23. Dallas's margin of victory over Washington last night. The Wings won 92-69 in a game that was effectively decided by halftime. The Mystics shot poorly, did not get to the line, and never executed the half-court offense Nina wrote about Monday. The conviction architecture took its first loss in six applications.
The ledger.
PHO at half stake the previous Saturday — WIN. ATL at full stake the next Tuesday — WIN. MIN at full stake the next day — WIN. TOR at full stake the next Friday — WIN. SEA at full stake last Sunday — WIN. WAS at full stake last night — LOSS.
Five for six. Eighty-three percent over 11 days.
The honest read on what went wrong.
The architectural shape was there. Dallas had the Q4 problem we wrote about. Washington was the model's HIGH-confidence pick. Edge was plus 7.7. Nina's tactical preview said the structural advantages were the same as the wins. Same shape, different result.
What did not go to plan. Paige Bueckers had her best game as a pro. She scored from the perimeter and got to the line, and the Wings did not have to depend on Ogunbowale iso. When Dallas's offense has two creators clicking at the same time, the Q4 weakness becomes irrelevant because they are up by 20 going into the fourth and Brondello-style closing structure does not matter. The Mystics also could not buy a shot. Stephanie White's offense generated decent looks per the broadcast cameras and the shooters missed open threes. Variance plus a Bueckers breakout was the loss combination.
Two ways to think about this. Same framing as the POR pass two weeks ago.
The pessimist read. The architecture is overfit to small samples. Five wins do not mean an edge. Last night was the regression.
The optimist read. Eighty-three percent over six samples is well above breakeven. The architecture is real. The one loss was on a night Dallas's two best players had simultaneous good games and Washington's shooting reverted to mean. The model still says the architecture is the highest-confidence pattern available. The framework holds.
I lean optimist this morning. Two specific reasons.
First, the variance check. Pre-loss, the architecture was running at 5-for-5 (100 percent). The true probability of the conviction architecture is somewhere between 60 and 75 percent based on prior multi-season data Maya and I analyzed in March. A 5-for-5 stretch in a 60 to 75 percent system has a base rate of about 7 to 24 percent per six-game stretch. Hitting one out of every 4 to 14 stretches. Five in a row was on the optimistic end of the distribution. Six in a row would have been even more optimistic. The expected-value math always had a loss coming.
Second, the architectural confirmation matters. Last night WAS at DAL had all three confirmations: model has road team favored, agent has road team winning the late quarters (WAS Q4 plus 4.0 HIGH), home team has structural Q4 weakness. Three of three. The loss happened anyway. That is variance, not a framework break. If we had two of three (less architectural confidence) and the result was the same, we would still be 4-1 with a smaller-stakes bet. The framework's job is to size correctly. Last night we sized correctly and lost. That is part of the game.
The action this morning is twofold. One, do not change the framework based on one loss. Two, do change the sizing on tonight's call because tonight's call is NOT a full conviction.
Tonight is small. One game on the board.
Toronto at Phoenix. STRONG LEAN TOR plus 7.5 per the card. Model edge plus 7.81. Sounds like a conviction call.
It is not. Here is why.
The architectural framework has three confirmations. (1) Model has road team favored at neutral. (2) Agent has road team winning the most-relevant late quarters. (3) Home team has structural Q4 weakness.
Tonight's agent report on TOR at PHO has zero quarter edges. Zero prop edges. Zero SGP correlations. The HIGH-confidence plays count is 0. The only thing the framework has is condition (1) — the model edge on the spread. Conditions (2) and (3) are not confirmed because the agent has not flagged Phoenix as a structural Q4 problem and has not given us a multi-quarter edge favoring Toronto.
One of three is not the conviction architecture. One of three is a model lean.
Half stake on TOR plus 7.5. Per the discipline framework. The same way we sized PHO at half stake the first time we wrote about the architecture two weekends ago because only one of three confirmations was present.
Phoenix is also not the Sparks. The Mercury are a top-half team this year with Alyssa Thomas defending the rim, Kahleah Copper closing games, and Satou Sabally creating offense. Jordan's piece this afternoon will go deeper on the matchup. The summary is: Toronto is a better road team than I would have expected three weeks ago, but Phoenix is a real team and the architecture is not screaming the way it has on the wins.
The other thing tonight. The total. MODEST UNDER 170.5, edge minus 7.55. Skip per the discipline tier downgrade on MODEST signals. Totals model is now 8-6 and trending right but we do not chase MODEST signals when the architecture does not confirm.
Tonight's recommended action.
HALF STAKE: TOR plus 7.5 at Phoenix. Model lean, not conviction. Architectural framework partially confirmed.
PASS: Total. MODEST signal, no structural confirmation.
The previewable thing for tomorrow.
POR plus 13.5 at IND. The model has Portland as a 4.9 favorite at neutral. Edge plus 8.6. HIGH-confidence published pick. This is the kind of architectural setup that has been working. The Fever just got blown out by Seattle Sunday. Portland is a slow team that walks the ball up and grinds out games. The architecture has both (1) the model spread edge and a likely (2) the agent quarter edges (the agent report for tomorrow has not been generated yet at the time of writing) and possibly (3) Indiana's Q4 structure question after the Caitlin Clark fatigue read.
If tomorrow's agent confirms the architectural conditions, POR plus 13.5 will be a full-stake conviction call and we will write the bankroll piece in the morning. Possibly the seventh application of the framework.
The cumulative ledger through May 18.
Spread STRONG: 9-5 (64 percent). One loss last night, framework still well above breakeven. Total STRONG: 8-6 (57 percent). Up one from the CON at POR under cleared by 9. Combined STRONG: 17-11 (61 percent). The model is profitable on both tiers.
The architecture moved from 5-for-5 to 5-for-6. Eighty-three percent over six applications. That is still the highest-confidence pattern available. Tonight we wait. Tomorrow we size up.
[ End Report ]
Share This