clock menu more-arrow no yes mobile

Filed under:

On Sample Size, Mortal Enemy of the New York Islanders

Recency bias is one hell of a drug.

Bruce Bennett/Getty Images

Hockey is the most what-have-you-done-for-me-lately of the four major sports. We have nothing scientific to back up this assumption, but we're pretty comfortable stating it as fact in the wake of the Islanders' recent subpar stretch of games. Twitter being the echo chamber that it is, and all.

Perhaps recent events thoroughly dominate a hockey fan's thoughts—more so than a basketball fan's or a football fan's—because of the level at which randomness impacts the on-ice action.

We've all seen games where one team is dominating zone time, outshooting their opponent, getting all manner of scoring chances, but ends up on the wrong side of the result after a deflection off a defenseman's skate begets a tough-luck 1-0 loss.

Maybe recency plagues hockey fans because of how much randomness impacts on-ice action

The random, shifting, follow-the-bouncing-puck nature of hockey as opposed to, say, baseball, where everything comes to a standstill before each pitch is delivered, is what often causes hockey fans to base their perceptions of a team on what they last saw rather than on the underlying numbers.

Especially if those numbers stand in apparent contrast to what was last seen. Psychology is weird, man. As is hockey.

Because hockey is played a) on ice and b) by players on skates and c) at very high speeds, random event fluctuation is more pronounced and tends to affect our thinking in a major way when it comes to what we've most recently witnessed.

This recency bias also gives rise to beliefs like "wins are the direct results of one team ‘wanting it more' than the other team," because it's easy to associate losses with a lack of effort. Just pick a couple bad plays and boom: "team didn't try hard tonight. That's why they lost."

(An aside: if all it took to win hockey games was "wanting it more than the other team," things like skating, passing, shooting, and overall talent wouldn't be important to coaches and general managers. But, those things are important, meaning a team has to do more than just "want it more" to win games.)

Recency bias is dangerous when it comes to assessing hockey clubs because it can obscure the underlying numbers that have strong (and proven) predictive power in quantifying how well a team is truly playing, and what results can generally be expected in the future.

By succumbing to recency bias and avoiding stats entirely, we're essentially putting all our stock in a small sample size of recently played games and therefore disregarding the most important measures used to determine whether a team is playing poorly, or if it's simply running cold.

So, you were talking about the Islanders before...?

Right, about those guys. The Islanders have been one of the league's best teams all season long, which isn't a platitude. It's a fact.

They've been at or near the top of the Metropolitan Division standings and, not coincidentally, at or near the top of three key advanced stats categories for 75 games (and counting): score-adjusted Corsi (SaC), score-adjusted Fenwick (SaF), and score-adjusted Shots on Goal (SaSOG).

(For those wondering why score-adjusted metrics are so important: Eric T. over at Broad Street Hockey summarized why adjusting for score effects is the best way to get a comprehensive picture of a team's puck-possession ability.)

Basically, consistently recording more shot attempts than your opponents has been proven to have a strong correlation to winning games and, in turn, to sustained postseason success. The fact that the Isles have posted elite possession numbers across an entire season's worth of games is not something that should be taken lightly.

But that's exactly what's been happening in Isles comment sections over the past week. Because apparently a seven-game sample (Mar. 10-26) holds more weight than the preceding 68-game subset.

"But they need to be playing their best hockey at this time of year!" Again, that's recency bias at work. Which generally isn't the principle on which you want to build your analytical case.

Admittedly, the Islanders haven't gotten many wins over the past two weeks—which is ultimately what matters in this (or any) sport, as we've been needlessly reminded—and it's brought about an interesting/terrible question: are the 2014-15 New York Islanders really any good?

If we accept that a team is what the numbers say it is, then we can confidently state that the Islanders are a good team. One of the league's best, in fact. Record-wise and possession-wise. And it's more likely that they return to form (i.e. their stellar season-long numbers) rather than they suddenly start playing like the Sabres.

(Sorry, Buffalo. At least you get that McDavid kid.)

Even though the wins haven't come as regularly in recent weeks as we've grown accustomed to seeing this season, the Isles are still out-attempting their opponents more often than not; this is important to note because it indicates the Isles are simply running cold, and not that they've fallen off a cliff from which they won't recover.

Since the All Star break—the period of time which many people will point to in terms of the Isles' middling record—the club still ranks third in the league in SaC (53.3%), second in SaF (54.1%), and second in SaSOG (54.7%). They're controlling the puck at a higher rate than their opponents, which is a sign that the bounces have gone against them, and that wins are probably on the way.

If they were losing games and losing the possession battles night after night after night, it'd be time to worry. As it stands, the Isles are just going through a rough patch, as all teams do in an 82-game season. The timing of the Isles' particular rough patch is what's making it seem worse than it really is.

As another manner of consolation to the ledge-jumpers out there: the Isles' PDO is a laughably low 96.4 (96.4!) in March. They've only posted plus-100 PDOs in two of their 11 games this month. In short, they'll bounce back sooner rather than later since PDO naturally regresses to 100 over time.

No team can run that cold for that long.

But I thought end-of-season form was the most important thing ever!

Funny how that works. Not to beat the concept of recency bias to death—OK to totally beat the concept of recency bias to death—picking a small sample size of games in which a team has gotten unlucky and assigning it the "they're not winning because they're not trying hard" descriptor isn't a good method by which we can evaluate that team's potential for future (read: postseason) success.

Especially if that sample size comes near the end of the season. Here, we'll prove it.

To those who believe an 11-game stretch in the year's third calendar month is the be-all, end-all of a team's season, we present a sampling of 10 NHL clubs and their records over 11-game stretches during the month of March:


































What do all of the above teams have in common? Only that they're the last 10 teams to win the Stanley Cup: (A) Los Angeles '14,  (B) Chicago '13,  (C) Los Angeles '12, (D) Boston '11, (E) Chicago '10, (F) Pittsburgh '09, (G) Detroit '08, (H) Anaheim '07, (I) Carolina '06, and (J) Tampa Bay '04.

What don't they have in common? Uniformly fantastic 11-game records in March of their Cup-campaign years. Does a team absolutely have to be playing its best hockey at this time of year in order to win in the playoffs? It helps, but it's by no means a prerequisite.

We're not saying this means that Islanders will win the Cup—even at 3-5-3 so far in March, they're certainly on the list of contenders—just that strong possession numbers tend to outweigh short stretches of bad luck.

To make a cross-sport parallel: it's the same concept as when a 90-percent free-throw shooter misses four consecutive shots from the line; his average says he'll probably make the next 36 free throws in a row.

Knowledge check

There is no stat for ‘heart' or ‘grit' or ‘wants it more,' which makes it very convenient (easy?) for the eye-test crowd to attribute wins to MOAR HEART and losses to NOT ENUFF EFFORT. There's no statistical way of disproving those statements.

At the same time, having strong fancy stats does not mean automatic wins. No one has ever said that, nor will they ever. Advanced metrics are just a method by which one can quantify how well a team possesses the puck.

Corsis and Fenwicks are not points and wins; they are simply proxies for possession and they do not translate directly to points in the standings. (We cannot stress this enough.)

They do, however, provide insight to a team's current form (control of play, bounces for and against, etc.) and have a strong correlation to future performance: teams above 50 percent in the possession department tend to win more often, just as teams below the 50-percent mark tend to lose more often.

As we've discussed above, and have seen play out year after year, score-adjusted measures are even more predictive. The Islanders are good by both raw and score-adjusted measures; they've simply been unlucky (for lack of a better term) of late.

They're not getting blown out and they're certainly not consistently losing the possession battles; don't let recent results trick you into thinking the Isles are stick-a-fork-in-them done.

But, if you still believe that the only games to really matter during the course of a season are the ones that have happened these past two weeks, we're going to assume that you weren't celebrating the Islanders' successes in October (6-4-0, 12 pts), or November (11-3-0, 22 pts), or December (8-4-1, 17 pts), or January (7-5-0, 14 pts), or February (9-5-1, 19pts).

Because hey: those games didn't count. Right?

(All manner of stats not easily calculated via the counting-on-our-fingers-and-toes method are based on even-strength 5-on-5 play and are courtesy of and They are the best sites.)