NHL analytics: The next level of numbers

Rick Nash (Getty Images)

If you want to succeed in a hockey pool, you need to find players who put up more points than your competition expects. One way to go about that is to take advantage of the fact that people have a tendency to weight recent performance too heavily and fail to consider whether a player has benefited from unusually good or bad luck in achieving his results.

More NHL on Sportsnet:
Subscribe: Rogers GameCentre Live
Rogers Hometown Hockey | Broadcast Schedule

Hockey analytics has come up with a number of tools that can be employed to help determine just that—and whether a player’s most recent season causes him to be overvalued or undervalued come draft day. We’ll look at these tools in the context of forwards at even strength (5v5).

Catch NHL Special: A Numbers Game, tonight at 7 p.m. ET on Sportsnet

The first of these is a stat known as Individual Points Percentage (IPP). This measures the percentage of 5v5 goals scored when a player is on the ice in which he gets a point. In the long run, most forwards tend to get points on about 70 percent of the goals scored when they’re on the ice. Of forwards who played at least 2,000 5v5 minutes from 2007–08 through 2013–14, 65 percent were credited with a point on between 65 and 75 percent of the 5v5 goals scored when they were on the ice.

In the short term, things are a little bit different. Among forwards, 346 played at least 500 5v5 minutes last season. Only 44 percent earned points on 65 to 75 percent of the goals scored when they were on the ice. In other words, a player coming off a year in which he had an unusually high or low IPP may have posted a point total that doesn’t reflect his likely output in the future.

It should be mentioned that star forwards do tend to get points on a higher percentage of the goals scored when they’re on the ice than scrubs. From 2007–08 through 2013–14, for example, Sidney Crosby was credited with a point on 82 percent of the 5v5 goals scored while he was on the ice. Colton Orr, meanwhile, was credited with a point on just 42 percent. So while most guys do fall into that 65- to 75-percent range, there are special cases, both good and bad.

In addition to IPP, 5v5 point output can be distorted by on-ice shooting percentage—the number of goals a player’s team scores when’s he’s on the ice divided by the number of shots his team takes when he’s on the ice.

Alex Ovechkin, for example, scored 21 5v5 goals on 224 5v5 shots last year. He shot 9.4 percent at 5v5. The Capitals scored 34 goals on 567 5v5 shots with him on the ice—so Ovechkin’s on-ice shooting percentage is just six percent. That is an outrageously low on-ice shooting percentage for a first-line player and a big reason he had such a terrible plus-minus last season.

Fortunately for Ovechkin (and fantasy owners who understand this concept), unusually high or low on-ice shooting percentages don’t tend to persist. If we look again at forwards who played at least 2,000 minutes between 2007–08 and 2013–14, we see that about 64 percent of them had on-ice shooting percentages of 6.7 to 8.9 percent. If we look just at last year, only 55 percent of forwards who played at least 500 minutes fell in that range. So there are outliers there as well.

With both IPP and on-ice shooting percentage, the difference between players in the long run is smaller than the difference between players in the short term. There are seasons in which players happen to get points on a lot of goals and seasons in which they don’t. They can have seasons where everything their team touches goes in and seasons where it just doesn’t, for whatever reason. Those two factors, largely out of their control, can still have a significant impact on the point totals they produce.

To identify players who might be overvalued or undervalued based on last season, it makes sense to look for players who had an IPP or on-ice shooting percentage that was significantly different from their long-term norms. For players who haven’t played 2,000 5v5 minutes in the NHL, I’ve just assumed that they can be expected to get 5v5 points on 70 percent of goals, and that their teams will shoot 7.8 percent with them on the ice.

This produces a list of players for whom offensive production at 5v5 last year may not be an accurate reflection of what we can expect this year. To the extent that people put too much weight on the most recent season, these players might be steals later in the draft or millstones if drafted too high.

As with many models, a little knowledge to go with this straight information can improve outcomes. Despite his presence in the list of potential bargains, for instance, drafting Ville Leino is probably not a good idea, given that it’s hard to put up points in the NHL while not playing there. Better circumstances might also help some players who seem primed for regression based on this list—Gustav Nyquist, for example, is unlikely to generate as many points per 60 minutes of play, but is likely to play more this season.

But for the players on these lists who play in similar circumstances and have similar amounts of ice time to last season, they’re likely to see their point totals rise or fall accordingly. A fantasy owner who is aware of this on draft day is likely to be a fantasy owner who’s happier at the end of the campaign.

When submitting content, please abide by our submission guidelines, and avoid posting profanity, personal attacks or harassment. Should you violate our submissions guidelines, we reserve the right to remove your comments and block your account. Sportsnet reserves the right to close a story’s comment section at any time.