Week 1, Settled

Why those two numbers differ

48–47 across the full board is a coin flip. +8.22u on 102.5 risked is not what a coin flip pays — and that gap is the entire point of the machine. Every pick carries a price and a stake, the board leans toward plus‑money, and it sizes up where our number sits furthest from the market’s. Land the bigger, longer‑priced ones and a near‑even record finishes comfortably ahead. That is exactly what happened.

The featured slate is the same machinery aimed at 18 picks instead of 98, and this week it aimed badly in one specific place. We treat that as a tuning problem with a known cause, not a mystery — and we walk through it in full below rather than leaving it to a footnote.

Of 16 games, the board played 7, called no edge — stay away on 6, and skipped 3 more purely on price. More than half the slate got a pass. That discipline is a feature you are paying for: most services find a play in every game because a full card sells better than an honest one.

The tier ladder did what it is built to do

Every pick is tiered by how far our number sits from the market’s, and the tier sets the stake. Week 1 was the first live test of whether that ordering carries real information. It does — and the separation was clean.

Strong edge cleared its own break‑even line by 9.4 points and returned +8.40u on 54 risked. Edge landed within 1.5 points of its number. The higher tier carried the week, the lower one held its ground, and they finished in the order the model said they would — which is the single most encouraging thing in this report. The bars are wide because the record is four days old; they tighten fast from here.

Where the +8.22u came from

Five markets became published plays. This is the week’s return broken out across them — not five separate claims about skill, just the arithmetic of where a published total came from.

Percentages arrive as each market matures. We publish a hit rate once a slice reaches 30 picks and a return‑on‑risk figure at 200, so this page gains detail every week rather than losing it. Receptions already clears the first bar and hit 50.7% against the 48.8% it needed. The rest will follow as the season fills them in.

Voids are picks that never truly settled — a player who did not dress, or a quarterback who left injured in the first half. The stake comes back and the pick is excluded in both directions, which is why the three columns are shown separately rather than folded into a tidier number.

That rule runs both ways. Kyler Murray and Sam Darnold both went off hurt in the first half, so their props void. On this slate that helped us — both were picks we were losing. But Murray also carried five props on our internal board, all unders, all of which landed precisely because he left early, and the same rule voids those too. Applied honestly it took 3.3 units off our internal number. A rule you only invoke when it suits you is not a rule.

Receptions is the engine: 68 of the 98 picks, 66 of the 102.5 units risked, and +7.67u returned. Passing TDs added +5.87u off a light 6.5‑unit allocation. The one genuine miss was anytime touchdown at 3–7, and it is the piece of the board getting the most attention this week.

The one thing we got wrong, and what happens next

The featured slate went 6–10 and returned −3.80u on 29.5 units risked, while the 79 picks we did not highlight went 42–37 for +12.01u. We would rather hand you that number plainly than let you find it yourself.

The useful part is where it went wrong. Five featured anytime‑touchdown picks went 1–4 and cost −5.12u. The other thirteen featured picks returned +1.33u between them. Strip out that single market and the highlighted slate finishes the week ahead. This is not a model that lost its way — it is one component sized too aggressively, and that is a fixable thing.

Update, Week 3. Three changes. The highlighted card is again open to every market we can measure, ranked by how far our probability sits from the market’s, with no single market allowed more than half the card. Edge‑tier picks drop to a flat half unit: they are 49–62 through two weeks, and a full unit was not earned. Expert agreement no longer moves a pick up or down the ranking; what the beat coverage says now sits inside each pick’s See why. Separately, we found the probability model had been squeezing most picks onto two or three identical values and re‑fitted it, so probabilities and tiers will look more spread out from this week.

Worth keeping in view: every pick on the board is published with its own tier, price and stake before kickoff. The highlight is a shortcut through the board, not the board itself — and all 98 picks below were available to you, including the 79 that carried the week.

How the week was shaped across the days

The return was not spread evenly. The Wednesday and Thursday openers — 14 picks risking 14 units — returned +6.85u between them, a standout start to the season. Sunday, carrying 76 of the graded picks and 79 of the units, came in at +1.69u. Monday’s six picks finished at −0.32u.

Fourteen picks driving most of a week’s return is a concentrated result, and we point at it rather than let it sit quietly inside the total. Sunday is the slate with real volume behind it, and Sunday landed right on its own number — 37–39 against the 48.7% those prices required. A big slate holding its line while the small ones outperform is a fine way to start.

Every pick, all 98

Every pick, ordered by game day, at the stake printed on the card — 0.5 to 2.5 units, set by the tier. Nothing here is selected after the fact: this is the same list that was live before kickoff. A void is a pick that never truly settled — the player did not dress, or left injured before the game was decided — and it returns the stake.

Swipe the table sideways for price, result and units.