The summer of 2010. Don’t you remember it like it happened yesterday? Ethan Moreau was waived, Sheldon Souray was sent to the minors, Patrick O’Sullivan revealed himself to be engaging in intergalactic espionage… ah yes, heady days indeed. But at that moment, new hope was burbling up through the sticky, putrefying remains of the 09-10 Oiler season. The kids. The three Amigos. Taylor Hall, Jordan Eberle, and Magnus Paajarvi.

Tolkien referred to this concept as “eucatastrophe” — when a light of deliverance breaks through the darkness of despair. Yes, it brightened up the team’s prospects all the way to another 30th place finish, while the kids were certainly young, exciting, and young.

Going into the shortened 2013 NHL season, it seems much of the Oilers’ plan to improve the team’s fortunes sits on the shoulders of all the sparkly young assets they’ve accumulated over the last few years. This post will try to forecast what kind of seasons Oilers fans should expect from Hall, Eberle, and MPS, the venerable third-year Oilers who very incrementally delivered us from despair 2.5 short years ago.

Now, I’m an economist, a stats/data kind of guy. I generally approve of all the advanced metrics the hockey blogosphere has created over the last half decade, and find their insights useful. In this post, however, I’ll be sticking to the general arithmetic of traditional statistics — namely, points per game.

**Experimental Design**

To start, I compiled a list of all wingers who had rookie years of more than 20 games in the NHL after the first lockout (1995) between the ages of 18 and 21. Of course, all three Oiler 3rd years fit into that category. Then I followed suit with a list of wingers who had their second years (>20 games) between the ages of 19 and 22, and finally a list of wingers who had third years between the ages of 20 and 23. What I was trying to identify was the population of wingers who had three consecutive years of more than 20 games in the NHL at a young age. It turns out that 71 players match this set of facts.

Now, many bloggers would at this point parcel out the comparable players to try to tell a story of where our 3 protagonists may end up if they follow their comparables’ paths. But as a stats guy, I question why we don’t keep this entire population in our analysis. Why voluntarily restrict your sample size when there’s perfectly good insights to be gleaned from the first three years of Jared Boll’s career as there is from Pat Kane’s? We can learn something from all 71 of these players.

My question was as follows: what 3rd year points per game level can we predict for a certain winger by using their points per game totals from their first and second seasons? I set up a multiple regression model, with 3rd year points-per-game (PPG) as the dependent variable and 1st & 2nd year PPG as the two independent variables using the population of 71 players who met my criteria above. I specified zero as the intercept to fit the logic of the model.

The model came up with an r-square of 0.9265, meaning that the equation it came up with could explain 92.7% of the variance with respect to 3rd year PPG with the variance seen in 1st and 2nd year PPG. Not too shabby. The equation is as follows:

3rd year PPG forecast = (1st year PPG * 0.498) + (2nd year PPG * 0.639)

For instance, here’s the equation for Jarome Iginla: 3rd year PPG forecast = (0.61 * 0.498) + (0.46 * 0.639) = 0.60. This model predicts that in Jarome’s 3rd season he’d put up 0.60 PPG — fittingly enough, he put up a mark of 0.62 PPG.

Both of the coefficients above have generous enough t-Stats (4.00 and 6.24, respectively) to reject the hypothesis that 1st and 2nd year PPG have no explanatory power with respect to 3rd year PPG.

**Predicting the Three Amigos & Friends**

Using this equation, we can then estimate what the PPG will be for Hall, Eberle, and MPS, along with a handful of other 3rd year NHLers entering the 2013 season under similar parameters.

The model suggests Taylor Hall will score 42 points in a 48 game season, Eberle will score 45 points in a 48 game season, and MPS would score 16 points in a 48 game season. Of course, this doesn’t take into account advanced stat regressions of any kind — essentially the model assumes everything else is held constant, the only thing changing here is being one year older. If I was smarter or had more time I’d try to build those mechanisms into an expanded model. For now, this is just something fun to look at.

So, I guess the more important question is does the model suggest the 3rd year players will progress enough to help carry the team to the playoffs? I’d suggest, not really. Taylor Hall is predicted to have only a slight 0.01 PPG increase in output — though over a full schedule his impact will be non-trivially larger. Eberle is predicted to drop from 0.97 PPG to 0.93 PPG, which jives a bit with others’ views that his shooting percentage is unsustainable. Make no mistake, he is an impact player, but comparing his results to his peers from the past suggests he has a bit of moderating to do. MPS is predicted to rise from 0.2 PPG in his sophomore season to 0.34 in his trophomore season. Over an 82 game schedule that would translate into 28 points, meaning he’d be up to his elbows trying to outscore Ryan Jones (33 pts last year). I remain convinced he has 3rd line potential, and these numbers suggest that line of thinking is appropriate.

Here’s the full list of 71 players from the past who met my criteria for inclusion in this study:

## 2 Comments

Excuse my dullness, but where do the coefficients come from?

Good work on the post, by the way!

The coefficients are created through multiple regression — they’re the constants that fit a line that has the least squared errors relative to the dependent variable… I hope that answers your question…