If you’ve played pickup basketball, you’ve asked this question. Every gym in the world has their own system, and you might think I’m going to say there’s no one best system. But there totally is. It’s the way we do it at the Embarcadero YMCA right here in San Francisco, California.

First, I’ll talk about some properties a good system should have. Then I’ll go over some existing systems and why they’re bad. Next, I’ll talk about the Embarcadero YMCA system (which I’m gonna try branding as “SF-style” to make it more memorable) and why it’s the best I’ve ever seen. Finally, I’ll reluctantly admit that it has some flaws.

What makes a system good?

  1. Even teams
    Everyone likes to win, but overall people have a better time when the games are close and competitive

  2. First come, first served
    The earlier you show up, the earlier you get to play.

  3. Equal playing time per person
    Good and bad players are equals in the eyes of god.

  4. Speed
    The court should have as close to 100% utilization as possible.

  5. Orderly succession
    The system should be clear, simple, and comprehensive, and should feel fair to participants.
    • A. It’s boring to watch people argue
    • B. It’s stressful to have to do the arguing

  6. Politesse
    Ideally the system doesn’t make bad players feel bad

  7. Winner stays on
    It’s good to give people an incentive to win so they’ll play team ball and the game doesn’t collapse into LaMelo-esque folly.

Existing systems and the problems with them: Initial teams

As far as which 10 people play, there’s pretty universal agreement that it’s the first 10 to show up, which is great, although occasionally there are arguments over when people got there so ideally there’s a chalkboard where people can write their name down.

As far as sorting those 10 into two teams of 5, I haven’t seen THAT many different systems:

“Shoot for it”, with the first five to make it on a team

This is the most common custom I see, and it’s awful — it’s as close as you can get to a recipe for producing maximally unequal teams (I guess maybe if you lined people up by height and did tallest vs. shortest that would be less equal?), AND it also takes a while since you’re waiting on 5 people to make a shot.

“Shoot for it”, but alternate, so 1-3-5-7-9 are on a team

Results in fairer teams than the above but takes twice as long (actually more than twice as long because 6-9 aren’t as good at shooting as 1-5 and take more shots on average to make their shot).

Two captains pick teams

This tends to result in relatively fair teams and isn’t bad honestly, but it has a few issues:

  1. It can be hard to holistically balance the teams e.g. if there’s one player who’s by far the best and one who’s by far the worst you ideally want to pair them, or if the three best are approximately equally good and everyone else is far below, you want to make sure the team with two of them has deficiencies elsewhere on the roster
  2. It can feel bad for the players who are picked later (although it’s probably not news to them anyway so I don’t care much about this)
  3. The captains might prioritize things like choosing their friends over trying to pick best player available, which can lead to team imbalance

Pair up + rock paper scissors

I’ve only ever seen this used at the Presidio YMCA, but listing it here for completeness. The way it works is everyone matches up with someone they think they’re roughly as good as, they play rock paper scissors with that person, and all the winners are on a team against all the losers. Like the “captains” approach, this results in teams that are relatively equal, but no two players are exactly as good as each other, so you do get situations where all the slightly-better members of the pairs win and the teams wind up somewhat imbalanced.

Existing systems and the problems with them: Succession

Who are the next five to play?

When there are more than 5 people waiting, it’s usually just the next five people who were waiting, in the order they started waiting. This is good and how it should be. Some gyms mess this up though by requiring that you “have a 5”, meaning that if there are three or four good players who all want to play together and “have next”, they can exclude fourth or fifth players from their squad and pick someone later but better, or even pick up the best player from the losing team. Obviously this is terrible along almost all dimensions — it means bad players play less and feel worse, it leads to superteams, it violates first come first served, it results in arguments, and it has nothing to recommend it. But sometimes people think they can get away with it.

Where it gets complicated though is that because the losing team all lost together, they definitionally all started waiting at the same time. Imagine there are two newcomers who want to play (twelve total in the gym) — which of the five losers gets to join up with the two newcomers?

Mostly there are two paths here:

  1. Shoot for it
    This has the virtue of simplicity and feels fair, but like all other shoot-for-it-based approaches it takes a bit and means bad players play less.

  2. Newcomers choose
    Also means bad players play less, and feels less fair/creates more bad vibes than shooting for it. At least it’s faster.

If there are 11-14 people, whoever from the losing team didn’t make it on is almost always locked for the following game. If there are more than 15, typically the next five have already been prenegotiated, so the losers have to join up with the “remainder” after the full fives that are waiting have played.

What happens if someone leaves the winning team, via injury or just because they have to go home? In some gyms, they get to pick up whoever they want (unfair, superteam); in others they shoot for it (takes a long time, bad players play less).

SF Style: The Complete Guide

Initial Teams

When you show up, you write your initials on a chalkboard. When there are 10 initials on the chalkboard, it’s time to gather everyone and choose teams. Typically by the time people gather, there’s an 11th person. This person chooses the teams. This aligns incentives perfectly because the more equal they make the teams, the worse on average the winning team will be, so they’ll have a better chance to beat them the following game. They can take a holistic outlook and try to make sure that both teams have someone who can dribble, someone who can get rebounds, the defensive matchups roughly make sense, and skill differentials at one spot on the roster are compensated at another point, and they can do it all “in their heads” without having to spell out exactly who the best and worst players are.

If the 11th person is new to the game and doesn’t know the players, they can defer to the 12th person, or if there are only 10 a couple trustworthy oldheads can figure it out amongst themselves.

Succession

When a game ends, do the losers sprint to the chalkboard to write their name above everyone else’s? No. Instead, they huddle up and each quietly chooses a number 1-5. Once that’s done, someone who’s coming on and didn’t watch them choose shouts out the numbers 1-5 in any order they want. That’s the order people come on in. So if there are three people waiting, and the called-out numbers are 5-3-2-4-1, then whoever chose #5 and #3 will play, and #2-#4-#1 are on next. If there are 8 people on the board, than #5 and #3 will play with the next three after the first five play.

If someone departs the winning team after the game, the next person on the board plays (so in our scenario with three people waiting, #2 (the third number chosen) joins the winners.

So it’s mostly pretty simple! But some other details for comprehensiveness:

  • If a player gets injured in the middle of the game, the next person on the board hops on and gets a sort of “free option” — if they win, they get to stay on with the winners; if they lose, they stay on for the next game as they would if they hadn’t played. This is needed to make sure that people aren’t disincentivized to join an in-progress game because they might lose their opportunity to play a full game.
  • If you show up before the game ends (i.e. you see the final shot or see time expire), you play before the losers; if you show up when the game has already ended (while the number-choosing process is underway), you’re behind them.
  • Sometimes we start out 4v4 if there are only eight or nine people and then transition to fives when more show up; when this happens, if there are exactly two people, the better player goes to the losing team and the worse player goes to the winning team. If there are more than two people, they all come on together as their own team, pick up players from the losing team in number-choosing order, then the winning team picks up the next player from the losing team.

Virtues of the SF System

  • It’s purely objective, so no one feels cheated
  • The random ordering means everyone plays roughly the same amount.
  • While the randomness doesn’t prevent superteams (imagine three good people are waiting and happen to be joined by the best two players on the losing team), it doesn’t actively cause them like all the other systems.
  • It’s quick to execute and you don’t spend a lot of time waiting for people to make shots
  • Most importantly: I can’t emphasize enough how pleasant it is to have a clear system with no arguments and no need to jostle for position and try to form alliances.

Flaws in the SF System

There are a few so I’ll call them out.

  1. You need a chalkboard or at least it goes a lot smoother with a chalkboard. In theory you can do without the chalkboard if everyone’s on the same page and there’s a trusted person with a decent memory who’s supervising.
  2. It’s hard to generate perfectly random numbers, and most people don’t try, so there is a minor landgrab to get 3 and 4. Me, when I choose the numbers, I make them Actually Random by doing something like choosing someone on the court and adding up all the letters in their name (A=1 through Z=26), then taking the remainder mod 5. Also in theory [infohazard alert] the best players can cheat the system by developing a rep for always picking the same number — we had one guy who used to do that but he switched gyms.
  3. It’s unusual and not the easiest for newcomers to understand (obviously if it were more widespread this would go away).
  4. It’s not self-correcting — I think in the ideal system if you got a bad number the first time around, you’d somehow have priority in future random drawings, but this is probably more trouble than it’s worth to fix
  5. Better players still play more than worse players because their teams win more. (But to the lowest degree possible without abandoning “winner stays on”.)

Conclusion

SF-style is best style! Talk about it to the people in your run today.