Research · Model books

Model tearsheets

Operator view: design evidence is muted; out-of-sample accrual is the decision surface. The append-only consensus ledger is the only LIVE row. Default order is status group then OOS Sharpe descending; killed books remain visible at the bottom.

LIVEbook_liveNo backfilled design series; this row is ledger-only.2026-07-06 → 2026-07-14-1.9%Append-only book_ledger provenance='live'; 4 ledger sessions, n=3 settled; pending 2026-07-17.-2.02-5.2%3 days4 ledger sessions total-0.27live
REPLAYseat_v6n22026-04-27 → 2026-06-30+138.2%Selection window. CURRENT-corpus replay using current dashboard scoring semantics; historical membership and extraction can differ from what was available contemporaneously (corpus drift). Replay, not live fills.2026-07-01 → 2026-07-16+1.3%Current-corpus replay; live: consensus only.0.79-11.8%11 daysOOS replay days0.17OOS
REPLAYseat_v9n2026-04-27 → 2026-06-30+30.6%Selection window. CURRENT-corpus replay using current dashboard scoring semantics; historical membership and extraction can differ from what was available contemporaneously (corpus drift). Replay, not live fills.2026-07-01 → 2026-07-16-4.8%Current-corpus replay; live: consensus only.-2.92-9.8%11 daysOOS replay days-0.65OOS
REPLAYseat_modelV312026-04-27 → 2026-06-30+30.4%Selection window. CURRENT-corpus replay using current dashboard scoring semantics; historical membership and extraction can differ from what was available contemporaneously (corpus drift). Replay, not live fills.2026-07-01 → 2026-07-16-16.6%Current-corpus replay; live: consensus only.-8.49-16.6%11 daysOOS replay days-1.89OOS
REPLAYseat_v6n2026-04-27 → 2026-06-30+261.3%Selection window. CURRENT-corpus replay using current dashboard scoring semantics; historical membership and extraction can differ from what was available contemporaneously (corpus drift). Replay, not live fills.2026-07-01 → 2026-07-16-35.8%Current-corpus replay; live: consensus only.-12.74-36.7%11 daysOOS replay days-2.82OOS
BENCHEDbenched_v92026-04-27 → 2026-06-30+92.4%Own-era selection window. benched 2026-06-30; replayed on current corpus — corpus-drift caveat applies; July continuation shown for seat-review comparison2026-07-01 → 2026-07-16+10.8%Post-bench current-corpus continuation replay; corpus-drift caveat applies.5.61-5.5%11 daysOOS replay days1.23OOS
BENCHEDbenched_v6b2026-04-27 → 2026-06-30+125.8%Own-era selection window. benched 2026-06-30; replayed on current corpus — corpus-drift caveat applies; July continuation shown for seat-review comparison2026-07-01 → 2026-07-16+18.2%Post-bench current-corpus continuation replay; corpus-drift caveat applies.4.91-10.9%11 daysOOS replay days1.07OOS
BENCHEDbenched_v6c2026-04-27 → 2026-06-30+131.5%Own-era selection window. benched 2026-06-30; replayed on current corpus — corpus-drift caveat applies; July continuation shown for seat-review comparison2026-07-01 → 2026-07-16+3.0%Post-bench current-corpus continuation replay; corpus-drift caveat applies.1.44-11.2%11 daysOOS replay days0.31OOS
PAPERn1_fade2026-05-19 → 2026-07-13+33.6%Single chartered window — chartered (bars frozen pre-result); OOS accrual from charter date. Structural horizon: the last eligible N1 entry is 2026-07-07 and its full five-session mark ends 2026-07-13. Live resolved-true outcomes through 2026-07-10 were evaluated; no later eligible bull-majority event has a complete locally marked hold by 2026-07-17.trades entered after 2026-07-07n=0Strictly post-charter entries only.n=0n=00 tradespost-charter tradesn=0OOS
VALIDATEDvictory_lap_outright2026-05-05 → 2026-07-16+9.1%Single chartered window — chartered (bars frozen pre-result); OOS accrual from charter date.trades entered after 2026-07-17n=0Strictly post-charter entries only.n=0n=00 tradespost-charter tradesn=0OOS
VALIDATEDvictory_lap_paired2026-05-05 → 2026-07-16+3.8%Single chartered window — chartered (bars frozen pre-result); OOS accrual from charter date.trades entered after 2026-07-17n=0Strictly post-charter entries only.n=0n=00 tradespost-charter tradesn=0OOS
CANDIDATEshort_vol_weekly2024-08 → 2026-04+2,644.3%Full discovery window; historical result is design evidence.formations after 2026-05-01 → 2026-06-08+31.3%n=13 trades across 4 weekly formations; OOS mostly ahead; paper pending. Structural horizon: the last qualifying locally priceable entry is 2026-06-08 and its full 21-session exit is 2026-07-09. Local EOD chain parquet ends 2026-07-10, and later formations either fail the frozen two-sided quote/spread/contract quality rules or require unavailable later exits. All qualifying local holds completing by 2026-07-10 are included; no network data was used.35.240.0%13 trades4 weekly formations11.37OOS
CANDIDATEmodel_optx2024-09-25 → 2025-09-24 (purged train)First-year fit only; no in-sample return is presented as evidence.2025-09-26 → 2026-07-16-19.4%Frozen second-year one-look OOS. Strict one-look MC3S-b OOS book. Purged first-year train; OOS begins 2025-09-25. Last complete 44-session entry is 2026-05-13 and marks through 2026-07-16; later signals lack a complete local hold. Plain-NumPy depth-2 gradient boosting, 200 rounds, lr=0.05.-1.70-27.4%550 tradesfrozen OOS picks-3.46OOS
CANDIDATEn1_lap_capped2026-05-05 → 2026-07-16+42.2%Single chartered window — chartered (bars frozen pre-result); OOS accrual from charter date.trades entered after 2026-07-07n=0Strictly post-charter entries only.n=0n=00 tradespost-charter tradesn=0OOS
CANDIDATEoiA2024-09-26 → 2026-07-16-13.3%MC3S chartered 2026-07-19 with bars frozen pre-result; historical replay is prior evidence, forward paper is the judge.trades entered after 2026-07-19n=0Forward paper is the judge; no historical replay result is promoted to OOS.n=0n=00 tradespost-charter paper tradesn=0OOS
CANDIDATEoiT2026-04-22 → 2026-07-16+11.8%MC3S chartered 2026-07-19 with bars frozen pre-result; historical replay is prior evidence, forward paper is the judge. Joint Twitter/OI window starts 2026-04-20 and is severely data-limited.trades entered after 2026-07-19n=0Forward paper is the judge; no historical replay result is promoted to OOS.n=0n=00 tradespost-charter paper tradesn=0OOS
DEADdead_gs_long_straddles2024-08-05 → 2026-06-08-98.9%Structural horizon: the last qualifying locally priceable entry is 2026-06-08 and its full 21-session exit is 2026-07-09. Local EOD chain parquet ends 2026-07-10, and later formations either fail the frozen two-sided quote/spread/contract quality rules or require unavailable later exits. All qualifying local holds completing by 2026-07-10 are included; no network data was used.Killed; no OOS accrual.1,060 tradesfull killed window-22.47full
DEADdead_radar_straddles2024-09-23 → 2026-06-22-87.2%Structural horizon: the last qualifying locally priceable entry is 2026-06-25 and its full 10-session exit is 2026-07-10. Local EOD chain parquet ends 2026-07-10, and later formations either fail the frozen two-sided quote/spread/contract quality rules or require unavailable later exits. All qualifying local holds completing by 2026-07-10 are included; no network data was used.Killed; no OOS accrual.4,334 tradesfull killed window-23.85full
DEADdead_silence_fade2026-05-25 → 2026-07-06-2.0%Structural horizon: the last eligible weekly entry is 2026-07-06 and its full five-session exit is 2026-07-13. The next weekly construction cannot complete its full hold by 2026-07-17; the hub span is the Monday label of the entry week.Killed; no OOS accrual.329 tradesfull killed window-0.41full

Production tracking — accruing & recording

Forward-judged tracks with no return series yet: paper overlays still accruing and live instruments recording. No OOS return is shown — each is judged only at its pre-registered clock, not before.

StatusTrackKindThesisJudgment clockOOS
TRACKINGaccruing — forward judgedt2b_hollow_lappaper overlayVictory-lap fades pay when the options crowd never positioned (low own-history NDOI): +4.9-5.7%/trade, 67-76% hit.n=25, one quarter, hypothesis-inverted discovery 2026-07-19.regrade at n>=60 hollow-lap events or 2 quartersaccruing (n~25)
TRACKINGrecording — forward judgedu1_residualinstrumentOptions-vs-twin disagreement grades confidence of tweet-book trades (options model beat twin +7pp hit OOS).daily ledger live since 2026-07-19 (284 names/day).first read after ~40 forward sessionsrecording (n=0)

Click any column header to sort. Window results are span-true compounded returns. OOS statistics use only the displayed OOS return marks; when a statistic is not defined, the cell reports n=X. “Obs” is the OOS count where OOS exists and the full count for killed books. Clustered t uses OOS entry cohorts where OOS exists and the full available window otherwise.

Report conventions retained: annualized CAGR is shown only for spans of at least 180 calendar days. QuantStats Sortino uses mean/downside-deviation × √(periods/year), which differs from plain-downside conventions. Paired returns keep the verified v1 sum-of-legs convention; inference uses 10,000 entry-cohort bootstrap draws. Premium-normalized books assume a 10%-of-capital weekly premium budget; per-trade premium returns are shown unscaled in Avg obs.