I mentioned Crosby as people often extrapolate out his shortened seasons, particularly 2011-12 as some of the best in the cap era. But as you stated small sample sizes can be problematic.
Ok fair enough but there's also there's an important distinction to be made for goals and assists, especially in regards to the upper echelon players. Goal scoring is more prone to fluctuations compared to points. Take Crosby in 16-17, the way he start off that season had people talking of a potential 60-goal year, something he ofc ultimately fell well short of. He had 27 goals in his first 37 games, a scoring rate of 0.73/gm but in his last 37 games he had 17 goals for a scoring rate of 0.46 That's a decline of 59% But while his goal scoring pace declined tremendously, his point totals in contrast only declined by 28% While that is still a large decline, it's less than half as much as his goal scoring declined by. When one aspect of their game declines the greats are able to compensate by focusing on another aspect, just like McDavid is this season right?
Secondly it's unreasonable to make predictions to a very specific benchmark. First off the number who were on pace for 70 goals drops from 31 to 24 when you use the much better 60 games sample size vs the 41 game one, but a further 7 of those 24 seasons were just barely on pace for it with a 2% margin of error(and no other season within 2% of it). If we went with projections to 80 games, only 17 seasons were on pace for it. But interestingly enough, in an 80 game schedule more seasons were on pace for 100 assists - 18. While 82 games is the correct sample size to use now, is it really more correct than using an 80 game projection which is when nearly all of the 70 goals and 100 assist seasons happened? Anyhow, all that said and dispute my repeated posts in favor of goals, I do think the feats of 70 goals and 100 assists are very close - I said so much before starting this thread. I've only been drawn back in repeating into this argument because of the brashness of a certain fan base.
I’m of the opinion that extrapolated totals are even more meaningless than any arbitrary games played cutoff to project from. It’s really impossible to predict. I made a list at the end of last season of the best points per game a player reached with at least 70% of an 82 games played mark (57 GP). I think that’s “better” than 41 games but at the end of the day it’s hard to pick one. If I picked 80% of 82 games played then ‘93 Lemieux and other phenomenal seasons wouldn’t make the cut so I thought close to 60 games as the cutoff made sense.
The question I’m asking you is what do you think? What sample size do you think is sufficient? I genuinely am interested what you think so I’ll defer to you here.
This is something I spoke to in that Lemieux vs Gretzky thread where I stated that I agree a 1/2 season sample size isn't large enough for an accurate extrapolation, but that a 3/4 size is especially when further context can be applied. That further context being a players age and their scoring constancy. A player in his peak is more likely to maintain a similar scoring pace over the course of a season, particularly an exceptional one eg; 25yo Gretzky in his 8th year; 2.575@40gm, 2.70@gm60 2.69@final, or 26yo McDavid in his 8th year; ppg 1.875@40gm, 1.88@60gm, 1.865@final. And a player who scores points in 90% or more of his games is likewise indicative of being more likely to sustain his scoring rate than one who is scoring in only around 70-odd% of his games.
20 extra games may not on the surface seem like a big difference, but the potential for variances in pace actually decreases in a greater than linear fashion due to compounding. For example at the half way point of the 95-96, Lemieux was scoring at a rate of 2.66ppg, which is
15.5% higher than his end of season result. It was well known(in hindsight) that he was running hot to begin the year and it was in fact by far the least consistent season of any year he scored over 100 points. But by the 50 game mark he was at 2.4ppg, which was off by just
4.3% of his final rate and by the 3/4rd point of
his season(gm 53 of 70) his scoring rate diverged by only
3.4% Getting off topic here, but over the course of an 80 game span a difference of 15.5% would result in a massive difference of 29 points(213 vs 184!) That kind of accuracy isn't very useful. But 3.4%? That's a difference of a mere 6 points(190 vs 184). I believe that's now a worthwhile point, where I would draw the line is with anything more than a single digit/5%. For the record the difference in Lemieux's scoring rates at the 3/4rd mark(of his gms) in his big near-full seasons was 2.9%('88 @gm58of77) 3.2%('89 @gm57of76) 3.4%('96 @gm53of70).
Another example was the Crosby season I mentioned above. His scoring rate from the halfway mark of the '17 season(gm37 of 75) to his final mark diverged by
13.9% But the difference between his mark from game 55 to the end was just a mere
2.6% and from game 60
1.7%. I think it's very clear that there's a
massive difference in accuracy of projections between using a half season sample verses using 3/4rds of a season.