At the Sloan Sports Conference earlier this year, Adam Gold, a University of Missouri PhD student, proposed a way to change how the draft order is determined. Instead of using the reverse standings, he suggested using the number of wins or points generated after teams were eliminated.
This would, in theory, remove the incentives to tank once a team can no longer make the playoffs, while still ensuring that the worst teams (who are eliminated earlier) maintain their high draft positions. It got noticed – by Oilers fans and followers, to no surprise – and debated, but I wondered what would actually happen under such a system.
We can’t look at Edmonton’s results last year after being eliminated and conclude anything, because they had no incentive to win those games. In fact, Gold shows that teams who are eliminated go from winning 45% of their games to just 28% – clearly, they wouldn’t try to lose nearly three-quarters of the time if wins were rewarded instead of losses.
But even if we had the right conditions to use last year’s data, or a couple of years, we’d still be at the whims of random variation. Anything can happen in 82 games, and that’s all the more true in the five or 10 games a team plays after being eliminated.
So we have to try something else: we’re going to play a million seasons.
We’ll go through every game from last year’s schedule and assign a winner and loser. In 24.3% of the games in the last three years, the losing team has received a point (i.e., the game has gone to at least overtime), and in 13.9% of all games, the result was decided by a shootout. In shootout games, we’ll decide the winner by coin toss, and award points accordingly. For all other games, including those decided in OT, we’ll figure out the winner based on the two teams’ SRS, as well as the standard home-ice advantage in the NHL.
Before each game during a season, we’ll check if either team is already eliminated and if so, we’ll keep track of their games played, wins, and points after being eliminated. Then the team with the most wins after elimination receives the first overall pick, and so on down the line. And we’ll repeat all of that one million times to smooth out the inevitable randomness.
But when is a team eliminated? That’s a crucial question. To simplify matters here, we’ll say a team is eliminated when they can no longer surpass (in total points) the eighth-place team in their conference. This isn’t quite true but close enough for our purposes. A team may still be alive, technically, if they’re 10 points back with five left to play, but that requires at least a five-game win streak and a five-game pointless streak (insert joke here about the Islanders having season-long pointless streaks). Which is very unlikely.
Of course, there are going to be unintended consequences if this system ever takes over, many of which we won’t know until they happen. A team that knows they won’t make the playoffs could tank before they’re officially eliminated, get eliminated very early, then go nuts with the “draft points.” In some ways, that’s the exact same scenario we have already — the tanking is moved from March/April to October/November. Every year we see teams who aren’t mathematically eliminated but have no realistic chance at a playoff spot. How do we know they don’t give up before being eliminated, once they realize it means a slightly better chance at that #1 pick?
The first column is how often this team made the playoffs in our 1,000,000 replays of 2011-12, and the second is their average number of points overall (in all 82 games). The third column is how many games they played, on average, after being eliminated (GAE), and the fourth is how many wins they averaged after being eliminated (WAE) in those seasons (which is what we’re sorting by).
The last column is how often they would receive the first overall pick outright (ignoring years where teams tied for the first draft spot).
Team Elim% Pts GAE WAE First% CBJ 2.5% 76.1 8.2 3.2 17% MIN 4.2% 78.9 7.2 2.9 14% NYI 5.8% 78.3 6.6 2.7 9% TBL 6.7% 79.9 6.1 2.5 10% TOR 15.3% 83.4 4.4 2.2 6% EDM 13.6% 84.7 4.5 2.1 6% FLA 22.9% 85.6 3.7 2.1 5% ANA 16.5% 84.8 4.3 2.0 4% CAL 19.0% 85.7 3.8 2.0 5% CAR 17.5% 84.0 4.0 2.0 5% WPG 22.8% 86.3 3.4 1.9 4% MON 32.8% 88.2 2.6 1.8 2% DAL 30.8% 89.2 2.8 1.6 2% COL 30.0% 88.9 2.4 1.5 2% WAS 38.4% 89.9 2.1 1.4 2% BUF 37.1% 88.8 2.3 1.3 2% PHO 57.8% 95.2 1.2 1.4 1% LAK 58.8% 95.7 1.1 1.2 1% CHI 55.4% 94.6 1.3 1.2 1% NAS 76.7% 99.2 0.6 1.2 0% SJS 63.7% 96.7 1.0 1.2 1% OTT 60.6% 94.3 1.1 1.1 0% NJD 72.4% 97.0 0.6 1.1 0% VAN 91.8% 105.5 0.2 1.0 0% DET 89.9% 104.0 0.2 1.0 0% STL 89.8% 103.9 0.2 1.0 0% NYR 89.1% 102.5 0.2 0.9 0% PHI 83.5% 100.4 0.3 0.8 0% BOS 98.5% 109.7 0.0 0.8 0% PIT 97.0% 108.1 0.0 0.7 0%
Some teams did notably better or worse last year than in these simulated seasons, such as Edmonton or Columbus, which we can interpret to mean those two teams had somewhat unlucky years.
What’s encouraging about these results is they imply that this new system would maintain the entire reason to have a draft (distribute new talent to encourage parity) while removing its current worst feature (Fail for Nail, Stop Winnin’ For MacKinnon, etc.). Half the time, the first pick would go to one of the four worst teams — CBJ, MIN, NYI, TBL. Almost never does a 9th or 10th-place team receive it.
So the bad teams still pick earlier, but now they’d actually have to earn that pick. I can get behind that.
This wouldn’t be the way the entire draft order is determined — only the first 14 teams can be ordered based on “points after elimination”, obviously, and the rest are done with playoff advancement. But it has almost the same teams in the top 14 as the actual 2012 order did: only Florida didn’t originally have a draft pick in the top 14 this year, and they of course made the playoffs.
The Oilers and Canadiens would likely not be very happy about this new system, should the results above hold in reality. The Habs fell to last in the East this year but weren’t as bad, really, as the Hurricanes, Jets, Leafs, Lightning, and a bunch of other teams who picked after they did. Given how close the bottom of the East was, Montreal certainly wouldn’t always finish last if we played 2011-12 over again.
Edmonton’s a little different: they were undeniably a very bad team, but because their SRS was better than Minnesota and Columbus they were eliminated a few games later on average in these million seasons, after 4.5 games instead of 8.2 or 7.2. And with three games in hand, those other teams could obviously accumulate more points.
Which leads us to a big problem under this new system: determining when exactly a team is eliminated is not nearly as easy as it sounds. There’s a reason we simplified the process above. If there are more than a handful of games left, it actually requires complicated mathematical gymnastics (or some effort equivalent to that) to be completely accurate. Gold brings up the word “combinatorics” in his talk and he’s right to do so.
It’s not just a nitpicky problem, either. Adding an extra game to any team’s post-elimination schedule could change these standings. As an example, Edmonton had 32 wins in 82 games (in their actual results), or 0.39 per game. If we mistakenly identified them as being eliminated one game earlier than they really were, their average of 2.1 wins after elimination here would be inflated to a 2.5 – enough to vault them from sixth to fourth. Two more games would mean 0.8 more wins, or enough to overtake Minnesota for second.
And this shift one way or the other could happen with anyone. A team would probably accept losing the No. 1 pick if it’s taken away in a random draft lottery where the probabilities are known to all, but not when it’s taken way by some mathematician in a back office somewhere, regardless of the accuracy of that person’s work.
You could probably avoid this problem by announcing officially every morning who has been eliminated, so all teams have the same information and there’s no ambiguity. Columbus or whoever could argue they should already be eliminated, but if they know they’re not, at least they’re aware of the situation and it’s not decided after the fact.
So it may work or it may not work. There’s a lot that could happen that we can’t anticipate. We can say that this new system will probably not change the order significantly – bad teams first, mediocre teams later, playoff teams last – and that it would remove the need to balance winning against draft picks. (Presumably, this would also remove the draft lottery, which is no longer needed to guard against the worst tanking.)
Besides, if two teams were tied in the draft standings and were playing each other in Game 82, you don’t think that would be an outstanding game to watch? That might be enough reason to adopt this system right there.