Each rule was frozen before its first game. Every pick is graded against the closing line.
Lab data unavailable.
Beating the close means getting a better price than the final market price before the game. It is the best early sign a rule has an edge. What's the close?
The race · every graded pick is a dotWho beats the close
One pick that got a better price than the closeOne pick that got a worse priceThe rule's averageAvg needed to make money
Most rules beat the close by a little. Making money takes about +2 a pick on average, the dashed line.
Outside builders and cappers · graded on DraftKings closeBuilders and cappers racing next to our lab
Not racing for the Standard this season.
Everyone is graded on DraftKings prices. Anyone who bets exchanges or other books may get better prices than these grades show.
An outside builder or capper posts picks in public before puck drop. We grade them exactly as we grade our own rules: the DraftKings price when the pick was posted against the DraftKings close on the same line, raw prices, no vig removed. A tie is a price that did not move. A total at a line DraftKings never offered is counted as not graded and says so. Both sides of one game, or picks past a builder's daily cap, are void: counted, shown, never graded. No percentage or average is shown until 25 picks are graded. The won-lost record is for reference; the bar is the close. The builder's own quoted price is recorded, never graded.
Head to head · same game, different sidesThey disagree
Every pick is a frozen rule's decision on a recorded DraftKings price, written down before the start and graded against the closing price on the same line. Picks marked posted went to Discord before the start; earlier picks were recorded privately while the lab was being built. How far a pick beat the close is in points: positive means the price moved toward the pick after it was recorded. A rule with fewer than 10 graded picks is shown faded: too few picks to judge. A rule with no pick yet is marked that way. Lab numbers are computed by the exporter and never enter the Standard's counts or any model's record. Amended closes were missed by a name-matching bug and corrected on Sep 28; the original records are kept.