Robertson’s Rant: On Accounting and Incentives

Gary Bettman takes a lot of crap for things that probably aren’t really his fault, or at least not his alone. Yet as commissioner, he is the public face and cheerleader for all manner of unpopular and/or boneheaded moves, and thus gets jeered by fans across the NHL. (Funny enough, it doesn’t work the other way around: good moves tend to be attributed to the Board of Governors, the GMs, the Competition Committee, or even individuals like rising-star League executive Brendan Shanahan.) Take, for example, the “Gary Bettman loser point.” It probably actually came about as the brainchild of some GM or Hockey Operations type, who figured that by reducing the risk inherent to overtime, teams might be encouraged to go for it once they got there, instead of taking the risk-averse route and playing for the tie. It was a nice idea in theory, but ultimately an incomplete one.

At first, the loser point appeared to be doing its job and reducing the number of tied games1, but over time, coaches figured out how to game the system, and as they are wont to do, promptly ruined everyone’s fun. Teams started playing for the tie more in regulation, in order to guarantee themselves at least one point.2 As the perceived value of a standings point increased in the final third or so of the season, they probably played for the tie more in overtime, too: after all, why risk giving up the third point to a team you’re in a race with when you can trap it up and deny them that extra edge?3

The introduction of the shootout in 2005 was supposed to fix this by removing the tie outright, but instead it merely moved the goalposts. While at first, coaches may have resisted taking a chance on something unfamiliar, once they realized that it operated on the Any Given Sunday principle, i.e. that anyone could win a game with just three shots, they started playing for that instead – especially if their team wasn’t very good at the main game.4 Shootouts are on the rise with each passing year, and various proposals have been put forward to try to limit them in some fashion, including removing shootout wins from the tiebreaker, as was done this year5, and extending overtime to include a 3-on-3 component, which will likely happen in the next couple of years.

Wait, where are you going with all of this?

As I see it, there are two separate but related problems with the standings system. The first is that, because some games are worth three points and some only two, the standings are, frankly, a sham: teams now earn playoff spots and home ice, not on the basis of their ability to win hockey games, but on their ability to force overtime and win a skills competition. The second, which I’ve been laboriously working toward, is that the incentive structure is all wrong. A victory, regardless of how it comes, is worth the same two points it has been since time immemorial; a loss, depending on how it comes, could still earn you a point, without costing the other guy anything. There is a benefit to forcing overtime, but no corresponding cost to balance the system.

To demonstrate, let’s look at Eddie’s LA Kings, now mired in 11th place in the Western Conference despite a +18 goal differential, the fourth-best total in the West. In looking at their record, I immediately saw why: the Kings have just one OTL point all year, while no other team in the West has fewer than four, and most teams have five or six. Turn five of LA’s regulation losses into OT losses, and they move from 24-20-1 to 24-15-6 – one point out of fourth! If you go the other way, and lump all losses together, the Kings have the fourth-fewest aggregate losses in the West (21), behind only the three division leaders. Yet because the Kings have failed to force overtime to the same extent as their peers, they’re on the outside looking in. That doesn’t make a lick of sense.

So, what would you propose?

For years now, I’ve been a supporter of the three-point system used internationally: three points for a regulation win, two for an OT win, and one for an OT loss. While any system that rewards losing still strikes me as a bit wrong-headed, and certainly doesn’t address the incentive problem in overtime, it lessen the impact of said problem by encouraging teams to try to win it in regulation instead. After all, in a tight playoff race, who’s going to take the chance of giving a rival a point and sacrificing one in the process?

I don’t follow European hockey, so I don’t have numbers to support this, but anecdotally, I can point to the 2010 Olympics. In the round robin, Canada and Switzerland were tied with about ten minutes to go. Canada poured it on like a medal was on the line, because they knew that if they lost that third point, and didn’t beat the US, they’d be out of contention for a bye to the quarter-finals. While they were ultimately unable to forestall overtime (or beat the US, at least in the round robin), the fact that Canada played like they did beforehand tells me that at least in my sample of one, the incentive structure worked as intended.

The other benefit, of course, is that the standings would finally make sense again: every game would be worth the same, .500 would mean what it used to mean, and teams would be appropriately rewarded for winning in regulation. Sure, the record book would fly out the window, but I would argue that that already happened, back when the OTL and later shootout were implemented: wins and points totals are already inflated because of the systems now in place.6 I believe this will lead to more exciting finishes in regulation, as teams try to grab that third point for the win, and thus fewer shootouts, especially if coupled with something like Ken Holland’s 3-on-3 overtime plan. I think both results would be beneficial, since I’ve always felt that shootouts were a gimmicky and fundamentally unfair way to decide a team game.

*          *          *

Postscript: Because I’m sure someone will be interested, here are the standings up to Monday night’s games, under the current system (L) and under the proposed three-point system (R). For the sake of a consistent comparison with different numbers of games played, I sorted it by points percentage, rather than points. (Note: NSOW = “non-shootout wins,” the NHL’s new tiebreaker for 2010-11; RW/RL are regulation wins/losses.) Unfortunately, it can’t really illustrate my key point – because it’s based on this year’s data, it’s based on this year’s incentives, so the Kings still get boned – but it at least shows what a three-point standings system would look like.

1 The average number of ties per team per season held steady around 9-10 ties/team from the re-introduction of regular-season overtime in 1983 until the mid-‘90s; from there, it steadily climbed to around 12 ties/team by 1999, when the OTL point was introduced. The experiment worked, kinda, for a couple of years: the average dropped to around 10 ties/team for a couple of years before climbing back up to 11.3 ties/team the last year before the lockout. Incidentally, if you’re looking for a reason for the increase in ties through the ‘90s, it’s probably related to the general slogging-down that crept into the game over that period of time: teams went from averaging just under 300 goals per season through the late ‘80s and early ‘90s to just over 200 goals per season between the late ‘90s and the lockout.

2 The number of overtime games per team per season increased from 16.5 in the last two years pre-OTL to around 18 the first three years after, and nearly 21 (!) in the final two years before the lockout.

3 I really wish I had a tally of how many ties/shootouts happened at a given point of the season (by halves, quarters, or months; any would work) to back this assertion up.

4 Indeed, the average number of overtime games per team per season held steadily at 18 from 2005-09, before increasing to 20 last year; meanwhile, the average number of shootouts went from 9.7 in the first post-lockout year to 10.6 two years ago to 12.3 last year.

5 Removing the shootout win from the tiebreaker may be working: the number of OT games is down from 20 to 18.7 (prorated) this year. Meanwhile, the average number of shootout games has decreased from 12.3 to 8.9 (prorated) this year, the lowest post-lockout total, and in line with the number of ties we saw in the ‘80s. These could be year-by-year fluctuations, with last year more of a blip, or they could be evidence of the system working as intended. We probably won’t find out, because they’ll change the system long before we have enough data, but it’s interesting nonetheless.

6 Average wins and points, respectively, since the lockout: 41 and 91.4; between 1983-99, adjusted for an 82-game schedule, they were 35.9 and 82.

About Doug Stolhand 27106 Articles
Doug Stolhand is one of the co-founders and co-hosts of the Puck Podcast and has been a member of the NHL media since the show's inception in 2006.


  1. Teams will always play to the best numerical advantage. What I don’t get is how your method would change that. You’re setting the game back to an ideal of how the game *should* be played and won, and that’s fine but I think you need to strap on your onion belt and yell “Get off my lawn!” Sports were never about which team is objectively better (if it were even possible to determine that). The best team is always the team that games the system most effectively. A 3 point game wouldn’t change anything aside from making that 0.01% who is obsessed with the math of these things breathe easier.

    There are all sorts of different ways to crunch numbers. The ’06 oilers wouldn’t have made the playoffs in a 3GS system but they were a win away from a cup. Every different calc will include and exclude different teams. How is your method any better or less axiomatic than the others?

    But we both know the answer to that. This axiom is closer to history. Y’know what else is closer to history? Clutch/grab. At some point you need to accept that the game evolves. Rules, scorekeeping and all.

  2. I think you’re missing my point. What I’m saying is that not having all games be worth the same number of points is inherently ridiculous, and that the way the points are distributed now favours playing for the tie, which is not terribly exciting. By “gaming the system better,” as you say, we would get more situations where a team goes for it with five to go instead of accepting the tie and trying again in OT/SO. None of this has anything to do with The Way We Were or any such nonsense; just the opposite, in fact. If that was my thing, I would be pounding the W-L-T, get-rid-of-the-shootout drum. Note also that I’m perfectly willing to bin the record book in favour of a standings system that makes better sense in terms of both accounting and incentives.

    As for the ’06 Oilers, I think that’s entirely irrelevant. As I said in the postscript, there’s only so much you can learn by converting standings, because at the end of the day, different incentives will produce different results. If we’d been under a 3PS in 2005-06, the Oilers’ results as they were generated in the 2.3PS would have had them out, but maybe the Oilers don’t play a League-leading 16 shootouts in that scenario, either. It’d only take a couple of those one-point OTLs becoming three-point RWs (for example) to change the standings again.

    And finally, clutch and grab? Really? Why not bring up bench-clearing brawls, too? I know you’re better than that.

  3. Incidentally, to follow up footnote #3, here is a post from last year showing teams playing for OT during playoff races since the lockout. Apparently, it didn’t become a particularly pronounced thing until after the SO was implemented. Make of that what you will.

  4. My point is that you can’t change the system without creating new problems. The loss of clutch/grab has increased the speed and then increased the frequency and severity of headshots. Maybe the 3pt system would be better, but better than what? I mean, any change that way is fundamentally changing how the game is played.

    There is absolutely no evidence supporting the idea that a 3pt system would be more fair or better for the game. At that point I think the only argument you have is the historical one.

  5. I’m fuzzy on one thing. If a win is 3, an OTW is 2, a shootout win is 1 and a loss is 0, what incentive does a team have, in a tied game to go for it in the last 5min? Wouldn’t that make them *more* conservative? More to lose, less to gain?

  6. I guess I’m not seeing how the historical argument plays in at all, because a three-point system has nothing to do with history except for “all games are worth the same,” which seems like a pretty basic thing. Your point is well-taken about unintended consequences: I don’t think the NHL had planned on everyone playing for the tie when they introduced the OTL. Maybe they do play more conservatively because of the difference between three points and none, but no matter what, they’re gonna lose a point if they don’t win in regulation, which I would think, in the thick of a playoff race, they would be loathe to do. Even if they then win in OT, it’s the difference between gaining one point and gaining three on a rival in the playoff hunt. Only way to know for sure would be to try it.

  7. I agree. I just don’t think there’s any impetus to try it other than really caring about what .500 means (which is where I was going with “historical,” though I realize that probably wasn’t clear at all.

  8. I don’t think that’s a history argument per se, so much as a math argument. 50% should be right in the middle: right now only the five draft lottery teams are below .500. That’s bloody absurd.

    Also, something to consider is that it’s not just about encouraging teams to go for it in a tied situation, as may not even be where the incentive has its best impact. In thinking about it more, it seems more likely that the change would come when a team is up one or two goals. Right now, they have a fallback (OT), and it costs them nothing to give up that tying goal, so they play passively, don’t take chances so the coach doesn’t yell at them, etc. If there’s a guarantee of losing at least one point, and possibly two, I would think players would stay aggressive for longer, especially in a playoff race.

    So since we’re discussing this, it seems like the other alternative would be a simple, baseball-style W-L system. Instead of increasing the incentive for winning in regulation, you’d be decreasing it for going to OT, by having all games be all-or-nothing. How do you think that would change the equation?

  9. In a winner-take all system, the shootout would be necessary. I guess that would come down to coaching. If the other team is really good at the shootout you’d be trying your damnest to win in reg and vice versa.

Comments are closed.