4. The margin the LVA at NYL UNDER 173.9 cleared the wrong direction Monday. Final NYL 93, LVA 85. Game went 178. The agent's projected quarter totals had it at 177.8 pre-game. The projection was correct to within half a point.
We sized half stake on the under despite the agent's 3.3-point OVER projection sitting right at the rule's pause threshold from Issue 40. The rule maintenance update said "if the agent projects OVER by 3-plus points, the loosened UNDER rule pauses." Monday I framed the projection as being "right on the threshold" and half staked anyway on the injury-structure argument.
The framework lesson to encode. When the agent's projected total is within a point of the pause threshold, the rule pauses. Do not half stake through the caveat. The projection landed within 0.2 points of the market total; the injury adjustment I argued would trim the projection by 5-8 points did not materialize because both teams' scoring inputs held roughly at their per-game averages despite the absences.
Cumulative through Monday's slate.
Spread STRONG (all bot-labeled STRONG LEAN cards per the grader): 33-31 (51.6 percent, minus 1.56 percent ROI). Unchanged from yesterday. Total STRONG: 32-34 (48.5 percent, minus 7.44 percent ROI). Down one game. The totals side has been the framework's persistent bleed. 66 STRONG UNDER cards through the season and the tier is below breakeven with the ROI dragging.
The framework recalibration window from Issue 42 continues. The totals rule loosening from June 9 has produced one recent loss (Monday's UNDER 173.9) and no wins. The rule needs 15-20 observations to evaluate honestly; we are at maybe 8 since the loosening. Continue per spec.
Tonight is three games. Zero HIGH-confidence picks published.
Atlanta at Washington. Card: MODEST LEAN ATL minus 5.8, edge plus 4.05. MODEST UNDER 165.9, edge minus 6.41. Both signals are in MODEST tier. Pass per tier downgrade.
The agent's read is interesting. ATL has two HIGH quarter signals aligned with the model's ATL pick direction (Q2 plus 3.30, Q4 plus 6.60). But the agent also has HIGH UNDER props on both Rhyne Howard and Allisha Gray — the Dream's two primary scorers. That is the contradictory signal profile: the agent says Atlanta wins big in Q2 and Q4 while the two players who would normally drive those quarters both go UNDER their point totals. Either the read is that Atlanta's role players carry the offensive load, or the read is internally inconsistent and the agent's per-quarter math is not fully accounting for the individual usage adjustments.
Bot did not elevate the pick to HIGH despite the aligned agent signals, so we do not size. If the pick had been elevated to HIGH the aligned agent quarters would have called for a full-stake read; the fact that it stayed MODEST suggests the bot's own uncertainty calibration on this matchup is not clean. PASS.
Dallas at Connecticut. Card labels are confusing (MODEST LEAN CON plus 8.8 with a negative edge, which I read as the model actually preferring DAL to cover the CON minus 8.8 line — the negative edge sign means the card model disagrees with the labeled pick). Model spread has DAL favored by 4.6 at neutral. Market has CON favored by 8.8 at home. The model thinks DAL covers.
Agent aligns with the model — DAL Q3 plus 5.00 HIGH, DAL Q4 plus 4.00 HIGH. Two HIGH signals favoring DAL. If this had been a HIGH-confidence bot call, it would be a half-stake bet with aligned architecture. The bot did not elevate. PASS per the framework's no-HIGH-no-bet default.
Seattle at Phoenix. Card: STRONG LEAN SEA plus 5.5, edge plus 7.1. The only STRONG-tier spread on the slate. Model has SEA favored by 1.6 at neutral. Market has PHO minus 5.5 at home.
Framework check. Non-HIGH bot pick despite STRONG LEAN card label. That means the bot's own confidence calibration lowered the pick out of the HIGH tier. Likely because Seattle is missing three rotation pieces (Horston, Mair, Magbegor) and the model adjustment for the absences eats the raw spread edge.
Conviction architecture check. (1) Model has road team favored at neutral — yes, SEA plus 1.6 at neutral. (2) Agent has road team winning late quarters — no. Agent has PHO with a MEDIUM Q1 edge and a MEDIUM Q3 edge; no SEA HIGH signals. Architecture fails. (3) Home team structural Q4 weakness — no. Phoenix has been solid in late quarters through June.
One of three architectural conditions. PASS per the framework — non-HIGH cards do not get half-stake defaults; only architecture fires get full stake.
Tonight's recommended action.
PASS: All three games.
Zero exposure. The framework's discipline holds through a slate that would have been half-stake exposure earlier in the season under looser rules.
Tempo has no game tonight or tomorrow per the schedule data I have. The next Tempo game per the bot's card generation status looks like Sunday at earliest, though the schedule may shift as the league finalizes its post-Fourth-of-July slate. Jordan will have the Tempo Report when the next Tempo card lands with cards for me to read against.
The Mabrey timeline is the franchise question. Yesterday's TOR injury list per the picks file's injuries_index had Mabrey listed alongside Sykes. Both leading scorers on the injury report is the kind of situation that reshapes the second-half projection. If both are back by Sunday, the team's offensive ceiling returns. If either misses extended time, the 11-8 team we have watched through June is going to look different in July.
No editorial action tonight. Bankroll stays.
[ End Report ]
Share This