3. The units of missed profit the framework's PASS-by-default rule has produced over the last five days of slates. Five consecutive slates with STRONG LEAN card signals we passed per the rule, and every one of them hit the wrong direction from PASS.
The tally.
Friday July 4. GSV plus 11.25 at Atlanta — bot HIGH, we did not size in editorial. GSV won outright 88-83, covered by 16.25 points. Missed 0.45 units. Sunday July 5. IND plus 5.5 at LVA — STRONG LEAN card, non-HIGH bot pick. IND won outright 84-68, covered by 15.5. STRONG UNDER 183.5 same game, cleared by 31 points. Both passed. Missed approximately 0.9 units at half stake per bet. Monday July 6. CON at MIN — STRONG LEAN card, non-HIGH bot pick, CON plus 13.9. CON won outright 90-89. Covered by 14.9. Missed 0.45 units. Same slate: SEA at LAS STRONG LEAN SEA plus 3.6, plus a STRONG UNDER 174.6. SEA won by 18 and total cleared by 28.6. Both passed. Missed approximately 0.9 units. Tuesday July 7. DAL plus 5.25 at NYL — bot HIGH pick, we did not size in editorial. DAL won outright 88-77, covered by 16.25. STRONG UNDER 172.8 CHI at PHO cleared by 30 points. STRONG UNDER 175.1 DAL at NYL cleared by 10. Three signals, all winners, we passed. Missed approximately 1.35 units.
Cumulative missed profit over the five-day window: approximately 3.6 units.
The framework's job is to size real edge and pass on marginal edge. What has happened over the last five days is not marginal — the misses are large-margin covers on STRONG LEAN cards with meaningful edges. The PASS-by-default rule on non-HIGH cards was calibrated when non-HIGH STRONG LEAN cards were performing near the STRONG tier baseline of 50 percent. Recent observations suggest that specific subset (STRONG LEAN spread cards where the bot did not publish HIGH) is performing at closer to 60-65 percent, which is well above the threshold that justifies exposure.
The recalibration.
Effective tonight, the framework adds a rule for STRONG LEAN card signals with edge greater than 5 points where the bot did not publish HIGH. When either of those conditions is met — the card labels the pick STRONG LEAN and the edge on the arc card is 5-plus points — the pick gets quarter-stake exposure. Not half stake, because the tier-tracking data still suggests HIGH publication is a meaningful additional filter. But quarter stake produces exposure at a size that captures some of the missed edge without giving the framework any single-night exposure larger than one unit total.
Similarly for STRONG UNDER signals with edge greater than 8 points where the bot did not publish TOTAL as HIGH. Quarter stake by default. The 3-point OVER agent pause threshold from Issue 40 still applies.
These are provisional rules pending 15-observation-window evaluation. If quarter stake on non-HIGH STRONG LEAN cards produces positive expected value over the next two weeks, the sizing gets bumped to half stake by rule. If it produces flat or negative results, the rule gets tightened back up.
The cumulative ledger per the grader through Tuesday.
Spread STRONG (all bot-labeled STRONG LEAN cards): 36-33 (52.2 percent, minus 0.40 percent ROI). Best position of the season, up three games this week alone from the wins the framework passed on. Total STRONG: 36-36 (50 percent, minus 4.55 percent ROI). Back to breakeven after multiple weeks below. MODEST spread: 21-22 (48.8 percent). MODEST UNDER: 23-24 (48.9 percent).
The bot's model is producing an edge. The framework has been sizing that edge too conservatively. Tonight's rule change starts capturing more of the edge.
Tonight is three games. Zero HIGH-confidence picks published. The bot's card generation has not produced the July 8 arc cards yet, so the framework cannot cleanly evaluate the spread edges via the STRONG LEAN card labels. The recalibration rule requires the arc card to fire. Without arc cards, we fall back to the strict default: PASS unless the picks file publishes HIGH.
The agent reports for tonight.
Connecticut at Minnesota. Agent has all four MIN quarters HIGH (Q1 plus 3.50, Q2 plus 3.80, Q3 plus 2.80, Q4 plus 5.80). Same dominance pattern as Monday's CON at MIN matchup that CON covered outright. The agent's directional read was wrong Monday. Tonight the same pattern appears in the same matchup, same direction. Two possibilities: the agent read is right this time and Monday was variance, or the agent's calibration is systematically wrong about the MIN home matchup and tonight's read is also wrong.
Without the arc card, I cannot see the current market spread and cannot evaluate whether the model math agrees with the agent. Per the strict default, PASS.
Las Vegas at Indiana. Agent has IND Q1 plus 5.80 HIGH, Q2 plus 3.70 HIGH, Q3 plus 2.60 HIGH. Three HIGH signals for Indiana at home. Caitlin Clark is Out per the picks file's injury index. Indiana without Clark against Las Vegas without Wilson (Wilson is confirmed Out tonight after Thursday's D2D) is a heavy-injury matchup on both sides. Projected total 178.8.
Without the arc card, PASS the spread per default.
Toronto at Golden State. Agent has GSV Q2 plus 6.90 HIGH, plus one MEDIUM signal. One HIGH agent signal — not a dominance pattern. Jordan's Tempo Report goes deeper on the matchup this afternoon.
Without the arc card, PASS the spread per default.
Tonight's recommended action.
PASS: All three games. Framework recalibration is effective tonight but requires arc card data to fire quarter-stake exposure. Without cards, the default holds.
The recalibration is on the record. The next slate that produces STRONG LEAN cards will be the first live application of the quarter-stake rule for non-HIGH cards with 5-plus edge. If the arc cards for tonight get published later today, we may catch a quarter-stake exposure retroactively (framework rules do not fire retroactively for editorial purposes but the framework tracking will note whether we would have covered).
Tempo returns from the 11-day layoff at Golden State. Mabrey is off the injury list. Sykes is still Out. Jordan has the Tempo Report with the road-identity read.
Talk tomorrow morning. The framework recalibration begins.
[ End Report ]
Share This