Track Record | EdgeStat

📈 CLV Per Bet — beating the close--

Cumulative CLV is the gold-standard proof you bet sharp. Positive CLV = you got value, even on losing plays. Long-run CLV correlates with long-run ROI.

Overall CLV

--

Trailing 30d CLV

--

Props w/ CLV

--

Total Settled

--

📅 Day-of-Week Performance--

Settled track record sliced by day-of-week. Use this to spot personal patterns (better on Tuesdays? bad on Sundays?).

🏟️ Performance by Sport--

Every sport the model grades, settled against real box scores / results. Includes the experimental prop firehose (tracked to train the model, not the curated bet list) — so blended ROI runs below the Alpha headline above. "Pend/Void" shows grading coverage: a sport with picks but nothing settled has no outcome feed wired yet.

🎯 Performance by Strategy--

Realized ROI per pick source. Featured = the strategies on the actual bet list (Locks, Top 25, Whale, Sharp, Consensus-hit, POD). Everything else is the experimental firehose, tracked to train the model — shown here, in the open, including the families that bleed.

🎯 Honest at the Source — Calibrated Reliability--

Every pick in the ledger carries two probabilities: the generator's raw number (kept untouched as the learning signal) and an isotonic-calibrated one learned from settled outcomes. Scored against the same real results: lower ECE means the displayed % is closer to how often it actually happens. In-sample; the held-out validation lives on the calibration map. Test it yourself in the strategy simulator.

📈 Cumulative P&Llast 30 plays

Running profit/loss over the last 30 settled plays. Smooth-ish upward trend is the point — flat days are normal in a 55% win-rate game.

📊 Full History — Honest Book vs Firehose--

Every settled pick at flat 1u, replayed chronologically. Green = the honest book (proven-negative families curated out — what the boards actually surface). Red = the raw firehose, including every experimental family we track to train the model. Same model, same real outcomes — curation is the difference. Run your own rules in the Strategy Simulator.

Date	Play	Price	Stake	P/L
Loading…

Date	Play	Price	Stake	P/L
Loading…

📋 Full Settled Logfilter: last 30

Bet type: Result: Confidence:

Date	Play	Price	Stake	Result	P/L	CLV

🧾 Methodology Note

All plays are graded against the price displayed when published. We log the closing line at game start to compute CLV. We do not retroactively remove "bad" plays, change stakes after publication, or hide losing months. The total units shown above include all plays — pre-game, in-game, and live.

Why does this matter? Because most "experts" you'll find online will quietly delete their losses and trumpet their wins. We do the opposite — we log everything precisely so you can validate that the model actually works. A 55-58% hit rate at –110 average vig is the truth of a sharp operation, and that's what we'll deliver if the underlying model is calibrated.

Public Track Record