 games for three different teams, barely earning six minutes per game.
4. Jiri Hudler, who we'd call enigmatic if he were Russian, beat Denis Hamel by scoring 97 points in 76 games at age 22 for the Grand Rapids Griffins, but scored just 25 points in 76 NHL games the next year for Detroit. He did work his way up to 42 points and then 57 points before a point-per-game season in the KHL, but came back to just 37 points in 73 games last year.
People come and go from the AHL for different reasonsyoung players needing the extra ice-time to develop, established players rehabbing injuries, or mediocre defensemen signed to ridiculous contracts. Expectations should be different for each of these groups, and it would be useful to devise a system more sophisticated that just multiplying everyone by 0.45.
Introducing league equivalencies to hockey seven years ago was a big step forward, and great work has been done to further revise the approach by breaking down goals and assists and factoring in age (as some examples).
The next step is an evolutionary one, where the linear approach is replaced with a more multivariate model that seeks to understand the context of a player's scoring in the other league, translate that to the equivalent NHL role, and then apply the necessary league adjustments. Until then, the best approach is to continue to use the existing translation factors for average players, and try to find comparative examples for the more talented.
Robert Vollman is an author of Hockey Prospectus.
You can contact Robert by clicking here or click here to see Robert's other articles.