Advanced Stats Helping Explain The Way Things Are

So a couple of days ago if you read the From The Branches, or if you keep tabs on what Gabe Desjardins is up to at BehindTheNet.ca (or Arctic Ice Hockey), you may have noticed that the Fenwick Percentages for the 2011-12 season were added to the mix of advanced stats that are maintained on his site.

This is great news for those of you that are interested in explanations of what they've seen so far this young NHL season.  People like numbers when they can summarize things quickly.  It helps when they know what the heck they mean, but being able to compare teams quickly and accurately in a fashion that's meaningful is really what makes things in sports stats worthwhile (in my opinion).

We've obviously got the point system for that as far as the standings go, but the problem with using points and standings is that we don't have an easy way to account for strength of schedule, competition, good or bad refereeing, luck, etc.  It's hard to figure out half way though the year which teams are good, and which teams are just lucky.

Knowing this, we look for ways to compare teams.  Well, we can compare how well their goalies are playing, but that doesn't mean much for the skaters.  We can tell who is putting up more goals, and letting more go in, but again, we know in the short term that a lot of that variation is largely due to luck.  So what can we follow that is repeatable, consistent, and will have an impact over the long term?

In sports like football (all versions of it - Canadian, American, European, Rugby, etc.) the key stat for most astute observers would be possession.  Teams that control the ball tend to control the play, and thus are usually more likely to win.  They don't ALWAYS win, but they tend to.  Unfortunately, the NHL doesn't keep or produce regular possession statistics.  Even if they did, we'd be relying on score keepers to track things like giveaways, takeaways and the like, and there's issues with that - generally we'd like to avoid those as much as possible.

So what proxies do we have for possession in hockey? Well, generally the team that has the puck more in the oppositions end, is likely to have more opportunities to attempt to score... that is to say, fire shots in the direction of their opposition's net.  Not all shots will make it to their target.  Many miss, and some are blocked and the majority are actually stopped by the goalie.  But what we can do is compare the number of attempts for each team on the ice, and use this comparison as an effort to determine who is controlling play more frequently.

Two numbers exist for this purpose.  The first, devised by Buffalo goalie coach Jim Corsi and given the same name (Corsi Number) is simply a plus / minus statistic that adds up all of the shot attempts for a team (SOG, MSF, and BSA) and subtracts all the shot attempts they surrender (SOG, MSA, BSF).  It's simple, and easy to observe, but the point has been raised that blocked shots are rarely a favourable outcome of any sort, and including them may or may not cloud the picture.

As a result of this argument, Calgary Flames blogger Matt Fenwick proposed the removal of blocked shots from the assessment. The resulting value which is the plus / minus of shot attempts (those on target and those that miss) for the team and against them, is termed the Fenwick number. A Fenwick Event is considered any shot attempt that hits the net or misses it, but is not blocked.

For the sake of easy comparison, Fenwick Events can be used in a number form (like Corsi) and provided on a per 60 minute basis, or as has been more recently suggested, presented as a percentage.  If you assign all Fenwick Events to the home and away team, each team will have a percentage of the total number of events in any given game.  Similarly, if you add up all the Fenwick Events in favour of a given team over the course of their season, and divide them by the total number of events in the games they play in, you would determine a Fenwick Events Percentage on a seasonal basis.  This is something you can easily compare; it's ONE number.

Ok so now we know a decent number that we can use as a reasonable proxy for possession, but there are some other issues to take into consideration. 1) Score effects: when the a team trails it tends to work harder to play catch up vs. the opposition, and when they are in the lead they tend to sit back and protect the lead.  This skews Fenwick and Corsi numbers quite drastically.  2) Scorers in different stadiums tend to award Fenwick Events at different rates.  Shots and Missed Shots vary quite a bit from stadium to stadium, and this inconsistency is difficult to deal with, particularly when a team's home scorer skews things drastically one way or another.

In an effort to compensate for these two problems by looking at the Fenwick Event Percentages when the score is tied (nobody is playing catch up or sitting back), and when a team plays on the road (their home scorer's bias won't show up in the data).

So here's a number we can put some faith in over the longer haul of a season to compare teams with: Road Fenwick Percentage When Tied.  So how does that shake out so far in the season? Here's a look at things as of November 23rd (I don't believe the data will include tonight's games though).  I've also included PDO to give you a quick read on the type of "puck luck" the team has had at even strength so far this year (remember that teams tend to regress towards scores of 1000 as the season progresses).

So what does this tell us? Well I think it's clear that Minnesota probably isn't the best team in the NHL, despite being 1st overall right now. Their goaltending has carried them so far, and the likelihood of it lasting all year is far from high. The same could be said of Boston, whose recent 10 game winning streak has been piled upon a pile of backup goalies subbing for injured starters, and teams in the midst of scoring problems. It also helps that they have Tim Thomas and Rask in goal, but I think by now we can admit that for them .930+ goaltending is "sustainable".

At the other end of the spectrum we have Colorado getting seriously jobbed right now by some horrid luck. Largely this stems from their crappy shooting percentage. They get a lot of shots on goal, and eventually they'll start going in. Similarly, though to a lesser extent, this could be said for Ottawa, New Jersey, Columbus, and Vancouver. Unfortunately for Devils' fans I don't know if they should expect serious offense or drastically improved goaltending anytime soon though.

Ottawa's goaltending carousel isn't settled, Columbus needs to ride out this stormy early part of the season, and Vancouver is hoping Luongo can sort his game out sooner rather than later.

All of the teams with Road Fenwick % While Tied over 50%, that have a PDO close to 1000 are sitting fairly pretty as their season isn't likely to shift suddenly, and they're playing well over the first 20ish games. This bodes well for Chicago, Pittsburgh, Washington, and to a lesser extent Toronto.  The first 3 are relatively legit contenders whose records are unlikely to fall out of playoff contention.

Toronto's numbers indicate that their early season burst isn't really a mirage.  They're not outperforming anything in particular.  Yes they're scoring a lot but their goal tending has been bad, and odds are when one gets worse, the other will improve.  Their is room to improve on the Fenwick front, but the team does look close to turning a corner towards being competitive.

Unfortunate reality needs to set in for Calgary and Tampa Bay.  Neither team is particularly productive in terms of controlling play offensively and they give up far too much.  Their PDO values also indicate that they aren't likely to change much going forward.  These are teams that likely need to make trades if they hope to have a shot at the post season.

Let me know if you have any questions and feel free to debate the worth or value of the stat - I know some of you want to.