A snake draft evens the picks, not the totals

What serpentine order actually equalises

The mechanism is the one described in our guide to picking teams without an argument : the Weighted Team Generator sorts everyone by rating, deals them out one per team, and reverses direction each round so the last picker of one round takes the first pick of the next. What that post does not spell out — and what this one is about — is the difference between what serpentine order guarantees and what people assume it guarantees.

It equalises pick position. Everybody's average slot in the draft comes out close to level, which is precisely why fantasy leagues use it instead of a plain repeating deal. It promises nothing at all about the sum of the weights on each side. Those two things coincide when your ratings are neatly paired and come apart when they are not. The sample list on the tool page is the tidy case: 5, 5, 4, 4, 3, 3, 2 and 2 across two teams gives you 14 and 14, dead level, because every rating has a twin that lands on the other side.

The six-player case that lands two points short

Now take six players rated 5, 4, 3, 3, 3 and 2, split into two teams of three. The snake deals the 5 to team one and the 4 to team two, turns around and gives team two a 3 and team one a 3, turns again and gives team one the last 3 while team two takes the 2. Team one holds 5, 3 and 3 for eleven points. Team two holds 4, 3 and 2 for nine. A two-point gap out of twenty.

There is a perfectly even split sitting right there and the draft walked past it. Put the 5 with a 3 and the 2 for ten, and the 4 with the two remaining 3s for ten as well. Same team sizes, same players, zero gap. The snake could not find it because the snake never looks at totals — it only ever asks whose turn it is. Finding that 10–10 split means checking combinations against a running sum, which is a different and much slower kind of algorithm, and it is not the one running here.

Why clicking again will not fix it

This is the part that catches people out, and it is worth understanding because it saves you a lot of pointless re-rolling. Before the sort runs, the tool shuffles the players. That shuffle is real, and with a seed in the Options box it is repeatable. But the sort that follows is by weight, descending, and a sort by weight only leaves the shuffle showing where two players are tied. Everything else gets dragged into strict rating order no matter what the shuffle did.

So the shuffle decides which of your three equally-rated 3s is drafted first. It cannot decide whether a 3 goes before the 4, because 4 is bigger. And since those three 3s are worth exactly the same, swapping which one lands where changes the names printed on each card and changes nothing whatsoever about the totals. Eleven and nine, every single time. Re-rolling a weighted draft moves people between equally-rated slots; it does not search for a better balance, because it is not searching at all.

The flip side is genuinely useful: if every player has a distinct rating, the shuffle has nothing left to reorder, and the tool becomes fully deterministic. Same list, same team count, same teams — seed or no seed. That is a feature when you want to hand someone a result they can reproduce, and a surprise if you were expecting variety. If you want variety with distinct ratings, you have to change the ratings, not the seed. Seeds do a lot more work on a weighted single draw, where the tilt is genuinely probabilistic , than they do here.

Read the totals, they are printed for you

Every team card shows its total in the heading, and that number is the entire quality check. Build the teams, glance at the totals, and decide whether the spread is small enough to live with. A point or two across a casual five-a-side is noise. The same gap across a four-player list where one person carries half the weight is the whole game.

When the gap is too big, the fix is to edit the ratings rather than re-roll. Weights are the only lever the draft responds to, so nudging one player from a 3 to a 4 because they have been playing well changes the sort order and therefore changes the split. This is also the moment to be honest that your ratings are opinions: a one-to-five scale compresses a lot of real difference, and two players you both rated 4 may not be interchangeable at all. The draft cannot know that. You can.

The unrated player problem

Ratings use the same suffix syntax as the rest of the site — the weighted draw post covers the accepted formats , and a list written for one tool pastes straight into the other. What matters here is the consequence of leaving a rating off. A bare name is not rejected; it is taken as a 1, and in a draft sorted highest-first a 1 means dead last. In a weighted draw that default is harmless, because a weight of 1 still wins sometimes. In a draft it is decisive: that player is picked last, every time, by construction.

So a name you simply forgot to rate does not get average treatment, it gets the floor — worth a scan of the list before you build, especially if you pasted the names in from somewhere else. Two smaller edges: asking for more teams than you have players is allowed and returns empty team cards rather than an error, which is occasionally useful when you are setting a board up to fill in later, and the team count floors at two, so you cannot accidentally build one team containing everybody.

When a weighted draft is the wrong tool

Reach for weights only when skill genuinely differs and people care about the result. For a trivia night among friends, unweighted random teams are faster to set up and cause fewer arguments than publishing a numeric rating of everyone in the room — the choice between equal sizes and equal skill, and which tool fits which , is worth settling before you start typing ratings at all. Weighted drafting also assumes team strength is roughly the sum of its players, which holds up for pickup football and falls apart for anything with fixed roles, where one specialist matters more than two generalists.

Used within those limits it is quick and it is good enough. Rate the players, build the weighted teams , read the two totals, and either start the game or move one player across by hand. The draft is there to get you to a sensible starting point in five seconds, not to prove that no better arrangement exists — and knowing which of those it is doing means you will stop clicking rebuild and hoping.