13. The line on Connecticut tonight against New York. CON +13 at home. The card has the spread as STRONG LEAN on Connecticut at minus 5.61 edge against the model. The model wants the line at CON minus 7.39 at neutral. Market has it at CON minus 13. Take CON +13 by raw model math.
The agent's read is the opposite.
NYL Q1 plus 4.60 HIGH. NYL Q2 plus 1.50 medium. NYL Q3 plus 4.40 HIGH. NYL Q4 plus 3.60 HIGH. Three HIGH-confidence quarter signals on New York's side with a "Close game, NYL closes" SGP correlation. The agent's 2025 net rating basis sees New York winning every quarter of this matchup despite Ionescu being Out.
The model-versus-agent conflict rule fires.
The rule is written for exactly this configuration. When the model picks one team and the agent has three-plus HIGH-confidence quarter signals on the opposite side, PASS the spread. The rule's record this season is 3-2 across five prior firings. The two correct firings of the rule that I personally wrote about — CON at ATL on June 2 (agent had four HIGH on ATL, bot's picks file had CON STRONG, ATL won by 16) and GSV at NYL on May 27 (agent had GSV by 27, bot had NYL, GSV blew NYL out) — both produced correct PASS calls on bot STRONG cards that lost by double digits.
Tonight's firing is the sixth.
The shape of the disagreement.
New York is missing Sabrina Ionescu (Out) per the morning injury report. Her ID appears in stars_out on the card and in home_injured_stars on the picks file. The Liberty without Ionescu have produced two wins in two games since the injury landed — the home win over Toronto last Wednesday (97 to 82) and the home win over Indiana on Saturday (83 to 75). Stewart-Jones interior plus a more distributed perimeter offense has produced enough output to win without the team's primary creator.
Connecticut is missing Brittney Griner (Out). That removes the Sun's interior anchor. Kennedy Burke is Day-To-Day. The home team is more thinned at this exact matchup than New York is.
The agent's read on the quarter shape is that New York's offense without Ionescu still has a meaningful structural advantage over Connecticut's offense without Griner. Stewart against a Connecticut interior missing Griner is the matchup that produces clean post entries and the kind of usage rate Stewart has not had since the Ionescu injury. The Q1 agent edge of plus 4.60 HIGH reads as Stewart establishing the post early. The Q3 edge of plus 4.40 HIGH reads as the second-half rotation where Connecticut's bench gets exposed without Griner anchoring the defense.
The card's case for Connecticut.
The card model has the line at CON minus 7.39 at neutral. The market line of minus 13 implies a 5.6-point market overshoot on the home dog's points. The card label of STRONG LEAN on CON reflects exactly that: the market is paying too much in points to the home team that has been getting blown out lately.
The card model's case is real. Connecticut's home record this season is below 0.500. The Sun have not produced the kind of structural advantages at home that justify market pricing them as a thirteen-point dog. The line implies a Liberty cover by single digits. The card argues the cover should be much bigger.
The agent's case is the opposite. Agent thinks Liberty wins by a Q1-aligned double-digit margin, possibly more if the third quarter develops the way the agent's plus 4.40 HIGH read implies. The agent's projection of the quarter shape implies a NYL win by 12-plus, which means CON +13 is the wrong side of the variance window.
The conflict rule is built on exactly this kind of disagreement.
The agent's 2025 net rating basis catches what the 2026 model can miss. Tonight's example is the agent reading the Stewart-Jones-with-Ionescu-out configuration as still structurally favorable against a Connecticut team operating without Griner. The 2026 model gives Connecticut too much credit for being at home without Griner because the home court adjustment is being applied to the wrong roster.
The verdict.
PASS the spread per the conflict rule. PASS the total (card has it at PASS already, edge is 0.17 which is essentially nothing).
The rule's record after tonight will be 3-3 if New York wins by single digits or fewer, and 4-2 if New York wins by double digits or covers the road by more than 13. The single-game variance does not change the rule's structural logic. The agent's 2025 net ratings caught a roster-mismatch shape that the card model's recent-form weighting missed. The rule continues to be the framework's correctly-firing edge filter.
What I am watching specifically.
The first quarter. The agent's plus 4.60 HIGH read on the Liberty Q1 is the strongest individual signal in the entire conflict reading. If New York opens with Stewart at 8-plus points in the first eight minutes and the score is 20-12 in their favor, the agent's read is on track and the rule fires correctly. If Connecticut opens with the home crowd carrying them to an early run and the Q1 is even or close, the agent's case is in trouble immediately.
The Stewart usage rate. Without Ionescu, Stewart has been carrying usage above 32 percent in the two prior wins. Connecticut's interior without Griner cannot match that usage with the kind of post defense the matchup requires. If Stewart's first three or four touches in the post produce clean buckets, the quarter math holds. If Connecticut's bench interior anchors hold up early, the matchup is closer than the agent expects.
The Burke status. Kennedy Burke is Day-To-Day for Connecticut. If she clears and plays full minutes, Connecticut has a perimeter creator who can manufacture half-court offense against a Liberty switching defense that does not have Ionescu's perimeter help. If Burke sits, Connecticut's offense without Griner AND without Burke gets very thin very fast.
Tip 7 PM ET at Mohegan Sun Arena. The conflict rule is firing for the sixth time. The discipline says PASS. The result will tell us whether the rule moves to 4-2 or stays at 3-3 at the unit-count level. Either resolution is consistent with the rule's structural logic. The framework continues to filter exactly the scenario it was built to filter.
[ End Report ]