NBA: 58–23 (71.6%)
Moneyline 11–0 (100%) · Spread 38–18 (67.9%) · Over/Under 9–5 (64.3%)
Track Record
The scoreboard the engine has to live with.
Every pick flagged as an EDGE is logged the moment the engine posts it, then graded the moment the game ends. No edits, no deletions, no cherry-picking.
Overall edge record
109–53 (67.3%) across 162 graded edge picks, counting moneyline, spread, and totals independently.
By sport
Sports appear here once they have at least 20 graded edge picks: below that, a win percentage is noise, not signal.
MLB: 23–10 (69.7%)
Moneyline 3–0 (100%) · Spread 12–4 (75%) · Over/Under 8–6 (57.1%)
NCAAF: 12–8 (60%)
Moneyline 1–2 (33.3%) · Spread 7–4 (63.6%) · Over/Under 4–2 (66.7%)
Does higher confidence actually win more?
Confidence is a calibrated probability, so higher-conviction picks should win more often. Here is whether they do, across every graded moneyline call, not just edge picks.
High conviction (70%+): 500–131 (79.2%)
Medium conviction (60–69%): 442–263 (62.7%)
Low conviction (52–59%): 892–697 (56.1%)
Recently graded edge picks
The newest graded results, straight from the log. Pending picks never appear here, only finished games.
NBA · MEM @ UTA · WIN
ATS: MEM · 2026-10-26
NBA · LAC @ OKC · WIN
ATS: LAC · 2026-10-25
NBA · GS @ PHO · LOSS
ATS: GS · 2026-10-24
NBA · DET @ MIA · WIN
ATS: DET · 2026-10-23
NBA · DEN @ OKC · WIN
O/U: Under · 2026-10-22
NBA · IND @ NO · WIN
ATS: NO · 2026-10-21
NBA · BOS @ PHI · LOSS
O/U: Under · 2026-10-16
NCAAF · KENEST @ MSRST · WIN
O/U: Under · 2026-10-13
NFL · CHI @ GB · LOSS
ML: CHI · 2026-10-11
NCAAF · JMAD @ GAS · WIN
ML: JMAD · 2026-10-10
NCAAF · IOWA @ WASH · WIN
ATS: IOWA · 2026-10-09
NCAAF · JAXST @ KENEST · WIN
ATS: KENEST · 2026-10-07
NCAAF · SOUMIS @ TROY · LOSS
O/U: Under · 2026-10-06
NCAAF · ARMY @ LOUTCH · LOSS
ML: ARMY · 2026-10-03
NCAAF · LIBRTY @ DEL · WIN
O/U: Under · 2026-10-02
NFL · PHI @ CHI · LOSS
O/U: Over · 2026-09-28
NFL · MIN @ TB · WIN
ML: MIN · 2026-09-27
NCAAF · SALA @ UK · LOSS
ATS: SALA · 2026-09-26
NCAAF · CLMSN @ CAH · LOSS
ATS: CAH · 2026-09-25
NFL · NYG @ LAR · LOSS
ATS: NYG · 2026-09-21
NFL · GB @ NYJ · WIN
ATS: NYJ · 2026-09-20
NCAAF · HOU @ TXTECH · LOSS
O/U: Under · 2026-09-18
NCAAF · SYRA @ PITT · WIN
O/U: Under · 2026-09-17
NFL · DEN @ KC · LOSS
ATS: DEN · 2026-09-14
NFL · DAL @ NYG · WIN
ATS: NYG · 2026-09-13
Graded football spread and total picks (model picks, not Edge Verified)
Since the model bump, every NFL and college football spread or total at 52.4% or better on our model’s probability is a graded pick. This is their whole record at -110, pushes excluded, whether or not a pick was Edge Verified. The percentages on the cards are the model’s estimates; this is what happened.
64–56 (53.3%) across 120 graded picks · units at -110: +2.18
NFL: 11–12 (47.8%) · -2.00u
Spread 2–8 (20%) · Over/Under 9–4 (69.2%)
NCAAF: 53–44 (54.6%) · +4.18u
Spread 27–22 (55.1%) · Over/Under 26–22 (54.2%)
How this record is kept honest
Logged at post, graded at final
A pick enters the log the moment the engine publishes it. When the game ends, it is graded against the actual result. There is no window to edit, re-frame, or quietly drop a pick.
PASS is not a loss, and not a win
When the engine’s numbers do not back a side by enough it returns a PASS, and passes are excluded from the record entirely. Only recommended picks count. See what PASS means.
A lean is not a pick
On football boards a spread or total below our 52.4% gate shows as a lean: the side our projection favors and its percentage, with no grade. Leans are never counted here, in either direction. Only graded picks are.
Football spread and total percentages are the model’s, not this record
Football spread and total percentages on the cards are the model’s own probabilities, not this record’s hit rate. This page is the measured part: every graded pick, win or lose, at the line it was published at. If the model’s numbers run ahead of the results here, the results are what count.
Small samples are labeled, not hidden behind percentages
A sport only gets a public win rate once it has 20+ graded edge picks. Short-term streaks (hot or cold) are noise; the long-run aggregate is the signal. Past performance never guarantees future results.
See how the picks are made
The full methodology behind every projection is public: the five-anchor model, per-sport math, and the PASS discipline.