Last summer I looked into this a bit. I didn’t determine specific projection rates, but, rather, did an experiment; take the final pick in the first several rounds of past drafts, and find who a team would have chosen if they had been basing their choices purely on point totals. I started in 1980, and got bored and quit once i reached the mid 90’s. The results were eye opening, as my selections dwarfed the actual drafting record of any team during the period. I managed to nab a ton of “diamond in the rough” North Americans; Gilmour, Larmer, Hull, Robitaille, Fleury, Recchi, Sullivan, St. Louis, Rafalski, amoungst other succesful NHLers. The defining characteristic of these steals seems to be a lack of size and/or speed.

Unless something has changed in the last decade, I’m guessing your study will show that scoring totals, for all their problems as a predictive tool, are better at forecasting NHL production than whatever the hell else scouts base their judgements on.

It’s absolutely maddening. I mean, take Rafalski. Not even drafted! The man was draft eligible for eight years and nobody took a chance on him. He scored 45 points in 43 games as a senior at Wisconsin, and nobody even thought to use a ninth round pick! St. Louis averaged nearly two points a game at Vermont, and not a look. Fleury had 129 in 66 his draft year, and fell to the seventh round. Hull had 188 in 56 and fell to the sixth. It’s rediculous that someone with nothing more than boxscore stats can outperfom NHL scouts so often, but that is what happens when assumptions are unshaken by results.

Jon: I think the weighted proportion of team offense is a good piece of data to fit within a multi-faceted analysis. I assume you’re familiar with Mr. Desjardin’s work and website, but for the sake of completeness I’ll post the page that I immediately thought of when I read your thoughts here.

http://www.behindthenet.ca/juniorproj.html

It seems apparent that the analysis needs to include the following factors:

- The quality of the prospect’s 17 season. Is it exceptional? Is there too much variance for an analysis at 17 to be reliable?
- Consideration of performance from year to year. I think that it seems obvious that player’s who don’t dominate at the age of 18 in junior are less likely to be elite or even above average NHLers.
- As mentioned by other posters above, age of the player compared to their cohort. What is the mean age for draft eligible players within their CHL league? Might be difficult to quantify in a definitive manner the extent of this factor’s importance.
- Weighted proportion of offense

It seems that in order to get a sense of perspective on all of these factors would require a substantial historical analysis as well. There may also be issues of conflation in the analysis because scoring rates are obviously not stable. This might be more of an issue for the more traditional analyses suggested by Desjardin, and a place where the weight proportion of scoring might be particularly useful.

In any case, getting down to a multi-factorial analysis of all these factors is quite the task and probably deserving of publishing in an academic sports journal. It might be more heuristically simple to start the process from Desjardin’s existing analyses, but then make it subject to your weighted proportion analysis and then further subject to a quantification of the player’s age relative to the mean for their draft cohort.

Just spit-balling though.

It would also be interesting to see whether size and height were correlative of success or translation of offense to the NHL.

A negative correlation is still a correlative relationship my friend, no?