← setups index

Trader fill match: logged entries vs real 1m option bars (Nifty + Sensex)

Input: trader-log/trades_timestamped.csv (225 rows, Excel v2 reload 2026-10-06). In scope: 223 index option trades (Gold and the no-trade day are excluded). Each row is matched on logged strike + CE/PE + entry minute + entry premium against the real chain for that instrument (all live expiries, ±15 min). This is not an ATM guess.

Sensex path (this run): BSE Sensex 1m option bars under nifty-options-s3/data/options/sensex/1m/ (by_expiry + live). Logged Sensex strike / premium / points are used as-is (no N/S scale). match_source=sensex_chain when matched on those bars; proxy_legacy only if Sensex bars are missing for that day (0 rows). Nifty rows stay on nifty_chain.

Headline

• Nifty strike + premium (n=35): 10/35 = 29% loose (premium inside a bar range within ±15 min, or within 5%), and 1/35 = 3% strict (inside the bar range within ±5 min). Placebo (±1 week): 31% loose, 11% strict. The loose rate is at chance level, and the strict rate is below the placebo. 7 of 35 premiums are more than 25% away from anything that strike traded near the logged time.
• Sensex strike + premium (real Sensex chain, n=37): 13/37 = 35% loose, 4/37 = 11% strict. Placebo: 18% loose, 7% strict (above chance vs placebo). Sensex chain rows in scope: 112; proxy_legacy fallbacks: 0. 13 Sensex premiums are >25% away from the logged strike's traded range.
• Rows with no logged premium (151, mostly May–Jul): strike + side + time only. Bar found for 148/151.
• Win rate: log 78.0% vs sim 28.3% (n=212; logged SL/target replayed on the matched contract). Outcome reproduced on 41% of rows. Alt readings: index-points on real spot 27.5%; ATM guess 26.1%. Matched NIFTY contracts (n=10): log 60% vs sim 40%. Matched SENSEX contracts (n=13): log 77% vs sim 31%.
• All rows with a logged entry premium (n=72): log WR 83.3% vs sim WR 46.9% (n_sim=64); outcome reproduced on 52%. 1 rows (dated 2026-10-05) have no chain coverage (X_no_*).
• vs previous proxy run: Previous Nifty-proxy run (archived run_pre_sensex_chain_20261007): Sensex premium proxy-loose 14/41 · overall log WR 78.0% vs sim 29.8% (n_sim=218) · premium rows log 83.3% vs sim 45.6%.
• Verdict: logged outcomes still do not reproduce at the rate implied by the log. Remining on won_log remains unjustified while row-level agreement sits around 41%.

Match quality by group

Tier counts

Tiers: A = premium inside the bar range within ±1 min · B = within ±5 min · B2 = within ±15 min · C = within 5% of the range · D = no premium logged (contract/time only) · F = premium mismatch or strike not listed. _nextexpiry = the best fit was not the nearest weekly expiry.

Win rate: log vs simulation

Simulation: fill = logged premium if it lies inside the matched bar, else clipped to the bar. Rows with no premium fill at the open of the entry minute. SL = logged risk. Target = logged target (Jul/Aug–Sep sheets) or logged realised points for May–Jul wins. For May–Jul losses the target is 5.2×risk (the median May–Jul win R-multiple). First touch is checked on the 1m bars after the fill bar, stop first on a same-bar double touch, flat at 15:29.

Logged status vs simulated exit

Weekend-dated sessions (picked by premium fit: previous vs next session)

Premium rows on real NIFTY strikes (worst first)

Premium rows on real SENSEX strikes (worst first)

Files

• results/fill_match.csv: one row per log row, with match_source, tier, expiry, fill bar/time shift, implied strike, placebo, option/index/ATM sims
• results/match_quality_by_group.csv, match_quality_by_tier.csv, wr_compare.csv, log_status_vs_sim_exit.csv, nifty_premium_rows_detail.csv, sensex_premium_rows_detail.csv, weekend_day_resolution.csv, headline.json, params.json
• src/match_fills.py (matcher + sims) · src/report.py (this report)
• Prior Nifty-proxy results: archive/run_pre_sensex_chain_20261007/