Gold Drafting Now

July 3, 2020, Micah Blake McCurdy, @IneffectiveMath

It is winter. Your hockey team is bad. In October, you had imagined that they were going to be better this year. Cup contenders, perhaps not. If you're being completely honest with yourselves, they weren't particularly likely to make the playoffs. But they were going to play meaningful games down the stretch, they were gonna be in-and-amongst-it. Instead, they are Bad; again. With a familiar feeling of sadness, you slowly turn the ship of fandom towards hoping that your team will instead lose games, cheering when the other guys score, slowly ratcheting up the lottery odds as you idly contemplate getting back into clove cigarettes. You've set a notification for that one prospect twitter account you really like and you already wonder if that's maybe a bit excessive. There is, always, next year, and to make next year any good, this year has to be bad, as bad as possible, so scavenge some funny ascii-art tanks for the silly tweets and try to make the best of the rest of whatever this season still is. How many more games? Twenty-five? Oh God.

A solution to this worryingly-familiar-for-me-personally miasma is to completely remove the incentive for teams to ever lose games. The simplest way to do so was first isolated by Adam Gold in his 2012 Sloan paper: instead of giving the best draft picks to the teams who have the fewest points, give them to the teams who have the most points—but start the clock for each team when it is no longer possible for them to make the playoffs. When this season is a lost cause, the focus naturally shifts to next season, for which securing the best possible new players is obviously important. This scheme, so easily and pleasantly named after a man with a great last name, is known commonly as "Gold drafting", and the standings points that teams obtain after they are eliminated are called "Gold points". The team with the most Gold points gets the first overall pick, next-most the second overall pick, and so on. There is no more lottery; on-ice results determine the picks completely.

The benefit is obvious but bears repeating: joy at seeing your team win is the essence of what makes sports good, and Gold drafting gives you more of it. When your extremely good team is on the verge of making the cup final playing another extremely good team, you care a very great deal. When your middling team has a win-and-you're-in against a divisional rival to perhaps go on a goalie-fueled playoff run towards madness, you are invested. With Gold drafting, when your bad team has a late-season showdown against another bad team where the winner drafts first overall next year, you are in. As things stand today, you might as well tune out until Bill Daly flips over the cardboard cards in an off-ice spectacle that pales in comparison to actual sports.

The promise of Gold drafting is 82 exciting games where you want your team to win every night, no matter where they are in the standings. This article is to advocate for its adoption in the NHL specifically, which is my area of special concern, but in fact it would benefit most leagues with a draft, none of whom use a system which is as exciting or fair.

The vital point where team priorities change dramatically is, as noted previously, the moment that teams are eliminated. However, the moment when a team is eliminated can come quite a bit later than the more uncertainly-muddy stretch of time when your team feels eliminated but ridiculous people like me are outlining just how if a variety of your divisional rivals were to lose 20 out of their combined remaining 22 games and your team were to suddenly become a juggernaut of gilgameshian proportions dropping no fewer than three points in your next dozen-ish games, you might still be able to squeak into the playoffs on a technical tiebreaking rule that you once knew about but have since forgotten. In this murky time a team may not be sure what is best: to rattle off a few wins, let this season's faint hope grow to a merely small hope; or perhaps lose a few, to hasten the inevitable, and open your Gold account officially. To crystallize the pain of this indecision for front-offices and to give fanbases something interesting and important to yell at/about, the version of Gold drafting I describe here has a new tweak: teams who want to start accumulating Gold points (which, by definition, can only accrue to teams that are eliminated) can declare to the league and to the general public that they are no longer interested in a playoff spot for the current season. Regardless of how many points they finish the season with, they will not make the playoffs; they are eliminated and their points from that moment will be Gold points. At every point in every season, every team, and consequently every fan, wants to win. If they want to make the playoffs, they need to win. If they want to get a good draft pick, they must declare themselves eliminated, at which point they need Gold points, for which they need to win. Winning must be always.

The remainder of this article is a thorough set of rebuttals to the commonest objections that I've heard over the years; none of the objections I've heard stand up to scrutiny. However, it's important to remember the benefits:

Sean McIndoe wrote about Gold drafting not too long ago, and some of my arguments line up pretty closely with his, being, as they are, correct.

Objections With Which We Will Now Dispense

Objection 1: What if a team is really bad? They need to get better. They can't win enough Gold points, because they're bad.

By far the most common misconception about Gold drafting is the idea that it punishes weak teams. After all, if you need to win to get Gold points, and your team is bad at winning, it seems logical to imagine that your team will be disfavoured. However, since the weaker teams are eliminated first, they have longer to accumulate Gold points. As an analogy, imagine a timed race where the runner who covers the most distance at the end of an hour wins; but the slower runners get headstarts. If the headstarts are large, then the slow runners will win; if they are small, then the faster runners will overtake them. Knowing which effect will dominate is not easy to intuit but fairly easily to measure.

With a simple model of thirty-two fixed-quality teams, I simulated an eighty-two-game season. Each season, I randomly sampled thirty-two teams from a distribution of strengths chosen to replicate a plausible distribution of end-of-season standings points. The average end-of-season standings point total was 92.25 points, with a standard deviation of 12.16 points. I simulated a few million seasons so that I could get a decent number of seasons in which there was a team that finished with 45 points or fewer, since the 2019-2020 Detroit Red Wings were on pace to finish that season with 45 points. (In point of fact, for every million seasons simulated, I saw just over a thousand seasons with forty-five points or fewer).

Since the team strengths in this model are known ahead of time, it's easy to compute at any point during the simulated season where the playoff cutoffs are likely to land and how likely a given team is to reach them. As a convenient proxy for "eliminated", I used "a chance of making the playoffs of less than one in a million". With that in hand, it's easy to measure how many games each team played after they were eliminated (to see how much headstart they had), how many Gold points they obtained, and, by ranking those points, what draft pick they obtained.

How much headstart do weak teams get?

A lot, though the season-to-season variance is considerable. The darker the red colour, the more common a given outcome. The black dot shows the mean outcome for a given point total; for instance, a team finishing with ninety standings points was, on average, eliminated with three games left to play. Roughly, the typical behaviour is that a drop of one standings point means that a team will be eliminated one game sooner.

How many gold points do teams get?

Mostly, weaker teams earn more Gold points. Teams with over ninety standings points regularly qualify for the playoffs (or are eliminated in their final game) and so earn no Gold points. For very weak teams, in the 45-55 point range, the effect levels off slightly. Notice the striation for teams who are eliminated very close to the end of the season - earning exactly one Gold point (by losing post-regulation just once after being eliminated) is harder to to than earning either zero or two (by winning, or by losing in regulation). Because Gold points are just standings points earned in special circumstances, their totals in short stretches show the same tendency towards even numbers.

Where do teams actually pick?

Teams finishing with 65 points or fewer pick third or better, on average. Weak or very weak teams picking fourth or later is not unheard of, but typically weak teams are rewarded with very good picks almost all of the time. For instance, a forty-five point team (like the 2019-2020 Red Wings, currently slated to pick fourth overall after being jumped three times in the lottery) can expect to draft first under Gold drafting 62% of the time, second overall 25% of the time, and third overall 6% of the time. (In the lotteries used, their chances were 18.5% for first, 16.5% for second, and 14.4% for third.)

This objection is by far the most common and it's not a remotely silly thing to worry about. The only way to justify having a draft at all is to ensure that the league is balanced, so it's important to make sure that Gold drafting accomplishes this goal as well as it does. If the details of the league were to change substantially—if the season length were to change, if the existing competitive balance were to deteriorate substantially, if the fraction of teams making the playoffs were to change, it would be necessary to check again. Happily I remain devoted to doing such nerdy work and large swathes of the hockey-loving public remain devoted to paying me to do it, so I'll update this if necessary.

In conclusion:

Gold Drafting Gives the Best Picks to the Weaker Teams

Objection 2: What about the trade deadline? If everybody wants to win, where will I get my deadline blockbuster fix?

There will certainly be fewer obvious sellers, since every team will want to keep useful players. Not every team has the same timeframe in their outlook, though; some teams are sure that this year really is their year to win the cup, others need the revenue from that extra playoff round now, still others will have a pipeline of prospects shaping up nicely and will feel that they can spend picks a little more cheaply. Conversely, every weak team's internal estimate of "how long until we are contending again" will be different, so the variety of urgency around the league will still permit moves to be made. What little diminishment in trades will be repaid handsomely first of all in more exciting games for more teams, which is intrinsically good, since on-ice excitement is always to be preferred to off-ice. Some number of current deadline deals are of the "send the player on an expiring deal with whom we have a poor relationship to another team where they immediately re-sign" form; any excitement lost from those trades will be recouped during free agency.

In conclusion:

The Trade Deadline Will Be Slightly Smaller But It's a Net Win For Excitement

Objection 3: What happens with first-round picks that get traded? Doesn't that create some weird incentives?

Picks that are traded also accumulate Gold points based on their original owner. These teams obviously have no incentive to declare themselves eliminated; but eventually every team either makes the playoffs (in which case these picks are worth whatever they were previously) or they are eliminated, after which their picks begin to accumulate Gold points regardless of who has traded for them. If a team trades their first round pick away and then later is eliminated, it is true that their future wins help the pick holder instead of themselves. In particular, one could imagine a scenario where a team could play against the team that holds their first-round pick, where the first team has nothing to try to win for, and could hurt their opponent slightly if they were to throw the game. As perverse incentives go, it doesn't really rate; in particular, it's dramatically smaller than the incentives to tank induced by the lottery being weighted towards weaker teams.

Teams will have to learn a new way of valuing first-round picks, because the rules about how they're ordered are being changed. The total value of all of those picks collectively remains the same, though, so any picks that are devalued by any change will have compensating increases of value elsewhere.

In conclusion:

Gold Drafting Dramatically Reduces Perverse Incentives

Objections 4: Why would players try hard just so that their team gets a good player next year? That new guy will presumably take one of their roster spots!

Every professional player plays for (at least) three reasons:

Moving to Gold drafting doesn't change any of these incentives, least of all the last one, which is especially relevant to players who suspect that their existing place in their team may not be perfectly certain for next year—they will need to make sure that they play as well as possible to get a spot at another team. I expect nearly all of the players on a team in serious contention for first-overall by any scheme are largely delighted at the prospect of getting that much better and therefore that much closer to being able to win.

In conclusion:

Professional Athletes Like Winning and Money

Objection 5: How do the declarations actually happen? What happens if there are only three playoff spots left and four teams try to declare themselves eliminated?

Every day at noon eastern, the league can publish the list of teams that are known to be eliminated, either by their own previous declaration or by the circumstances of the remaining games. Everybody knows heading into the day's games which teams are playing for the playoffs and which ones for Gold points. At the end of the day's games, teams who want to can send their declarations of elimination to the league. If the league receives too many, then all of the declarations are silently rejected; the public is not notified about them. If any of those teams still want to, they can make another declaration the following day.

In Conclusion:

Implementing Gold Drafting Is As Easy As The Waiver Wire

Objection 6: What if a team declares themselves eliminated way too early, like even before the first game of the season? Isn't that bad for some reason I can't really specify? It seems bad.

It strikes me as unlikely that any team would lob this bomb into their own fanbase in October, even with a weakish roster and the prospect of a McDavidian reward. If anything, teams are much more prone to the opposite folly of being quite sure that Everything Will Work Out this year, especially in the fall. That said, Gold drafting retains all of its virtues if restricted in various ways like saying that declarations have to wait until, say, the all-star break; or Christmas, or either of the thanksgivings.

In Conclusion:

You Can Tinker With The Marginal Details To Make It Just How You Like It

Objection 7: How exactly did you run that model that you made all of those graphs from? I bet it was wrong somehow.

I used a Bradley-Terry model where the fixed team strengths in a given season were drawn from a normal distribution with mean 0 and standard deviation 0.25. Each "season" consisted of 82 repetitions of the following procedure: shuffling the 32 teams into a random order and have the first sixteen play the second sixteen; each game was assigned a winner or loser randomly according to their team strengths. All losses were divided into 0-point losses or 1-point losses according to a flip of a coin weighted with 75% and 25% probabilities, respectively.

In Conclusion:

You Can Always Improve A Model But This One Is Cheap And Cheerful