2.5. The Chicago Sky's cover margin at Las Vegas Thursday night. Final LVA 98, CHI 90. Wilson played through the leg — she was Day-To-Day pre-game, ultimately available. Vegas won the game by 8. CHI plus 10.5 covered by 2.5.
The bot's HIGH pick hit at half stake. Plus 0.45 units.
The UNDER on the same game did not. Game went 188. UNDER 182.3 cleared the OVER by 5.7. Half stake loss of 0.5 units.
Net Thursday: minus 0.05 units. Basically flat over the two correlated bets. The framework read on the game shape (CHI covers in a low-scoring game with Wilson's minutes limited) was directionally right on the spread but the pace of the game held higher than the agent had projected.
That was the last time we covered a slate editorially. The three days since then have produced two missed wins on non-HIGH picks the framework's default rule passed on.
Friday July 4. GSV plus 11.25 at Atlanta — bot HIGH pick, edge plus 11.25. No demote flags. Golden State won outright 88-83. GSV plus 11.25 covered by 16.25. At half stake per the default rule, this was worth plus 0.45 units. We did not cover it in editorial because I did not read the picks file that morning. Missed win.
Sunday July 5. IND plus 5.5 at Las Vegas — the card had it as a STRONG LEAN with a real edge, but the picks file had zero HIGH-confidence calls (n_picks equals 0). Framework default says non-HIGH cards stay PASS. Indiana won outright 84-68 as a 5.5-point road dog. Plus 10.5 cover margin. The STRONG UNDER 183.5 on the same game hit by 31 points. Both signals were the strongest of the season and we passed both per default.
Total missed profit over Friday and Sunday: approximately 1.35 units at what should have been half-stake exposure.
The framework question this creates.
The default rule from Issue 42 says only bet HIGH picks at half stake by default; non-HIGH cards stay PASS. The rule has been the right call over the sample (spread STRONG on non-HIGH cards is roughly break even; the HIGH subset is where the edge lives). But recent non-HIGH STRONG LEAN cards have hit at meaningfully higher rates than the framework's PASS-by-default discipline captures. Sunday's IND cover was a large-margin win on a card the bot did not elevate but the model math was still reading as significant.
The MODEST UNDER tier tells the same story. The last two weeks have produced MODEST UNDER hit rates that suggest the June 9 loosening of the STRONG UNDER rule may not have been generous enough — the MODEST UNDERS need their own loosening consideration. Per the grader through Sunday, MODEST UNDERS are 22-24 (47.8 percent), which is below breakeven, but the more recent 15-observation window is roughly 62 percent. That is not a sample size that supports a rule change yet. Another two weeks of data will tell.
The cumulative ledger per the grader.
Spread STRONG (all bot-labeled STRONG LEAN cards): 34-32 (51.5 percent, minus 1.65 percent ROI). Up one game after CHI covered. Total STRONG: 33-36 (47.8 percent, minus 8.7 percent ROI). Down three games this stretch after multiple missed unders got booked as losses against the STRONG ledger while the winning MODEST UNDERs got booked separately. MODEST spread: 21-21 (50 percent). MODEST UNDER: 22-24 (47.8 percent).
The system is essentially flat on spreads and materially negative on totals. The rule loosening window continues.
Tonight is three games. Zero HIGH-confidence picks published (n_picks equals 0). The bot's card generation has not yet produced the July 6 cards as of the time of writing, so the framework read is based on the picks file and the agent reports only.
Minnesota at Connecticut. Agent has four HIGH-confidence quarter signals — MIN Q1 plus 4.00, MIN Q2 plus 4.20, MIN Q3 plus 3.00, MIN Q4 plus 5.90. The SGP correlation reads "MIN dominant game." Projected quarter totals add to 169.6, well below any conservative total line the market would produce for a Lynx home matchup. If the arc card had been generated with a spread line, this would fit the conviction architecture profile — assuming MIN is the road team with the model favoring them and CON is the home team with structural weakness. Napheesa Collier is Out per the picks file's injuries index, which changes the MIN calibration meaningfully.
Without the arc card, the framework cannot cleanly evaluate the spread edge. If the market has MIN as a road favorite by less than 4-5 points (which is where the model math per the agent would trend), the conviction architecture arguably fires. If the market has CON favored, the pattern breaks.
Per the discipline framework's strict reading — no arc card, no bot HIGH pick — PASS the spread. The agent's four HIGH signals are meaningful and I want to flag them, but sizing exposure on an incomplete framework check has been the source of most losses this season.
Washington at Golden State. Agent has GSV Q2 plus 7.50 HIGH and WAS Q3 plus 3.40 HIGH — split HIGH signals across the two teams. The largest single-quarter edge on the slate (GSV Q2 +7.50) but not a dominant pattern. Projected total 162.7. PASS per no-HIGH-published default.
Las Vegas at Seattle. Weak agent read, no HIGH signals. Projected total 167.5. PASS per default.
Tonight's recommended action.
PASS: All three games spreads and totals.
Zero exposure. The framework's discipline holds through what may be another slate of missed wins if the agent's MIN read materializes. That is the discomfort the framework's default rule is meant to produce until the loosening consideration produces a rule change.
Tempo has not played since June 27 (the Phoenix home loss). Nine days off. The team's schedule per the bot's card generation status has no upcoming Tempo game visible, which suggests either a scheduled break or a card generation lag. The picks file's injuries index for TOR shows only Brittney Sykes listed — Marina Mabrey is no longer on the injury list, suggesting she has cleared. That is franchise-significant news. When the next Tempo card is generated, Jordan will have the Tempo Report addressing Mabrey's return status and the team's second-half positioning.
Talk tomorrow morning.
[ End Report ]
Share This