ASCEND
Saturday · 2026-09-26 · ASCEND-QX 4.2

Today's AI Picks

Every weekday the model scores thousands of markets across football, basketball, baseball, hockey, soccer, college and combat sports, then publishes five strong plays, three lighter value plays and three player props projected over or under the line. Each pick shows its modelled probability, the best price found across the books compared and the stake that probability justifies — here is how a pick qualifies.

Weekend slate — the core release schedule runs Monday through Friday. Weekend cards are bonus coverage.

Measured record

Nothing has settled in the published sports ledger yet, so no hit rate is shown. Picks are graded automatically against final scores from the same schedule feed that supplied the fixture, and these tiles fill in the moment results land.

Measured accuracy

No settled plays have been graded in the published ledger yet. This panel fills in from the settlement ledger as results land — nothing is shown before it is measured.

Probability environment

Market relationships, not isolated picks

Nodes represent priced matchups and player markets; arcs show the probability relationships the engine compares before a play can publish.

Today's scheduled games

Every real fixture on the calendar today, straight from the public scoreboard feed the model prices against. Scores update on their own as games go live and settle.

Loading today's fixture list…

🔥 5 Strong Plays

Highest composite signal of the day. Released Monday through Friday. The top-ranked play is published free — the remainder of the card is released to members.

#1 · Strong Play · MLB

Pirates — Total Runs

Pirates @ Tigers · Baseball · 17:10 UTC

Under 9
+107 · 0.2u · ESPN BET
Sportsbooks · 10 books
Model Proj.
7.2
Win Prob.
53.2%
Edge vs Market
+4.1%
Confidence
57/100

Rolling Form (L10) and Usage & Role are the dominant drivers (Rolling z=-1.16, Usage z=-1.2, Matchup z=-0.92). Composite signal -1.11σ against a market implying 60.8%. Priced at +107 (best of 10 books, ESPN BET) against a de-vigged consensus of 49.1%.

Why this prediction exists
Model probability
53.2%
Price implied
49.1%
Market consensus
49.1%
Edge vs stricter reference
+4.1 pts
Edge needed to publish
+2.0 pts
Data quality
85/100
  • tier calibration is provisional (0 of 60 settled results)
Model explanation
Speculative · 57/100
Model win prob.
53.2%
Market implied
49.1%
Edge
+4.1%
Kelly stake
0.2u

Key statistical drivers

  • Rolling Form (L10)z -1.16
  • Usage & Rolez -1.2
  • Matchup Ratingz -0.92

Why this is a Strong play, not a Value play

Strong plays require both a calibrated win probability at or above the model's conviction floor and agreement across the top-weighted signals. Here the model lands at 53.2% against an implied 49.1%, giving a wide pricing gap the market has not corrected. Rolling Form (L10), Usage & Role, Matchup Rating all push the same direction, so there is no internal disagreement to discount — which is why the Kelly allocator sizes it at 0.2u rather than a reduced stake.

Why Under 9

The projection sits 1.8 below the posted line of 9. The model takes the Under because the projected distribution clears the number 53.2% of the time, while the price only demands 49.1%. A gap this size is large enough to survive a normal usage or pace swing.

Probabilities are isotonic-calibrated outputs of ASCEND-QX; stakes use quarter-Kelly. Model output is analytical information, not a guarantee of outcome.

Verifying your access…
Live odds

Real book prices behind today's plays

Unlock the sports engine to see the sportsbook price, consensus probability and edge behind every published play.

Best bets by league

League-specific boards, each with its own free headline play, market coverage and daily release.

How these picks are made

How AI Picks works — the model, its edge and measured accuracy →

Daily results audit — every graded slate, win or lose →

Model health & error analysis — Brier, log loss, ROI and every losing segment →

Verified fixtures only

Every pick is attached to a real scheduled game, cross-checked against a second schedule source.

Selection criteria →

Real book prices

Priced against the matching market at multiple books, with the best posted price and book count shown.

Model weights →

Graded in public

Settled at the released price and restated weekly, losing tiers included.

Accuracy audit →

AI Picks, answered

What a pick is, how it qualifies, and how often the board changes.

What is an AI Pick?

A dated selection produced by Ascend's model, published with the probability the model assigned it, the best real sportsbook price found across the books compared, the stake that probability justifies at fractional Kelly, and the level that invalidates it.

How are AI Picks selected?

A candidate has to be attached to a verified scheduled fixture, priced against the matching market at multiple books, disagree with the de-vigged market probability by more than the model's own historical error, and survive sizing. Anything that fails is dropped rather than padded onto the card.

How often are AI Picks updated?

Boards rebuild every fifteen minutes while fixtures are open, so the prices on the page track the books. The core release runs Monday through Friday; weekend cards are bonus coverage.

Are the picks human-reviewed?

The model output is not overridden by hand — no published probability is edited after the fact. The feature set, weights, thresholds and written analysis around it are built and reviewed by William Merrill.

How is AI Picks accuracy measured?

Only settled selections count. Each pick is graded against the final result at the price it was released at, then rolled into hit rate, ROI and units. Open selections are excluded until they close, so the number moves as games finish rather than when someone refreshes the page.

What accuracy should I expect?

Anything above 52.4% at -110 is profitable long term. The rolling hit rate and ROI published on this page are the real figures from the settled ledger, including losing stretches — no win rate is guaranteed.

Are the accuracy figures independently audited?

No. They are self-published from a reproducible backtest harness in this codebase and graded at released prices, with the losing tiers shown alongside the winning ones. No third party has verified them, and no page here claims otherwise.