Thursday, July 14, 2016

Is 'Good In The Room' A Recursive Problem?

Ken Holland has been largely dormant on the trade and free agency front these last few years - the aftershocks of inking Mike Green and Frans Nielsen are nothing compared to the eruptions that were Brett Hull and Marian Hossa - and so he's gone native, making sure to re-up his own players.  This summer he's re-signed Justin Abdelkader, Darren Helm, and now Luke Glendening to extensions of 7, 5, and 4 years respectively.  

Luke Glendening is a glue guy, a guy you can win with.  Of course his territorial numbers are mediocre at best and his penalty killing skills are marginal at best and he doesn't score, but the Wings are going to pay him $1.8M a season for the next 4 years to play his version of tough hockey.

Googling 'Luke Glendening good in the room' gives me no result, but he has to be 'good in the room' - he's a bottom-liner with limited skills who also plays hard every night and is willing to sacrifice his body.  Why wouldn't he be beloved by his teammates?  

We see this phenomenon leaguewide.  We see it even in front offices who we think are smart, like Chicago.  They re-signed Bryan Bickell and Marcus Kruger to contracts beyond their NHL accomplishments.  Was that to keep a championship team together, or did they truly feel that Kruger is a $3M a year player?  In addition, they recently traded Teuvo Teravainen (with Bickell), claiming that Teravainen's refusal to spend the summer in Chicago is evidence that he's unwilling to make the sacrifices necessary to become a better hockey player.

We know that 'good in the room' 'willing to sell out' and phrases like this are waiting at the front of a hockey player or executive's brain to be spouted to a waiting tape recorder or camera.  We know that the concept of team is relentlessly emphasized.  Playing as one.  We've seen boardroom videos of executives talking about 'the way we play' and how certain players fit into that team concept.  This must filter down to the player level, right?  Coaches must emphasize this sort of thing constantly.  It's a limited sample but HBO's 24/7 on the outdoor game appears to confirm this.

What I wonder is:  We assume that executives value grit and toughness and fitting in with the locker room culture over goals and assists when it comes to certain players.  What if, in addition to that valuation, they know the players have been so indoctrinated about the important of grit and glue guys and whatnot that the executive knows it will have a large psychological impact if e.g. 'Bicker' or 'Glenner' were let go?   We sometimes hear GMs speak this way about certain players, that a team was 'never the same' once a certain player left.  And what are GMs but (mostly) former players themselves?

I know this account doesn't exactly adhere to Occam's Razor.  It's much more likely that GMs merely consider the perceived impact on the team and just think these guys are worth keeping around rather than considering that not signing a certain guy will throw their squad into an existential funk.  I just don't see any GMs who don't consider 'team chemistry' at all when making decisions, and so until that happens we'll never see a team decided presumably on merit rather than a Luke Glendening chancing to be the exact size of the perceived missing piece on the team.

Thursday, May 5, 2016

The Problem with Understanding Hockey Analytics

During last night's telecast of Capitals @ Penguins game 4, everyone's favorite hockey analyst Pierre McGuire couldn't help but make a snarky comment about analytics following Matt Cullen's goal in the 2nd period. For posterity, or until the NHL makes another terrible website design change, the video is embedded below:

"It's the little things that analytics won't tell you, Doc. Winning races, winning battles, getting to loose pucks, all of that stuff happened during that last sequence. What a job by Matt Cullen."
Now, I'm going to tell you something that you didn't expect: Pierre McGuire is right. No metric we have, or most likely are developing (at least until player tracking data becomes available) can measure these things. However, both Pierre, and I suspect, the average hockey fan tend to miss the forest for the trees when it comes to understating what the statistics we do have measure, and how their purpose is applied.

Time and again, we've been over how the best indicator of future success in the standings is some form of Corsi percentage. First it was score-tied, and thanks to a further application of something we knew five years ago, now it's score-adjusted. While Corsi, or anything else we have does not measure Matt Cullen's will to win a race, Tom Wilson's ability to fire up his teammates, or the fact that Ryan Callahan's grandmother is in the stands, it does measure, and until somebody comes up with something better, is the best measure of what Pierre's intangibles hope to achieve.

Put another way, what Pierre and countless others are hoping to do is measure a player's ability to help his team win games. That's the same thing that Corsi does, only Corsi has the advantage of actual, testable events leading to results. At the end of the day, accruing wins and standings points should be the goal of any successful coach, general manager, owner, and franchise. The way that anyone in these positions goes about achieving them often makes or breaks a team's success.

This isn't to say that what Pierre lists as intangibles aren't important, or that a good character guy in the room isn't necessary. The trick is to identify players that possess these traits, and at the same time, drive shot attempts towards the opposing team's net. You can harp on intangibles all you want, but if the result is utilizing a player like Tanner Glass or Zac Rinaldo for an extended period of time, your process is broken, plain and simple.

This brings me to my next point. Many analysts, like Pierre, are often asked to break down matchups on a play-by-play, game-by-game, or night-by-night basis. In a similar respect, the average fan often watches or attends one game at a time, and as a result, both don't stop to see the bigger picture. To me, the fact that we're still dealing with comments like Pierre's on national broadcasts isn't all that surprising. In fact, it's kind of expected, given how he, many of the analysts in his industry, and many fans across the world are viewing hockey under a microscope.

Adding to the problem, hockey may be the hardest sport to predict on a game-by-game basis. Whether you look at basic models or betting markets, one trend is blatantly obvious: in hockey, even controlling for everything we know, a vast majority of the matchups have odds between 55/45 and 65/35 either way. 3:1 favorites in a single game don't happen often, which is why in small samples, we see extreme results.

When you start to ask why, it becomes less and less surprising; hockey is a game of razor thin margins anywhere you look. For example, an above average team is going to control merely 2-4% more shot attempts than a below average one over the course of a season. An above-average goaltender stops one (1) percent more shots than an average one, or one (1) more shot over a sample of 100. If the margins are this thin over 82 games, it's no surprise that they're basically non-existent over one game or one playoff series. Coincidentally, one game or one playoff series are what analysts like Pierre are paid to comment on.

Still, despite the studies identifying score-adjusted metrics as the best predictor of future team success, fans, analysts, and even teams (see: Ducks, Anaheim) still can't rationalize the inherent randomness of losing as a 60% favorite in one game or one series.

To put this in different terms, if you've ever played poker, specifically No Limit Texas Hold 'em, you'll understand this concept. In poker, you might go one day, or even one week having your opponent's King-Queen offsuit beat your Ace-Jack suited whilst all-in during a high leverage tournament hand. However, we know that in the long run you're making the right play, given that you're going to win about 62% of the time. In hockey terms, the 2015-16 Anaheim Ducks were the team holding Ace-Jack. Their season came down to one shuffle of a weighted deck, and they lost. Instead of accepting their fate as a simple mark of bad luck, striving to continue "getting their money in with an edge," they overreacted and fired Bruce Boudreau. As a result, they're almost certain to end up with a less capable bench boss next season.

At the end of the day, the fact that mainstream hockey circles can't comprehend that 35-45% underdogs can win games and series, but don't over the long run, doesn't prove the metrics we have are broken. Instead, it proves that their understanding, and by association the overall understanding of the game, is broken.

Tuesday, May 5, 2015

On The Difficulty Of Starting From Scratch

The Devils yesterday hired Ray Shero to be their general manager, ending Lou Lamoriello's 28 year reign as GM.  There's some parallelism here - Lou, at least according to Wikipedia, named himself GM back in 1987, and has now fired himself.  Would that we all had a career like this.  Regardless, Shero is a curious choice - sure, he's one of the few available GMs with a Stanley Cup ring who is not eligible for Social Security, but given his policies with the Penguins, it's hard to see how he's the right guy to fix a Devils' ship that's run aground in a rapidly shrinking sea.  The Devils need to rebuild a forward corps badly damaged by poor drafting and the ravages of old age - Shero is famous for his trading of young players and draft picks.  He's also (in)famous for inheriting Sidney Crosby and Evgeni Malkin when he took over for Craig Patrick in 2006.  However, let's look closer at the Penguins of that era, because I feel like people miss something when they talk about Shero's good fortune.

The 2003-04 Penguins were truly one of the NHL's worst non-expansion teams.  They finished with a goal differential of -113.  The interesting thing about this club is that they were relatively young - their 9 leading scorers were under 30 - but they were also horrendous.  Of the players that played on that team, three would end up winning a Stanley Cup with the Penguins five years later - Brooks Orpik, Rob Scuderi, and Marc-Andre Fleury.  The Penguins also had several players in their organization at this point who would go on to serve vital roles - they had drafted Colby Armstrong, Ryan Whitney, Erik Christensen, Ryan Malone, and Maxime Talbot.  They would add Alex Goligoski, Malkin, and Tyler Kennedy in 2004.

Here's what needs pointing out though - the Penguins got basically nothing of value for anyone on the 2003-04 roster.  Everyone on this roster either left via free agency or was traded for a pittance.  When Ray Shero took over in 2006, most of this roster was already gone - of the top 9 scorers on the 2003-04 team, only one repeated in 2005-06, Ryan Malone.  Let's detail what happened to those other guys in a list:

Dick Tarnstrom:  Traded in 2006 for Cory Cross and Jani Rita, Cross dealt in 2006 for 4th round pick in 2007, Rita returned to Finland in 2006

Aleksey Morozov:  Returned to Russian League for good in 2004, retired in 2014

Ryan Malone:  Left via free agency in 2008, scored 65 goals in his 3 future seasons with the Penguins

Milan Kraft:  Returned to Czech League for good in 2004, retired in 2013

Rico Fata:  Waived

Konstantin Koltsov:  Returned to Russian League for good in 2006, still active

Richard Jackman:  Traded for Petr Taticek in 2006 - Taticek never played for PIT

Tomas Surovy:  Returned to Europe in 2006, still active

Tom Kostopoulos:  Left via free agency in 2005

There you have it - the Penguins' top 9 scorers from that year turned into a 4th round pick.  And that's often the true price of 'tanking'.  There were assets in the Penguins' system at the time, but this is an entire team that apart from Malone, Orpik, Scuderi, and Fleury (none of these anyone's idea of a franchise player) may as well have never existed.  Shero made some very bold moves as Penguins' GM - the boldest being the add of Marian Hossa and Pascal Dupuis at the expense of Armstrong, Christensen, a former 1st round pick, and their 1st round pick in 2008 - but he was always hamstrung by the fact that he inherited a team with few assets besides the obvious two.  He needed all his young talent to fill out the garbage roster + Crosby + Malkin (coming soon) that he'd been handed, but he needed to augment that roster with players to compete for a Stanley Cup.  It's no wonder that the Penguins now are struggling to find young forwards - that was certainly not ordained, but it is not surprising when you consider the Penguins of 2003-04 and where they wanted to be in the late 2000s.  The legacy of an empty team lasts for quite a long time.

Tuesday, June 10, 2014

Whither (or Wither) Defensive Defensemen? On The Disappearence Of Your Least Favorite Player

Hockey is won in the trenches, everyone knows that.  The 'trenches' in hockey are the tough areas - at the blueline, along the boards, right out in front of the net.  You want a guy - forget that, you want a MAN - who can protect these areas and makes sure that any opposing forward who arrives in them pays a price.  So you get a defensive defenseman, a man who makes life hard on the other team for 19 or 20 minutes a night, a man who'll stop pucks with his face and handles the puck like he's handling it with his face.  Off the glass and out.  Along the way he'll take some penalties, and along the way he'll score a goal or two a season, maybe four or five if it's a lucky year, and his teammates will mob him with joyous and somewhat dubious smiles.

The stats community has come to the conclusion that this sort of defenseman, besides when he is exceptional, usually doesn't help a team.  In fact, if your team has one of these guys and you are reading this blog, he is probably the guy you can't stand.  His errors lead to goals, and his positive contributions are impossible to measure.  His Corsi is usually bad, perhaps the worst on the team, and you think to yourself, 'If we could only get rid of this guy, we'd have a better squad'.  Well, have no fear.  My contention is that the game is aging these sorts of players out.  It's happening slowly.  General managers still love them and sign them for way too many years, but that's going to have to come to an end soon, I think, because I'm not sure where these guys are going to come from in the future.  There's a lot of guys wandering around 3rd pairings with these sorts of skill sets, but I don't see how many of them make the jump.

To test this contention, I used hockey-reference's amazing Season Finder feature to look up these players' careers.  Now, I think it's quite difficult to pinpoint what these players are - it's sort of what Potter Stewart said about pornography, you know it when you see it.  I used 4 criteria for my initial set of players:

A:  More than 500 NHL games played
B:  No greater than .25 points per game for his career
C:  Penalty Minutes .75 per game or above for his career
D:  Must've been a top 4 defenseman for multiple seasons (roughly 19-20 minutes per game)

Through an exhaustive search, I found these defensemen who fit those criteria above:

Bryan Allen, Mike Komisarek, Andrew Ference, Robyn Regehr, Willie Mitchell, Bryce Salvador, Anton Volchenkov, Hal Gill, Brooks Orpik, Barret Jackman, Cory Sarich, Ladislav Smid, Tim Gleason, Rostislav Klesla

These players collectively average .19 points per game (so approximately 16 points per 82 games) and .98 PIMs per season, or 80 penalty minutes.

I generated a second set of players for whom these things are mostly true and added a new criterion that could also be fuzzy, namely E:  Usually loathed by his Corsiati fanbase and generated these players who also fit the profile, although most of them either were too good early on or were always back-pairing D:

Chris Phillips, Doug Murray, Nicklas Grossmann, Matt Greene, Mark Stuart, Scott Hannan, Jan Hejda, Eric Brewer, John Erskine, Andrew Alberts, Brad Stuart

One thing the above list reveals is a staggering number of 1st round picks.  Allen, Komisarek, Regehr, Volchenkov, Orpik, Jackman, Gleason, and Klesla were all taken in the first round.  From our second list, Hannan, Brewer, Stuart, and Phillips were also first round selections.  Was this still the Lindros bogeyman hovering over the league?  If you recall, when Eric Lindros showed up in the league, every team was terrified of him and needed to get enormous defensemen who could handle defense against him and the huge forwards that would surely come after.  And indeed, most of these guys were drafted in the days when obstruction was still a perfectly cromulent form of defense, and cross-checking at the front of the net was expected.   Has that alteration in junior hockey changed the way some of these players might've developed?

I feel that these players fall into three categories - Degraded 'Two-Way' Defensemen, Eternal Defensive D, and The Rest.

First Category:  Degraded 'Two-Way' Defenseman

A player like Brad Stuart started out his career with offense.  He was supposed to be the next Scott Stevens when he was drafted.  He came into the league and put up 36 points in his first season.  He had .55 points per game in the 2005-06 season.  But if you look at Stuart's last five seasons, he's fallen under the .25 points per game threshold.  We see this with Eric Brewer and Chris Phillips too - they were once over the .25 points per game threshold but fell below it in recent seasons.  Paul Mara is another example of a guy whose offense simply died in his late 20s and he became a third pairing physical guy.  I lack the ability to sift through career arcs, but my suspicion would be that a lot of these guys had really good shots and weren't horrible with the puck in the offensive zone and so received PP time early in their careers, but that once their puck skills diminished or it was revealed they were no good at offense, they were taken out of any offensive role.

Players that fit in this category now:  Braydon Coburn, Kyle Quincey, with Dmitri Kulikov and Zach Bogosian being longshots.

Second Category:  Born Defensive D

These are players who were drafted high because they were supposed to be good on defense.  Their offense didn't develop yet.  Maybe it never will.

Players that fit this category now:  Erik Gudbranson, Luke Schenn, Jared Cowen, perhaps Dylan McIlrath, perhaps Griffin Reinhart, perhaps Matthew Dumba.

Third category:  The Rest

Here are the defensemen 28 or under who fit the criteria this year:  Brenden Dillon, Clayton Stoner, Mark Fraser, Robert Bortuzzo, Mike Weber, Roman Polak, Dmitri Kulikov, Eric Gryba, Nate Prosser, Zach Bogosian, Brett Bellemore, Dalton Prout.  Which of these players will become top 4 D?  Kulikov and Bogosian have had better offensive seasons, but they may become Degraded Two-Way guys.  Stoner, Bortuzzo, Fraser, Weber, Gryba, Prosser, Bellemore, and Prout are all 3rd pairing guys or 7th D at the moment, and it's hard to see any of them graduating to top 4 duty (although Buffalo is so atrocious that Weber could).  That leaves Polak, Dillon, and the aformentioned Ladislav Smid to carry the torch.  Maybe Gudbranson and Schenn will move into someone's top 4 - certainly Schenn's been there in years past, though it doesn't appear that he is improving at hockey.  It sure feels like teams just cannot find these players anymore, and that much like the enforcer who can play a regular shift in the late 90s, this player is slowly disappearing from the NHL - it's just difficult to find a player who is big, skates well, hits well, and is good enough at defensive coverage to play 20 minutes a night (and who doesn't provide offense).  His replacement is the Anton Stralman, Mark Fayne sort of player who doesn't hit and doesn't put up much offense but who can defend 1 on 1 without taking penalties and moves the puck adequately out of his own zone.  Regardless, just consider this when free agency rolls around the next few years and some of the guys on the first two lists are still being given chances and the non-hitter signs for well below what you'd expect - most of the general managers in the game developed their hockey intellects when obstruction and cross-checking was still du jour, and it's hard for them to wrap their minds around the notion that it's gone.

Wednesday, December 4, 2013

The Real Problem With Big Goalie Contracts

Twitter - that bottomless reservoir of smug - always gets in a particularly joyous uproar when a team signs a goalie long-term.  Today's casualty was Henrik Lundqvist, who signed a 7 year extension for nearly $60M, a deal that will finish a few months after his 39th birthday.  Others have no doubt broken down what happens to goalies in their late 30s, even elite goalies, and most of them don't fare particularly well.  The problem, however, is not just their faring not well - it's their faring at all.  Allow me to explain.

Lundqvist's cap hit of $8.5M won't be eclipsed by another goalie anytime soon - indeed, there's a dearth of elite goalies with long track records coming to market.  Still, with the salary cap going up, someone will break it, and it will continue to be broken as Lundqvist ages.  I could see a world where he's the 10th highest paid goalie in 2018.  Whatever the case, his cap hit isn't onerous if the salary cap rises as much as experts think it will.  What is onerous is the meaning of that cap hit - it means that Lundqvist is the de-facto starter for the next 7 years.  Is he playing poorly at some point, say in 2017?  Well, what about the ten years of experience he has previous to this where he played well?  The problem isn't just with potentially getting poor value out of the contract by Lundqvist failing to be merely elite over the course of this thing - teams make it worse when they are unable to cut bait with formerly successful players.  This isn't as large an issue with elite forwards and defensemen - we've seen guys like Mike Modano, Bryan Trottier, Brendan Shanahan, Gary Roberts, Chris Chelios, etc. who have ended their careers as role players.  Were they likely entrusted with too much responsibility because of their history?  Yeah, probably, but it's an issue of 15 minutes versus 11 minutes per game.  Chicken feed type stuff - one win at the most.  You can always give a skater less ice time and less responsibility, but the picture is murkier for netminders.  This is an issue of starting a goalie 60 games versus starting 20 games - issues that are enormous.  We should all know the numbers - the difference between a .920 goalie and a .905 goalie, assuming 30 shots on goal per game, is 25 goals.  That's roughly 5 wins, which could easily turn what should be a division leader into a fringe playoff team, or a fringe playoff team into finishing well out of the playoffs.  It's especially difficult to give a guy getting starter money backup minutes if he's not playing well, and just as hard to split his ice time.

I'm not saying Lundqvist will break down like this - he's clearly one of the game's best goalies and the elite have fared better later in their careers.  What I am saying is that if he does start to fade significantly, it will probably take the Rangers a year or two to figure it out, possibly more, during which time he will actively be hurting the team.  Normally I'd say this is an outlandish prediction, but there's nothing more certain in sports than management hoping the past can once again be the present - the longer the legacy, the more inescapable the vortex.

Monday, November 25, 2013

The New Type Of Trade

Martin Erat just publicly demanded a trade out of Washington.  Trouble for Martin Erat is that his contract is pretty large and so far his performance rather paltry - he's got a cap hit of $4.5M this season and next season.  While he's owed only $6M in this time, there aren't that many teams who can just stick that salary on their team.  There is, however, a way that Erat can be shipped out of town, the Caps can still be in the same cap trouble (by taking a contract back), and the Caps can pick up an asset along the way:  The dual Retained Salary Transaction.

We've seen the Retained Salary Transaction a few times since the new CBA was signed in January 2013 - Kris Versteeg, Matt Frattin, Jussi Jokinen, Thomas Vanek - all are currently on new teams who are not paying their entire contract.  What we haven't seen is a swap of retained salary transactions, and this disappoints me, because I think this could be quite a boon for a team.  An example - imagine a team with a player like Erat who has too high a cap hit and not enough production.  Let's say for the sake of argument that, unlike Marty Erat, this player lacks a no-trade or no-movement clause.  They find a team who has a similar player on a similar cap hit, let's say both guys are on deals that pay them $4M a season.  They swap players, each picking up half of the other team's contract.  Now not only have they shuffled some deck chairs, both teams have turned what was a potential liability into a potential asset.  They can keep this new player at his $2M salary, or if they don't like him, fob this new player off to someone else on half of his previous cap hit.  The key to this deal is that both teams have picked up potential assets by exchanging their problem with the other team, and in an NHL where few teams have cap room to spare, it opens up more potential for future exchanges.  I'm sure the union was in favor of Retained Salary Transactions - it enables some guys whose contracts would've been near-immobile to be sent elsewhere - but I think there'll be a few guys zipping around the league who aren't fans of having 3 separate teams holding parts of their contract.

Friday, November 1, 2013

On Corsi Rates and Corsi +/-

Occasionally I feel the need to look for old data, back when Corsi and Fenwick were new and when was updated at the beginning of every season, without exception.  That brings me to the old version of, which expressed our favorite numbers as a whole number, not as a rate.  Jack Johnson wasn't at 43.9% in 2008-09 according to timeonice, he was at -131 - that was his Corsi +/-.  This quickly became non-standard, and so old timeonice numbers became a pain - you had to do tedious calculations to get at territorial percentage numbers, which is all everyone seems to care about now.  Still, I miss the whole numbers for three reasons:

A:  You can approximate a player's expected +/- based on a 1000 PDO.  I don't have average Fenwick and Corsi shooting percentages handy - I know they're floating around somewhere - but we know that on-ice even strength shooting percentages are around 8%, typically.  So if Jack Johnson is -81 in shots in 2008-09, we know that goals are scored on around one out of every 12.5 even strength shots, so he should be approximately -6 with average goaltending and average shooting.  Henrik Zetterberg was +332 shots in 2007-08, meaning he 'should've been' +26 at 5 on 5 hockey.  With rates, there's no way to estimate this except 'higher = good, lower = bad'.

B:  High-event players get privileged over low-event players, which may be correct - Imagine a thought experiment - one player is on the ice for 90 shots/60 for his team and 60 shots/60 against.  Another is on the ice for 30 shots/60 and 20 shots against.  They have the same shots rate - 60% - but Player A is going to be putting up +30 shots/60 whereas Player B will only be doing +10.  Indeed, we can find real world examples of this - Jonathan Cheechoo and Alexei Ponikarovsky played about the same amount of ice time in 2007-08.  They put up similar shot %s - 57.5% for Ponikarovsky, 56.7% for Cheechoo.  Cheechoo even had a little more ice time than Ponikarovsky.  And yet, Ponikarovsky is +121 in Shots, Cheechoo +104, because Ponikarovsky's lines shot the puck more than Cheechoo's but also gave up more shots.  It's not much, but Ponikarovsky should've had an extra goal (and a half).  When we go by rates, we miss out on the extra value that higher event players may provide*.

C:  It's closer to a WAR-type thing.  Yes, we have to adjust for Zone Start.  Yes, we have to adjust for Quality of Teammates.  We may even have to adjust for Quality of Competition.  But we have a better sense of how many goals this player is adding to or subtracting from their team.  4th line players get dinged for having fewer total events.  Injured players get dinged for having been injured.  One and two game callups no longer look like superstars or complete disasters - they just look like exactly what they are: guys who might've had a good night, bad night, or an inbetween one.

I know that territorial metrics can't come close to approximating a player's actual value, but when we go by rates, we're missing a larger element - ultimately, it's not rates that win games, but shots.  We should be talking about rates for teams and more about whole numbers for individual players.

* - High event players may not be better than low event players because of the NHL's point structure - the OT point implicitly rewards low event teams, as games with fewer events should have fewer goals and games with fewer goals are more likely to be tied at the end of 60 minutes.  Still, were this not the case, high event players with the same shot % would be better than low event players.