We have published 29 projections since the Snepsts system was first developed three years ago, and they've been off by an average of just 1.25 points overall, which is excellent. The standard deviation has been 14.9 points on a group of players that average 68.2, which obviously doesn't sound very good at all.
In fact, eight of the 29 projections have been off by at least 20 points, the worst of which is Alex Tanguay, who scored 69 points in 79 games last year despite our projection of just 42 points in 82 games. Based on these numbers, 35% of projections will be off by at least 15 points, which doesn't sound like a terribly helpful system.
To be fair, VUKOTA was off by an average of 8.38 points on this same group, with a similar standard deviation of 14.6 points, with seven of the 29 projections off by at least 20 points, the worst of which was Nicklas Backstrom, who managed just 65 points in 77 games despite being projected for 104 prorated over 82 games (Snepsts called for a more reasonable 86).
So perhaps the problem isn't Snepsts or VUKOTA, but rather the sample group itself. These weren't randomly-chosen players, but in fact high-scoring players who were in the news for various reasons, including being traded, and/or having particularly good or bad seasons.
Snepsts vs. VUKOTA
Instead, let's evaluate Snepsts against all eligible NHL players last year. To be eligible, a player requires at least two seasons worth of data, though it is far more accurate when you have more. We'll also restrict the comparison to the 267 players with at least 10 historical matches.
In that much larger and more random group, Snepsts was off by an average of 2.4 points, with a much smaller standard deviation of 10.0 points, due to their lower-scoring nature overall. Now, 87% of them were projected within 15 points instead of 65%, and 71.2% within ten points and 45% actually being nailed with five points.
Compare that with VUKOTA, which was off by an average of 3.3 points, with a still-smaller standard deviation of 9.6, the exact same percentage of 87% were within 15 points, a higher 73% within ten points, but a smaller 40.4% within five points.
Generally speaking, Snepsts and VUKOTA are of similar accuracy, although in fairness to VUKOTA, Snepsts does not predict games played, cannot be applied to any player but just those with at least two seasons worth of data, and gets its butt kicked in cases where there are few historical matches.
Surprises and disappointments
According to Snepsts, the biggest surprises last year included included a couple of Ducks and a pair of Boltsperhaps the sunny weather confuses the system?
Biggest surprises according to Snepsts
Player PTS/GP Snepsts Diff
Lubomir Visnovsky 0.84 0.39 0.45
Alex Tanguay 0.87 0.45 0.42
Corey Perry 1.19 0.83 0.36
Clarke MacArthur 0.76 0.41 0.35
Claude Giroux 0.93 0.58 0.35
Martin St. Louis 1.21 0.87 0.34
Teddy Purcell 0.63 0.29 0.34
Alex Ponikarovsky 0.25 0.59 -0.34
Tomas Kopecky 0.52 0.20 0.32
Alex Ovechkin 1.08 1.38 -0.30
In the end, it's no surprise that the Snepsts System matches VUKOTA so well, since the latter is partially based on the same type of approach, and the two systems were developed by colleagues who have worked closely for so long and have historically been of such similar mind (Tom Awad and me). In fact, if you take the average of Snepsts and VUKOTA, you wind up with potentially the most accurate projection system there is:
The best results from a combination of Snepsts and VUKOTA?
Snepsts VUKOTA Average (both)
Average difference 2.4 PTS 3.3 PTS 0.4 PTS
Standard deviation 10.0 PTS 9.6 PTS 9.6 PTS
Within 15 points 86.9% 86.9% 87.6%
Within 10 points 71.2% 73.0% 75.2%
Within 5 points 45.1% 40.4% 44.6%
Harold Snepsts is an NHL Vancouver legend. A classic stay-at-home defenseman, Snepsts only once topped 23 points in his 17-season career that also included stints with the Minnesota North Stars, Detroit Red Wings, and the St. Louis Blues. Though lacking the finesse of the higher-priced talent, Snepsts won over the fans with his hard-working and physically intimidating defensive playnot to mention his colorful personality and trademark moustache.
Despite his lack of scoring, Snepsts competed in two All-Star Games thanks to his sound defensive play that also resulted in being named Vancouver's top defender virtually every year until a famous and costly turnover in the 1982 Stanley Cup, and a series of nasty injuries in 1982-83. One of the last NHL players to compete helmetless, Snepsts wasn't worried about suffering a head injury, joking "if it happens, I could always return as a forward."
In the end, Snepsts played 1,033 games, scored 233 points, and racked up 2,009 penalty minutes, 54th all-time. As a Canuck, Snepsts was fourth in games played with 781 before being passed by the Sedins last season, and fourth in penalty minutes.
Following his NHL career, Snepsts coached the Peoria Rivermen of the IHL to a 105-point season in his very first attempt at the job, before jumping to the NHL to assist Bob Plager of the St. Louis Blues in 1992-93. As of 2000, he's been a scout, first for the Central Scouting Service, and then for the Vancouver Canucks, who inducted him into their "Ring of Honour" this past March.
Just as Mick Vukota was not involved in the development of the projection system that bears his name, Harold Snepsts had no involvement in the development of this one. As an iconic hockey player to a generation of loyal fans, respected for making it based more on hard work and dedication than natural talent, it's our honor to name the system after him in tribute.
Robert Vollman is an author of Hockey Prospectus.
You can contact Robert by clicking here or click here to see Robert's other articles.