Minus one point two five units on Sunday. Plus zero point four five units on Monday. Tuesday off. Net minus zero point eight zero units across three days.
The Sunday recap is the important one because it validates a framework rule change I have been reluctant to formalize.
CON at DAL was the pick I flagged in the Sunday piece as fitting the 2-HIGH-late-quarter agent conflict pattern. The picks file had Connecticut as a HIGH pick at plus 6.75 on the spread. The agent had two HIGH-tier quarter edges on the Dallas side, both late-quarter — DAL Q3 plus 5.00 and DAL Q4 plus 4.70. No SGP idea to trigger my proposed 2-HIGH-plus-SGP rule, so the framework sized at half stake per the formal Issue 42 default with the pattern flagged in the write-up.
Dallas won 83-63. A 20-point margin. Connecticut was blown out by exactly the second-half dominance the agent projected.
That is now three separate times this season the agent's late-quarter dominance signal has been right against a picks-file HIGH pick on the opposite side. CON at IND on July 22 (Indiana blowout by 35). Now CON at DAL on August 2 (Dallas blowout by 20). And an earlier instance in June whose exact game I would need to look up but which followed the same pattern.
Three-for-three on the pattern is enough sample. The rule is now formal:
When the picks file lists a HIGH pick on the spread AND the 2025 agent has TWO HIGH-tier quarter edges on the opposite team AND both of those edges are in the second half (Q3 and Q4), the framework PASSes the pick regardless of the picks file edge magnitude.
That is more granular than the original 3-HIGH-conflict PASS rule. The intuition is that the 2025 agent's late-quarter dominance signal is a structurally different read than early-game quarter edges — it captures which team is likely to close the game, which is where blowout margins get built. When the picks file has one read on a game and the agent has late-quarter dominance on the opposite side, the closing quarters have been where the disagreement resolves in favor of the agent.
The original 3-HIGH conflict rule stays in place. This is an additive tighter trigger for the specific late-quarter case.
The two other Sunday losses were less pattern-clear.
IND at MIN spread IND HIGH — loss. Minnesota won 108-100 as a 5.8-point home favorite, IND lost outright by 8 as a plus-5.8 dog. Minnesota outscored Indiana across the second half decisively. No agent conflict trigger to PASS by. Just a losing HIGH pick.
IND at MIN total UNDER HIGH — loss. Game landed at 208 against a 193.2 line. Over by 15. Model had projected 179.85. Agent had projected 187.6. Both were well under. The game landed 20 points higher than the model expected. The quarter-stake sizing meant the loss was minus 0.25u instead of minus 0.5u — this is exactly what the totals-HIGH downgrade rule is designed to prevent.
Monday.
LVA at ATL spread LVA HIGH edge plus 6.05 — win. Vegas won 109-87 on the road as a 6-point favorite, covered easily. Half stake, plus 0.45u.
Tuesday.
One game on the board. No HIGH picks. Nothing to size.
Cumulative check per grader (through pending Wednesday grades):
Spread STRONG LEAN: 53-43-1 (55.2 percent, plus 5.34 percent ROI). Unchanged from last update — no spread STRONG cards graded in the interval.
Spread MODEST LEAN: 23-29 (44.2 percent, minus 15.56 percent ROI). Unchanged.
Total STRONG LEAN: 46-67 (40.7 percent, minus 22.28 percent ROI). Down another 0.7 percentage points. One new losing total STRONG card graded.
Total MODEST LEAN: 28-28 (50.0 percent, minus 4.55 percent ROI). Unchanged.
The totals STRONG ROI at minus 22.28 percent is approaching the minus 25 percent trigger where I flagged the next rule escalation (totals HIGH from quarter stake to PASS entirely). Not there yet, but close. The quarter-stake sizing has been correctly reducing loss size — the IND-MIN UNDER Sunday was a 0.25u loss instead of 0.5u — which is exactly the intended behavior.
Tonight (Wednesday August 5). Four games on the board, two HIGH picks.
Pick one: SEA at NYL total UNDER 183.4 HIGH edge minus 10.31.
Spread pick is SEA MEDIUM — PASS per formal rule.
Total UNDER HIGH. Sabally Out for New York (again). Agent projected total 171.3 (43.7 plus 41.9 plus 43.4 plus 42.3). Model 173.09. Both well under the market's 183.4. Full alignment. No agent conflict on the total side.
Sizing: QUARTER stake UNDER per totals downgrade rule. 0.25u.
Pick two: LAS at CHI spread LAS HIGH edge plus 5.25.
Diggins Out for Chicago. Injury edge component plus 2.0 but not injury-dominant. Model has LAS by 2.35, market CHI by 2.9. That is a 5.25-point spread edge favoring Los Angeles.
Agent read: only one MEDIUM-tier quarter edge (CHI Q1 plus 1.50). Zero HIGH agent edges either direction. No conflict trigger fires.
Sizing: HALF stake LAS per Issue 42 default. 0.5u.
Total exposure tonight: 0.75u across two bets on two games.
The Tempo. Not playing tonight. Jordan has the Tempo mid-crisis piece separately — Toronto has lost two straight badly (32 at home to Minnesota, 17 on the road to Golden State) and is now 14-13 with a real playoff-bubble concern. Next Tempo game Friday.
Framework state.
The formal rule change tonight is the 2-HIGH-late-quarter-agent-conflict PASS trigger. That is one more tighter filter on top of the existing rule stack. The next rule tune, if the totals STRONG tier keeps declining, is totals HIGH moving from quarter stake to PASS at ROI minus 25 percent or lower.
The three-day net at minus 0.80u is not great but not season-derailing. The Sunday bloodbath was the story; Monday's clean HIGH win partially offset it. The framework holds discipline and the rule stack keeps evolving as the data supports each change.
Back tomorrow with grades.
[ End Report ]
Share This