Rss

NHL teams are prone to bargain bin hunting for hidden gems. The Canucks themselves have adopted the reclamation project approach many times over Jim Benning’s tenure, looking for bargains with the likes of Derrick Pouliot, Sven Baertschi, and now, Matthew Highmore.

The reclamation project is a controversial move. When a player is available for cheap, it’s often for a good reason: the player just isn’t that good, and acquiring him doesn’t change that fact.

However, there is actually a route to a good reclamation project. It starts with a concept I talk about a lot here on the blog: the law of mean reversion.

The law of mean reversion says that outcomes deviating heavily from average will revert to average over time. A player scoring at an unsustainably high rate will eventually see his production tip to more normal levels, while a player whose scoring has run dry is likely to see better production going forward.

So why hasn’t Connor McDavid experienced mean reversion yet? Mean reversion is dependent on the degree to which an outcome is fuelled by luck. I’ll direct your attention to a couple of videos (again) featuring the great Michael Mauboussin on luck and skill. The first is on playoff performance, and the second is a more broad lecture on luck and skill:

As you can probably guess, a savvy GM isn’t grabbing bad players for cheap rates, rather, he’s looking for players that have played well but gotten unlucky. He can happily walk in and take the “problem” player off the selling GM’s hands for a well below-average cost and just watch the player improve with more usage.

To observe this effect in action, I’m going to do a study. We need two things:

  1. A measure of player quality with the luck stripped out.
  2. A measure of the difference between expected output and actual output. The greater the difference, the more unlucky the player, and the greater the possibility that there’s unseen value.

I chose EvolvingWild’s relTM metric for the task. RelTM effectively isolates for team quality to grab a player’s impact on either goal differential or expected goal differential. The first is close to a measure of “pure skill” while the second has degrees of luck.

I took every NHL forward’s season over the last decade with a minimum of 300 minutes on ice, and I subtracted relative actual production from relative expected production to get my “value factor”. Next, I filtered out the players who had negative relative expected goal differential, and positive actual on-ice results, eliminating the bad players and players who aren’t unlucky enough to look bad.

Here’s how both scores look:

I used a ranked methodology to put the factors on even footing, and took the average of the two ranks to compile a list of the “deepest value” forwards of the last decade.

That’s a lot, so to summarize: we have quality, defined team-relative (relTM) expected goal differential, and we have value, defined as the difference between actions and output. Both factors are evenly weighted, with players ranked on their average of both. To my pleasant surprise, Cody McCormick was ranked at the top of the list (dammit).

After that gaffe, the rest of the rankings were pretty sound. In the least scientific portion of this study, I took the top 100 reclamations to use as guinea pigs. For your interest, here are the top ten names and seasons:

I’d like to remind you: the idea of this list isn’t to rank the “best” players. Rather it’s to determine who might be the biggest “bang for your buck” in terms of perception vs reality. Evgeni Malkin, though never available via trade, was one of the best “value bets” in that sense — he was a great player with outputs that didn’t reflect that fact. In 2011, he scored at a measly (for him) pace of 37 pts in 43 games, only to rebound in 2012 with a 107 point Hart trophy season.

One of my favourite parts of this study was searching hockey twitter for opinions on various list names. For example, here’s Jordan Weal:

Don’t worry Corey, you had a good reason to believe.

Nick Shore was exceptionally underrated:

Martinook for Selke:

Of my 100 top reclamation opportunities of the last decade, I decided to isolate a couple of things. How did production change? Which reclamations actually got traded for their misfortune?

On average, the list’s reclamation projects increased their points per 60 by 0.31 against a relatively flat change in the broad population. Keep in mind that this is partially skewed by top players being unable to really increase production, but those do represent a minority of all players.

Being on the reclamation list was also predictive of switching teams the following season. Just over 20% of all reclamation players got shipped out (again, statistically significant), and some were driven out of the league entirely. Keep in mind, the study was limited to 300 minute player seasons, and there were more than a few reclamations traded multiple times with short looks following their unlucky season.

But, the switchers must have improved more slowly than the league right? No, in fact, they improved more than the players that stayed put (not by a statistically significant margin) and than the broader population (statistically significant).

 

There were a lot of crazy stories in the reclamations. Ever heard of Tim Stapleton? Jets fans have, as he put up just under 0.5 points a game in the year following his unlucky season in Atlanta only to get rewarded with the boot to Russia.

David Perron is another funny case. He was on the reclamation list in 2015, and the next season Pittsburgh shipped him out at the trade deadline. The change of scenery worked: an expansion draft claim and two sixty point seasons later, Perron is no longer on the reclamation list.

So now, I’ve invented a new team: team reclamation. While its construction is dependent on hindsight, team reclamation is every player on the reclamation list that played 300 minutes on a new team in the following season (so Perron doesn’t make it).

Here’s the new team and the players’ total change in points per 60 over the previous year:

And here is the team with each player’s total points in the following season, sorted by per-game point totals:

While I previously put Sham Sharron up against the Canucks amateur scouting department, I’m now inclined to have pro-scouting departments of various teams in the NHL take on “team reclamation” and its assistant captain Loui Eriksson (for what it’s worth, Loui’s failure to improve was a drag on reclamation’s performance).

Expect some writeups here on CanucksArmy on the art of the reclamation project in the future. While the reclamation list is imperfect, it gives me plenty of opportunities to try and refine the search for cheap players, and may yield value in the future.

Conclusion: deep value forever

To summarize the key findings in the study of the reclamations:

  1. The reclamations’ increased production in the following season at significantly higher rates than the overall population.
  2. Despite their production increases, the reclamations were more prone to switching teams than the rest of the NHL’s forwards.
  3. The reclamations that switched teams improved at the same rate as the other reclamations, and more significantly than the rest of the NHL.
  4. Among those reclamations, two “switchers”, Jonathan Marchessault and Carter Verhaeghe, each signed with Florida for below $1 million a year. One of my favourites, Jussi Jokinen, went for a 7th round pick and produced 57 pts in 81 games the following year. Columbus acquired Marian Gaborik, watched him produce, and then shipped him out to Los Angeles.
  5. A number of the reclamations were driven out of the league the following season or sent to the AHL.

I’m not suggesting to pay through the nose for reclamations — in fact, I’m suggesting the opposite. The point here is to figure out which players might be unjustifiably cheap at a given point in time. The Florida Panthers have proven multiple times that reclamation projects can work, you just have to bet on some better luck.


Important
This site makes use of cookies which may contain tracking information about visitors. By continuing to browse this site you agree to our use of cookies.