Every registered strategy plays forward as write-once paper tickets, and every ticket is graded against the closing line. This is the out-of-sample test the rest of the site's claims are measured by — which is why it leads with the settled count and mean CLV rather than with profit.
Settled
18
23 with a close
Mean CLV
+2.0¢
best t = 0.84 · under the bar
Paper P&L
+42.4u
noise at this n — reported, not believed
Open
15
3 voided
Promoted
0
of 6 · as expected
nba-fav-ml
NBA moneyline favorites
mlb-model-ml
MLB model vs market
mlb-model-parlay
MLB model parlays
kalshi-totals-pinnacle
Kalshi totals vs Pinnacle
edge-screen-ev
Edge screen +EV (exchange offers)
xvenue-lock
Cross-venue locks (PM ↔ Kalshi)
Every strategy ever registered stays on this list, including the ones that die. Removing a failure would shrink the denominator and make whatever remains look better than it earned — precisely the trick this ledger exists to refuse. A dash means the strategy has not settled enough tickets to have the number at all, which is a different statement from a zero.
mlb-model-parlay · 2026-07-29 · ledger: starter scratch: Griffin Canning → Kyle Hart
mlb-model-ml · 2026-07-28 · ledger: starter scratch: Griffin Canning → Kyle Hart
mlb-model-ml · 2026-07-24 · ledger: starter scratch: Noah Cameron → Beck Way
A void is not a loss and not a push — it leaves the sample entirely, because the row it graded no longer exists. The void cascades from the prediction to every paper ticket referencing it, which is why a voided row shows a dash in every graded column rather than a zero: a zero would silently add a data point to a sample the row was explicitly removed from.
Beating the closing line is detectable in hundreds of tickets; beating the book on profit needs thousands. So P&L is reported and ignored, and promotion is decided on closing-line value alone, against Pinnacle's close.
Eligibility for review only — P&L significance additionally needs thousands of tickets (a real ~5–8c edge failed p<0.05 at 672 bets).
Stakes are quarter-Kelly capped at 2.5% of a non-compounding 1,000-unit paper bankroll. Non-compounding on purpose: a stake that grew with the bankroll would make an unlucky streak look like a broken strategy, and a lucky one look like skill.
