Youth Matters, but how much?

There has been a lot of heated discussion going on in Tbird land these days and none may be more heated than the argument that this team is losing "because they are the youngest team in the league" or because of other factors.

I think we can all agree, at least a certain extent, that younger teams tend to be at the bottom of the league and older teams tend to be near the top of the league. This isn't to say that there aren't exceptions to that generalization. Prince Albert is the 3rd oldest team in the league and they currently sit in 9th place in the Eastern Conference and may miss the playoffs. On the other side of the equation we have Kamloops as the 2nd youngest team with 60 points, Vancouver, Portland and Kootenay with the 5th, 6th and 7th youngest teams in the league and 76, 77 and 79 points respectively.

The question is... how large is the impact of age? How much can we predict the success of a team based on the average age of the roster?

What I've attempted to do here is look at these numbers as objectively as possible. I've made no secret of the fact that I think Sumner is a decent coach and that Farwell is a decent General Manager. I've also never said I thought either was great, only that I thought neither should be fired (yes I know Farwell can't really be fired).

I've tried here to let the numbers tell me the story, rather than trying to mold the numbers around what I hope or want them to say. I think you can make some conclusions that support arguments on both sides of the fence (believe it or not).

So without further delay... here we go.

The first thing I did was to take the average age of each team as calculated by Alan Caldwell over at Small Things At Large. Caldwell is a numbers master and provides some of the best WHL content out there. If you're reading this and you actually haven't given his place a visit, do so.

So the study isn't perfect because we don't have the exact age of each team on the last day of the season, but it still gives us a pretty good idea of what each team looked like when we get their ages after the trade deadline. Keeping in mind the fact that if a team has a bad start to the season they are more likely to trade veteran players and if they have a good start they are likly to trade youth for veterans and that will affect the numbers.

We then chart each team's average age against the number of points they had at the end of the season (or currently for the 2009-2010 season). We then calculate a Linear Regression trendline.

Quick sidebar... Linear Regression is way to model the relationship between two variables. If "Age" is our "x" data point and "Points" are the "y" data point, the Linear Regression is a way of modeling how "y" is affected by "x" or... exactly what we are looking for... how is a team's performance affected by the age of the team.

Once we chart this Linear Regression, the "line" can be described by a formula and something called an R squared.

(Stay with me... I'm trying not to make this TOO complicated... I will explain what it all means in a minute)

R Squared is a number from 0 to 1 that represents the "strength" of the correlation between the "x" variable and the "y" variable OR how strong the connection is between "Age" and "Points". Everyone good with this?

Ok let's pound the numbers.

I did 4 charts (I will provide charts later, too hard to format in this post). First chart is Age against Points from two seasons ago (2007-2008).

Oldest team: Edmonton - 18.227 years old, points 55*
Youngest team: Prince George - 17.682 years old, points 44

Best team (regular season): Tri-City - 17.913 years old, 108 points
Worst team: Red Deer - 17.783 years old, 43 points

*Important to note here that Edmonton was an expansion team and was allowed to carry 5 O/A players and was a pretty big reason why they finished with 55 points. I could make a case that they should be taken out of the data but I will leave them in to keep things simple.

R Squared = 18.49%
Slope = 61.009

18.49% doesn't seem very high and if you take out Edmonton it gives you 28.72% with a slope of 77.584.

At this point... I'm sure I've either completely lost you or you're asking "ok Tyler, what the heck does all this mean".

Here goes.

I enlisted the help of a friend who is a math teacher and asked him where and when does our R Squared become statistically relevant. Or said a more simple way, at what level of R Squared can we say with some confidence that the relationship between Age and Points is significant. Allowing for a 5% level of significance (meaning we are ok with being wrong 5% of the time) the answer is 16.3%. Anything less than that and the result really isn't strong enough, anything over that and we are "on to something".

Even if we leave Edmonton in the data, the conclusion is that Age and Points are statistically related to a meaningful extent and if we take Edmonton out and the R Squared jumps to 28.72%, suddenly Age becomes a huge factor.

Now let's tackle the Slope. The slope of the line is 61.009. What does this mean?? A slope of 61 means that for every 1 unit of "x" variable you will increase "y" by 61. Whoa... wait a second.

This means that if a WHL team increased their age by 1 year in the 2007-08 season they could have reasonably and predictably expected an increase of 61 points in performance. Wow... really? Yes... really. If Prince George increased the average age of their players from 17.682 to 18.682 the next season they could expect to jump from a 44 point team to a 105 point team.

"That doesn't seem right..." is I'm sure what some or most of you are saying and you would be kind of right. The problem with that is that the oldest team in the league was Edmonton at 18.227. It's pretty unreasonable for us to think that in one season P.G. is going to go from the youngest team in the league to suddenly being .4 years older than the oldest team from the year before.

So let's look more practically at 6 months (or .5 years). If a team gains just 6 months in average age they can still expect a boost in performance of 30.5 points. That's still pretty large! Saskatoon, with the 2nd youngest team that season at 17.783 would go up in age by 6 months to 18.283 and improve somewhere around 30.5 points... going from a 67 point club to a 97.5 point club.

A 6 month jump in age is still significant, but much more reasonable. Medicine Hat went from 18.00 in 2007-2008 to 18.38 the next year. Brandon went from 18.08 last season to 18.50 this season with the hosting of the Memorial Cup. Even with the more reasonable expectation that a team could get older by 6 months, the results are the same. A team that gets older, gets better. Even without considering skill increases, coaching or player movement.

Is one season enough to prove anything? No... so let's look at the last two seasons.

R Squared:

2008-2009 - 36.85%
2009-2010 - 37.87%

Cumulative (all 3 seasons) - 23.96%

Again... Anything above 16.3% is statistically significant. So the correlation between Age and Points is extremely strong and significant. Now let's look at things in real terms.

Slope:

2008-2009 - 71.525!
2009-2010 - 47.386

Cumulative - 55.34

Over a 3 year period, if a team got older by 6 months (in terms of average) they gained over 27.67 points in the standings. That blows me away.

The other nice part about getting a slope and a formula is that we can "predict" how many points a team of a certain age "should" have had. This allows us to examine whether this Seattle team (or any team) is playing above or below their age... or said another way... whether they are overachieving or underachieving and by how much.

I won't run through each team in the WHL (but you could and I might eventually). Let's just focus on Seattle.

In 2007-2008, the Tbirds had an average age of 17.905. Using the formula, Seattle should have expected roughly 72.466 points. They earned 91. You can reasonably say that Seattle out played their age by nearly 20 points.

In 2008-2009, the Tbirds had an average age of 18.04. Using the formula from that season, Seattle should have expected roughly 82.911 points. They earned 75. Seattle under played their age by almost 8 points.

This season, the Tbirds have an average age of 17.57. Using the formula from this season, Seattle should currently have 43.672 points. They have earned 38. Seattle has under played their age by around 5.5 points.

CONCLUSION -

I'm hoping everyone made it through that.

Here is the point: Some Seattle fans have taken to the argument that "Youth is not an excuse for this team to lose this many games." The fact is that this team is performing close to what we all should have expected for a team of this age.

Age is a pretty strong way of predicting the success of a WHL team. Are there exceptions? Yes. Each season there are teams that overachieve and teams that underachieve their age. In the midst of a 15 game losing streak it's easy to feel like this team is playing way below the level with which they "should" be playing. The facts just don't point to that being the case. Are they underachieving? Yes.. but only by about 5.5 points or roughly 3 wins. For the people that think Sumner should have done more with the talent we had last year? You might have a point. Seattle should have had 8 more points (4 more wins) last season. They still would have finished with 83 points and Spokane still would have had home-ice advantage. They also managed to overachieve their age the season before, finishing with 91 points on a team that should be "expected" to only finish with around 72.

Later this week... we'll take a look at the franchises that have over and underachieved their ages the most over the last 3 seasons.