21.5. The margin Phoenix covered Dallas by against the spread last night. Final score Phoenix 85, Dallas 70 — Phoenix won outright by 15 against a market line of Dallas minus 6.5. The conviction architecture's three-condition firing was correct by 21.5 points beyond the line.
The Dallas exclusion rule kept us off the play.
That is the honest accounting on the rule's third live application. The rule we instituted May 30 said size down or pass on architecture firings at Dallas home. The first two live applications (SEA at half stake, SEA again under the loosened reading) lost. Tonight's third application would have been the first win — at half stake plus 0.45 units, at full stake plus 1 unit.
The non-Dallas architecture record stays 5-0. The architecture firings at Dallas, including the firings we passed, would have produced a record of 1-2 if we had counted last night (the PHO win plus the prior LVA and SEA losses). The exclusion rule's evidence base does not change in either direction with one data point, but the data point is the kind of evidence the rule's threshold review needs to track.
The structural argument for keeping the exclusion rule unchanged.
Three prior Dallas applications lost — WAS by 23 in May, LVA by 8 in May, SEA at half stake on June 1 by 23 points in our half-stake application. The variance from Bueckers's individual scoring runs was the documented failure mode the rule was built around. The fourth live application going the other direction (PHO covering by 21.5) is one game's outcome. The rule's structural logic on Bueckers variance has not yet been refuted; she did not have her usual 25-plus point game last night.
The structural argument for loosening the rule.
PHO at DAL produced exactly the kind of clean three-condition architecture firing the framework wants. Road favorite at neutral, agent on PHO Q4 HIGH, Dallas Q4 structural weakness at minus 3.0. The model's edge of 14 points was the largest single-game edge any matchup has shown this season. The exclusion rule cost us the bet at any sizing. If the next non-Bueckers-dominant game at Dallas home produces another architecture firing, the rule needs to flip to half-stake-default rather than PASS-default.
The threshold review marker for the rule: if the next two Dallas applications produce one more architecture firing that wins, the rule loosens to half stake. If both lose, the rule stays at PASS. If one wins and one loses, the rule stays unchanged and the threshold question goes to a four-application running window.
That is the same evidence-based framework that produced the under-side rule loosening on June 9. We do not change rules on single-game outcomes. We track the structural pattern across multiple applications.
The other side of last night's framework.
LVA at POR was the architecture's quietest miss of the night. The card had Portland plus 9.8 at STRONG LEAN and the STRONG UNDER 172 with edge minus 8.41. The picks file did not flag this game. The agent was completely silent on quarters. We PASSED per the partial-confirmation discipline and per the loosened UNDER rule (the structural conditions were marginal — Las Vegas thinned, Portland only Day-To-Day, pace condition unclear).
Final score Las Vegas 105, Portland 89. Las Vegas won by 16. Portland plus 9.8 lost by 6.2. The STRONG spread was correctly PASSED at full and half stake equivalents — saved approximately 1 unit.
Game total 194 went 22 points over the under line. The loosened UNDER rule's PASS was correct on the marginal conditions read. Saved approximately 0.5 units against a half-stake bet at the new rule.
Combined LVA at POR PASS: plus 1.5 units variance avoided.
CHI at IND was the third-game data point. The card had MODEST CHI plus 9.5 (PASS by tier) and MODEST UNDER 171.1 (PASS by tier default). Final score Indiana 114, Chicago 106. Chicago plus 9.5 covered (bot would have won at MODEST tier we did not size). Total 220 went 49 points over the under line. The MODEST UNDER PASS was the second saved bet at approximately 0.5 units.
NYL at ATL was a PASS-both card. Final New York 104, Atlanta 90 — both stars Out on both sides, card was right to PASS.
Combined ledger for last night.
Saved: approximately 2 units on LVA at POR spread and the UNDER markets (CHI-IND plus LVA-POR), plus the conservative PASS on the PHO architecture at Dallas was technically a missed plus 0.45 units of profit.
Net for the night: roughly plus 1.55 units of variance saved on the framework's PASS decisions, minus 0.45 units of missed profit on the PHO architecture. The cumulative discipline saved plus or minus 1 unit night on a slate where the bot would have gone 1-2 on STRONG-tier spreads and 0-3 on STRONG-tier totals.
The discipline continues to be net positive across multiple weeks. The Dallas exclusion rule's threshold question is the active framework-maintenance task.
Tonight.
Two games on the WNBA slate. The bot's picks file has one HIGH-confidence call.
TOR at WAS. Bot has Toronto at HIGH spread tier with edge plus 7.45. Card has STRONG LEAN on Toronto minus 2.3 with edge plus 7.16. Agent has TOR Q1 medium plus 1.60, Q3 medium plus 2.00, Q4 medium plus 2.10 — three medium signals on Toronto's side with a "Close game, TOR closes" SGP correlation.
The conviction architecture's three conditions are met at a non-Dallas home court. Road team favored at neutral (both models agree). Agent reads road team winning late quarters (medium tier, not HIGH). Home team has Q4 weakness (WAS at minus 0.3, the softer version of the structural weakness the rule wants).
This is partial architecture firing rather than the cleaner HIGH-tier alignments that produced the 5-0 non-Dallas record across May. The agent's medium-not-HIGH late-quarter reads are the rule's filtration mechanism.
Half stake on TOR minus 2.3 per the partial-confirmation rule. Not full because the agent's confirmation is medium tier. Not PASS because three of three architectural conditions are present at non-Dallas with the bot's HIGH-conviction call and the largest single spread edge of the slate.
Nina has the tactical piece on the specific case for and case against full sizing. Jordan has the Tempo Report on the Sabally Day-To-Day complication and the 7-6 road test.
GSV at SEA. Card has STRONG LEAN on Seattle plus 9.1 (home dog). Picks file does not have this game in HIGH tier. Agent has GSV Q1 plus 5.40 HIGH and Q2 plus 6.00 HIGH on Golden State's side — two HIGH-confidence quarter signals on the opposite side from the card's STRONG LEAN. The formal model-versus-agent conflict rule requires three-plus HIGH signals on the opposite side; tonight has only two. The rule does not formally fire.
But the threshold review evidence accumulates here too. Two HIGH agent signals on the opposite side from a card STRONG LEAN, with no SGP correlation, is the configuration the 2-HIGH-plus-SGP threshold variant would require. Tonight has the two HIGH without the SGP. The rule's strict 3-HIGH-plus reading produces a PASS by inaction, which is the correct framework call regardless of whether the rule formally fires.
PASS the GSV at SEA spread. PASS the total (card had it at PASS already with edge plus 0.54).
The verdict tonight.
One bet sized. Half stake on TOR minus 2.3. Roughly 0.5 units of exposure.
The cumulative ledger.
Spread STRONG editorial-sized: 12-13 entering tonight. The TOR half stake will produce the first sized bet in fifteen days. The silent-agent partial-confirmation PASS streak saved approximately 5 units of variance across the discipline's correct PASSes over two-plus weeks.
Total STRONG editorial-sized: 9-9. No bets sized since May 28. The loosened UNDER rule has not produced a sized bet because the structural conditions have not met the "cleanly met" standard since the rule changed.
The architecture's standing.
Non-Dallas applications: 5-0 across May. Tonight's softer firing at WAS is the first non-Dallas test since the early-month sample.
Dallas applications: 2-3 at the rule level. The PHO miss yesterday brings the unofficial Dallas record to 3-3 if we counted the games we passed. The exclusion rule's threshold review is the active discipline question.
Conflict rule: 4-3 going in. Tonight's potential application on GSV at SEA does not formally fire because the threshold requires three-plus HIGH agent signals. The two HIGH plus zero SGP configuration tonight produces a PASS by tier reasoning regardless.
Talk tomorrow.
[ End Report ]
Share This