Ranking Filter

Blind ranking strategy

There is a worked-out rule for where each card should go, and it dates from 1972. Mark the slots still open, say how good the card is, and the calculator gives the move.

Slots still open

Better than 70% of what could come up

Put it in slot 3

How to use it

  1. Tap the slots you have already filled, so only the open ones stay coloured.
  2. Slide to where this card sits among everything the list could deal you. A card you like more than three quarters of the list is about 75.
  3. Lock it into the slot shown.

The short version of what it tells you: early on, do not use slot 1 or slot 10 unless the card is extreme. With ten slots open, slot 1 is right only for a card in the top 15.0% of the list.

Where the rule comes from

Blind ranking is close to a known problem in decision mathematics, the sequential assignment problem. Cyrus Derman, Gerald Lieberman and Sheldon Ross solved it in Management Science in 1972: things arrive one at a time, each must be given at once to one of the remaining places, every place has a weight, and the aim is the highest expected total of weight times score. Their result is that the cut-off points depend on how many places are left and how the scores are spread, not on what each place is worth. With n slots open, the scale from worst to best is cut into n bands, and a card in the lowest band takes the worst open slot, the next band the next slot, and so on up.

The rule rests on three assumptions, and it is the proven best only when all three hold:

  • Cards arrive independently. The next card is no more likely to be good or bad because of what came before.
  • Your rating is a position, not a mood. The number from 0 to 100 is where the card sits among everything that could come up: 50 means half the list is better.
  • The score is a weighted total. The rule maximises a total in which better slots count for more. The blind score on this site counts pairs instead, so for that score it is a very good rule rather than a proven best.

Real lists bend the first two a little (once the best thing has gone it cannot come again), so treat the answer as the right instinct rather than a law.

The cut-off points

Worked from the paper's formula for a rating spread evenly from 0 to 100. Read a row as the borders between bands: with 3 slots open, below 37.5 take the worst open slot, above 62.5 take the best, otherwise the middle one.

Slots openCut-off points (rating out of 100)
250.0
337.5 · 62.5
430.5 · 50.0 · 69.5
525.8 · 42.1 · 57.9 · 74.2
622.5 · 36.6 · 50.0 · 63.4 · 77.5
720.0 · 32.4 · 44.2 · 55.8 · 67.6 · 80.0
818.0 · 29.2 · 39.7 · 50.0 · 60.3 · 70.8 · 82.0
916.4 · 26.5 · 36.1 · 45.4 · 54.6 · 63.9 · 73.5 · 83.6
1015.0 · 24.3 · 33.1 · 41.6 · 50.0 · 58.4 · 66.9 · 75.7 · 85.0

How much does it help?

We played 20,000 boards of ten in software, each way, with cards rated at random. The 1972 rule ended with 84.88% of pairs in the right order and 3.4 cards of ten in exactly the right slot. The plain instinct, "a card better than 70% of the list goes about 30% of the way down", scored 83.04% and 3.31.

So the instinct you already have is nearly as good as the mathematics. The gap is the cards near the edges: with the whole board open, the rule sends a card to slot 1 once it beats about 85% of the list, and to slot 10 once about 85% of the list beats it, a little sooner than feels natural. Even with the rule you land only about 3 cards of ten in exactly the right slot. That is the game.

Sources. The result is by C. Derman, G. J. Lieberman and S. M. Ross, "A Sequential Stochastic Assignment Problem", Management Science 18 (7), 1972, pages 349 to 355, cited here from its published record; we have not read its full text. The formula used is the one restated in Theorem 1 of A. Goldenshluger, Y. Malinovsky and A. Zeevi, "A Unified Approach for Solving Sequential Selection Problems", arXiv:1901.04183, read on 21 November 2026. The simulation is ours (seed 7); its code and tests are kept with this site's source.