COURT GUIDE

In a best-game league, the average is the wrong number

Every rankings page in fantasy basketball, ours included, ranks players on what they do on an average night. That is the right number in a category league, where every game counts, and in a classic points league, where the week is the sum of every game played. It is the wrong number in the two formats the hosts now default new leagues to. In High Score, six weekly starters each count their single best game of the week and nothing else. In Lock-In, one game per player counts, the one the manager locks before it tips off. In both, a player who scores 28, 30 and 32 in a three-game week is worth 32, while a player who scores 15, 45 and 20 is worth 45, and the two have the same average.

So this page ranks players on a different number: the expected value of their best game this week, from the games they actually play and the games they actually have.

Read the rest

The two formats, as their hosts describe them

High Score, the default for new leagues on the host most category managers use since the 2025-26 season, is a ten-man roster with six weekly starters, two guards, three frontcourt and a flex, scored on five stats in whole numbers: a point is one, a rebound one, an assist two, a block three and a steal three. Turnovers are gone. Each starter’s single best game of the week counts, and a manager may set the lineup once a week and walk away. Lock-In, the only mode the newest host offers for basketball, counts one game per player per week: the manager locks a game before that game tips off and it counts whatever happens; a player never locked scores his last game of the week. That host publishes which stats its default scoring counts but not the point weights, so the weights this page uses for Lock-In are the six-stat table our sport configuration carries as unverified, and the format note above the board says so. The third toggle, the eleven-stat points table, is the format where every game counts; it is here because a points manager choosing between two streamers also wants to know which one carries the higher ceiling.

Expected best game, and how it is computed

Take a player’s single-game scores under the format, every game on record, and sort them from worst to best. If his club plays n games this week and each one is a random draw from that record, the chance that the best of the n games is no better than his k-th worst game is (k divided by the number of games on record) to the power n, and the chance it is exactly his k-th game is that figure minus the same figure for k minus one. Multiply each game’s score by its chance of being the best and add them up: that is the expected best game, computed exactly from the record rather than from an average and a guess at the spread. With one game it is the plain average. With four it leans hard on the top of the list. The bar under the average on each row shows where the expected best sits between the average and the best game on record, so a wide bar is a player whose ceiling is far above his floor, which is exactly the player these formats reward.

Why the schedule enters, and why it enters twice

A player with four games this week has four draws at his best game; a player with two has two. The expected best of four draws is always higher than the expected best of two from the same record, so the games column is not a footnote here but a term in the number, and a good player on a two-game week can rank below a lesser one on a four-game week. That is the first way the schedule enters. The second is the back-to-back: the second night of one is marked in gold on the nights column, because it is the night a star is most likely to rest, and a rest night is a draw that never happens. The nights themselves, opponent and home or away, are listed on every row, which for a Lock-In manager is the whole decision: the expected best says how much a player is worth this week, the nights say which game to lock, and the tip-off order says when the lock has to be in.

The record, and the two seasons

The distribution comes from the public game-log feed built from the box scores. Once the season is under way a player is judged on this season’s games as soon as he has ten of them; until then, and for the whole of October, he is judged on last season’s, and the row says which in the season column. A player with fewer than five games in either season is not ranked, because five scores are not a distribution and the page would rather leave him off than invent one. Before opening night the feed carries last season only and the band says so. A rookie has no record and does not appear until he has played five games; a player who changed clubs is judged on his own games wherever they were played, which is right for this purpose because the number is about his scoring, not his teammates.

What it does not know

The expected best treats this week’s games as a random handful from the record. It does not know the opponent, the pace, a minutes restriction, a blowout or a coach’s plan; a player on a minutes limit after an injury has a record built at full minutes and will be overrated until the limit shows up in the games. The roster ceiling narrows the board to players rostered in fewer leagues than the slider says, the available view for a manager working the wire, but it is a proxy, not your league’s list. A player ruled out is left off; a day-to-day player stays on with the flag, because in basketball day-to-day usually means he plays. The format, the week, the ceiling and the position are written into the address, so the link you copy opens the same board.

The game logs are rebuilt nightly from the box scores; the schedule, the roster percentages and the injury report are refreshed on our own feeds through the day. If a feed is unavailable the page says so rather than showing a stale board.