Something that has puzzled me since learning about Nate Silver's PECOTA for baseball, where they can predict how a player's career will play out based on previous similar players of that caliber, has been at what age do hockey players actually regress?
[that was a very, very baseline description of PECOTA]
Often we hear that hockey players are in their prime from ages 27-31, and it's downhill from there; but is that really the case?
In order to find out, based purely on their ability to compile points, I took a dive into the stats to find out.
Using only forwards, since this is a scoring metric, who have played in 10 or more games in a given season (since the lockout), I compiled the points/60 for each age that came up in the analysis (18-43). The ten game mark was used to remove any noise from players aged 18 who only appeared in 9 games before their NHL team opted to return them to Juniors as to not burn a year of their ELC.
On a quick full run, here's what I found:
What does this tell us? Well, nothing, really. If you took this at face value and ran with it, you might come away thinking that hockey players just don't regress as we thought they did.
In order to combat too broad of a stroke on this analysis, I used a tiered method to place players into categories of first, second, third, and fourth lines. This was accomplished by using TOI/G as an indicator of what line each forward played on by ranking and percentiling each player's TOI/G (75% - 100% = first line; 50% - 74.999% = second line; 25% - 49.999% = third line; 0 - 24.999% = fourth line).
Another factor that needed to be taken into account was frequency, so the chart had to get a little more in-depth. Below, you'll see four charts, one for each line. The left vertical axis is our frequency measure. The right vertical axis is our p/60 measure. The horizontal axis is player age.
So here are our first liners, and really, the distribution of p/60 over age really resembles our full data set pull, but these are the elite players. Our lowest performing age group are the 38 year olds, coming in at 1.71 p/60. It becomes difficult to really gauge regression on a p/60 standpoint, because the players who are up there in age like Jagr, St. Louis, Iginla, etc... These guys are still in the league, because they can still play the game.
A better indicator of the regression we are looking for could be the frequency distribution. Almost normally distributed around the 26 year olds, with the mean age coming in at 27.4. There is a serious dropoff after age 30 in terms of players who are being deployed by their teams 5v5 as first liners.
It looks like the major drop off for second liners comes at the age of 28. Are these players jumping up to the first line, being relegated to the third line, or finding themselves out of the league? The possibility remains that as players go from 28 to 29 that they see more time on the first line, removing them from the second liners chart here; as the major dropoff for first liners does not occur until age 30.
Our worst performing age for second liners comes in the form of 19 year-olds operating at 1.53 p/60.
I know what you're thinking. Who is that 42 year-old giving all other 42 year-olds such a bad name? That would be Gary Roberts' 08-09 season where he played 30 games for the Lightning recording only 3 points at 5v5.
Third-liners come in with the lowest average age of just 26.2 years-old. This gives fuel to the fire that a lot of coaches like to bring along their younger players with third line responsibilities. Of the 1204 players in our data set for third-liners, 94 of them were aged 20 or younger, the highest frequency on any line.
Somewhat incredibly, third-liners seem to peak at age 23 and see a steady decline in frequency as you move up in age. Again, pointing to the third-line as being a development curve for young players. Players on the 2nd line also seem to peak at age 23, but the dropoff isn't near as consistent or steady as it is here on the third-line.
Looks like the top young players make the jump into their team's top-6 at age 23/24, and don't look back.
Who's that 43 year-old? Lucky for Gary Roberts, he hung up his skates after the 08-09 season, so we know it isn't him. The unlucky player here is Claude Lemieux's 18 games with the San Jose Sharks in the 08-09 season. One point on the season will do that to you.
Looking into the charts, it appears that if you are a great talent, have a healthy career for longevity purposes, there's no reason that regression will occur. However, it is quite rare as just 6.6% of forwards who met the criteria (10 games played in any season post-lockout) are 35 or older. With the addition of the 35+ contract rules to the CBA, this is a number that is surely to decrease as teams become more hesitant to hand out contracts to the 'aging vet'.
Points-per-60 by age group doesn't really take a hit (except in a few extreme cases) because the players still playing at that point or only playing because they still can.
The meat of NHL players seem to fall within ages 21-29, with a nearly normal distribution on the frequency, peaking at age 25. Which may lead you to argue that a player doesn't hit their prime at age 27, but may in fact already be on the downswing of his career. Prime seems to be more from the ages of 23-26.
If you're very young and in the NHL, it's probably because you are very, very good. If you're very old and in the NHL, it's probably because you are still very, very good. If you're anywhere in between, well, it's a crapshoot if you pick a player name out of a hat whether they're a good player or not - and chances are, that player won't get much better as time goes on.