Sunday, December 9, 2012

Short Seasons And Playoff Chances

We were sitting at the starting line on a cool autumn day in 2001, bristling with anticipation. The school board was holding its annual cross country running race for boys in grades 7 and 8. A gruelling 8-kilometre run across undulating terrain littered with 45 degree hills and muddied straightaways, and I’m lined up right next to Jimmy Wurner. Jimmy, standing at six foot one, was far and away the most accomplished distance runner in the school; we suspected that he had hit puberty 2 or 3 years earlier than the average 8th grader, thereby endowing him with superhuman strength and speed - the teenage equivalent of a Kryptonian emigrant.

99 times out of 100 Jimmy “the burner” Wurner would have no doubt mopped the proverbial floor with my sorry excuse of a 14 year old distance running body. But that day, there seemed to be something different in the air. Call it destiny, call it being hopped up on Fruit Loops, but whatever it was, something funky was going on.

For some indefinable reason, I ran my little heart out and ended up finishing 4th in the field of over 100; 3 spots ahead of the burner. Granted, the victory didn’t mean much, leaving me with a blue participation ribbon that would gather dust for years to come in a box at the bottom of my bedroom closet.

What that day reinforced is this; in most sports, if your sample size is small enough, weird and unpredictable things can happen. We tend to see this lot in the NFL - with a 16 game regular season, followed by single game elimination playoff rounds the league presents the smallest sample size of any major sport, lending credence to the popular cliché that on any given Sunday, anyone can win.

If NHL collective bargaining can finally come to a conclusion this month the season will be shortened a great deal, with the excepted number of games between 48 and 60. By looking at probability and statistical noise we can see that a shorter season could result in non traditional playoff teams sneaking into the postseason.

The simplest way to understand the impact of a small sample size is to look at coin flipping. When you flip a coin there is an equal probability of turning up heads or tails. If you were to flip a coin 10,000 times you're likely to end up with close to 5000 occurrences of heads and 5000 of tails. While you're unlikely to see exactly 5000 of either, the results would be close enough to render the difference immaterial.

Compare that finding with what happens when you flip a coin 10 times. The base probability is the same, 50% heads, 50% tails; however, the decreased sample size increases the chance of statistical noise. This is what happens when a sample size is too small, yielding results that are not indicative of the actual probabilities in play. 10 coin flips may result in 7 heads and 3 tails - this of course does not mean the probability of heads is now 70%.

Taking this concept and applying it to the NHL we could say that if the NHL season was long enough, the good teams would make the playoffs and the bad teams wouldn’t. 82 games, while a far cry from 182, is certainly a significant sample size, which allows, more often that not, the cream to rise to the top. However, if less than 82 games are played there is a greater opportunity for this statistical noise to creep in and for weird and unexpected things to happen with the standings.

A 48 game season represents a 42% reduction in the number of regular season games for each team. If you look back at the January and February standings of past years you can see how much playoff standings shift over the course of season. The more games that are played the more likely the league's best teams will separate from its lesser teams. Similar to the coin flipping effect, the more coins you flip the closer you generally get to the likely outcome of a 50/50 split in findings.

The Leafs have finished the last 5 seasons in 26th, 22nd, 29th, 24th, and 24th. While the addition of JVR and rumoured trade for Roberto Luongo could have a significantly positive impact, this is still a team with a recent tradition of missing the playoffs and probably hasn't done enough to shake that trend in an 82 game schedule. In all likelihood, if enough hockey games are played the Leafs would finish outside of the playoffs as they are currently constituted. What a Lockout shortened season could mean for teams like the Leafs is the chance to do the unexpected. A 20 to 25 game streak of games with a significant winning percentage could be all it takes to cement a playoff spot, as opposed to an 82 game season where the ebbs and flows of the season tends to even out the results over time.

While it’s still probable that Leafs still miss the playoffs, with a significantly smaller sample size the potential for statical aberrations is there. Yes, we would probably prefer the team qualifies for the post season on the strength of its burgeoning core and increased organizational depth – but after living through a decade of playoff exile, can we really afford to be that picky?


Anonymous said...

Apart from insinuating that the NHL playoff format actually does a good job of send bad teams packing instead of into the playoffs you've got the right idea.

Darren K said...

The overall playoff structure in the NHL is a separate discussion entirely. There are positives and negatives, especially when compared with other systems put in place in the NFL and MLB that can be at times frustrating.

For the NHL, having the divisions in place, and allowing for an automatic top 3 seed seems counter intuitive to putting the ‘best’ teams in.