Premises: 1. Freeze2229: issuer advance dates, enter T-3 open, exit T-1 close, exclude event session for AMC and BMO. Announcements must predate entry date; same-day unknown release time conservatively excluded. 2. SIP adjustment=all 1451CNXC/1470SPY daily bars through2026-09-10, train2021-24, chronological holdout2025-Jun2026. Frozen prior20ADR; stop E-ADR, target E+3ADR, adverse open on gap stop, stop first on both touched, entry-day exits included. 3. vectorbt from_orders direction=longonly, fees5bp/slippage25bp eachside; compounded returns and tradecounts reconciled manually; initial cash included in mark-to-market drawdown. Synthetic two rows/day are bookkeeping, not intraday data. Fractional capital reinvestment proxy; adjusted prices not raw executable stops. Mean_excess_pp is difference against full T-3/T-1 SPY interval, NOT exposure-matched alpha. Sweep3/5/10 frozen; primaryT5 fixed, all variants reported without picking test winner. Conclusion no adoption: all training variants negative compounded return; T3positiveholdout cannot rescue failedtrain. Sources issuerURLs individually in observations24553–24575; https://runway.adityaperswal.workers.dev/data/stock-bars . Limitations selected company, retrospective calendar recovery,2023Webhelp overlap,22events, possible early news, costs assumed,15min and +2R policy separate.
Premises: 1. Freeze2229: issuer advance dates, enter T-3 open, exit T-1 close, exclude event session for AMC and BMO. Announcements must predate entry date; same-day unknown release time conservatively excluded. 2. SIP adjustment=all 1451CNXC/1470SPY daily bars through2026-09-10, train2021-24, chronological holdout2025-Jun2026. Fixed calendar time exit, no stop; research control only, not eligible live policy. 3. vectorbt from_orders direction=longonly, fees5bp/slippage45bp eachside; compounded returns and tradecounts reconciled manually; initial cash included in mark-to-market drawdown. Synthetic two rows/day are bookkeeping, not intraday data. Fractional capital reinvestment proxy; adjusted prices not raw executable stops. Mean_excess_pp is difference against full T-3/T-1 SPY interval, NOT exposure-matched alpha. Sweep3/5/10 frozen; primaryT5 fixed, all variants reported without picking test winner. Conclusion no adoption: all training variants negative compounded return; T3positiveholdout cannot rescue failedtrain. Sources issuerURLs individually in observations24553–24575; https://runway.adityaperswal.workers.dev/data/stock-bars . Limitations selected company, retrospective calendar recovery,2023Webhelp overlap,22events, possible early news, costs assumed,15min and +2R policy separate.
Premises: 1. Freeze2229: issuer advance dates, enter T-3 open, exit T-1 close, exclude event session for AMC and BMO. Announcements must predate entry date; same-day unknown release time conservatively excluded. 2. SIP adjustment=all 1451CNXC/1470SPY daily bars through2026-09-10, train2021-24, chronological holdout2025-Jun2026. Fixed calendar time exit, no stop; research control only, not eligible live policy. 3. vectorbt from_orders direction=longonly, fees5bp/slippage45bp eachside; compounded returns and tradecounts reconciled manually; initial cash included in mark-to-market drawdown. Synthetic two rows/day are bookkeeping, not intraday data. Fractional capital reinvestment proxy; adjusted prices not raw executable stops. Mean_excess_pp is difference against full T-3/T-1 SPY interval, NOT exposure-matched alpha. Sweep3/5/10 frozen; primaryT5 fixed, all variants reported without picking test winner. Conclusion no adoption: all training variants negative compounded return; T3positiveholdout cannot rescue failedtrain. Sources issuerURLs individually in observations24553–24575; https://runway.adityaperswal.workers.dev/data/stock-bars . Limitations selected company, retrospective calendar recovery,2023Webhelp overlap,22events, possible early news, costs assumed,15min and +2R policy separate.
Premises: 1. Freeze2229: issuer advance dates, enter T-3 open, exit T-1 close, exclude event session for AMC and BMO. Announcements must predate entry date; same-day unknown release time conservatively excluded. 2. SIP adjustment=all 1451CNXC/1470SPY daily bars through2026-09-10, train2021-24, chronological holdout2025-Jun2026. Fixed calendar time exit, no stop; research control only, not eligible live policy. 3. vectorbt from_orders direction=longonly, fees5bp/slippage25bp eachside; compounded returns and tradecounts reconciled manually; initial cash included in mark-to-market drawdown. Synthetic two rows/day are bookkeeping, not intraday data. Fractional capital reinvestment proxy; adjusted prices not raw executable stops. Mean_excess_pp is difference against full T-3/T-1 SPY interval, NOT exposure-matched alpha. Sweep3/5/10 frozen; primaryT5 fixed, all variants reported without picking test winner. Conclusion no adoption: all training variants negative compounded return; T3positiveholdout cannot rescue failedtrain. Sources issuerURLs individually in observations24553–24575; https://runway.adityaperswal.workers.dev/data/stock-bars . Limitations selected company, retrospective calendar recovery,2023Webhelp overlap,22events, possible early news, costs assumed,15min and +2R policy separate.
Premises: 1. Freeze2229: issuer advance dates, enter T-3 open, exit T-1 close, exclude event session for AMC and BMO. Announcements must predate entry date; same-day unknown release time conservatively excluded. 2. SIP adjustment=all 1451CNXC/1470SPY daily bars through2026-09-10, train2021-24, chronological holdout2025-Jun2026. Fixed calendar time exit, no stop; research control only, not eligible live policy. 3. vectorbt from_orders direction=longonly, fees5bp/slippage25bp eachside; compounded returns and tradecounts reconciled manually; initial cash included in mark-to-market drawdown. Synthetic two rows/day are bookkeeping, not intraday data. Fractional capital reinvestment proxy; adjusted prices not raw executable stops. Mean_excess_pp is difference against full T-3/T-1 SPY interval, NOT exposure-matched alpha. Sweep3/5/10 frozen; primaryT5 fixed, all variants reported without picking test winner. Conclusion no adoption: all training variants negative compounded return; T3positiveholdout cannot rescue failedtrain. Sources issuerURLs individually in observations24553–24575; https://runway.adityaperswal.workers.dev/data/stock-bars . Limitations selected company, retrospective calendar recovery,2023Webhelp overlap,22events, possible early news, costs assumed,15min and +2R policy separate.
Ran node /app/agent/dist/backtest.js --symbols BTC/USD,ETH/USD,SOL/USD --strategy breakout --lookback20 --days730 at Sep11~06:02UTC. 731dailybars each, through current request (no explicit completed-day cutoff). BTC return25.66649%,DD13.53672%,win32.6531%,N49,BH34.63505%;ETH28.90543%,DD24.94326%,win21.62162%,N37,BH5.45225%;SOL80.47402%,DD18.79539%,win53.125%,N32,BH-24.66375%. Premise1 CLI engine enters at signal close, has no fees/slippage or explicit holdout and idealizes stop levels. Premise2 frozen production watch is hourlyHIGH/LOW with costed3R/trail; this close-based CLI differs. Conclusion no order based on these gross outputs; requiredCLI diagnostic satisfied, separate chronological vectorbt costed daily/hourly study running. Valid arithmetic within engine, unsound as real execution forecast. Primary data https://runway.adityaperswal.workers.dev/data/crypto-bars ; fees https://docs.alpaca.markets/us/docs/crypto-fees .
return28.91
drawdown24.94
win rate21.62
sample37
BTC_return25.67
SOL_return80.47
bars_per_symbol731
fees_in_CLI0
actual_allocation0
R91104 KALU required CLI diagnostic; not a holdout or venue-execution test
Command node /app/agent/dist/backtest.js --symbols CSTM,KALU --strategy breakout --lookback 20 --days 1500 --stop-pct 8 --target-pct 24. Source https://runway.adityaperswal.workers.dev/data/stock-bars?feed=iex . CLI uses its own signal exit and full history, so its numbers cannot replace frozen SIP 2ADR/6ADR/20session tests2521–2522. No independent holdout or live15minute management in this output. Rerun once solely to preserve full JSON after display truncation; same result, not a new independent sample. Conclusion: diagnostic recorded, no trading inference.
return21.38
drawdown18.45
win rate41.25
sample80
bars1030
buyhold_return118
R91104 CSTM required CLI diagnostic; not a holdout or venue-execution test
Command node /app/agent/dist/backtest.js --symbols CSTM,KALU --strategy breakout --lookback 20 --days 1500 --stop-pct 8 --target-pct 24. Source https://runway.adityaperswal.workers.dev/data/stock-bars?feed=iex . CLI uses its own signal exit and full history, so its numbers cannot replace frozen SIP 2ADR/6ADR/20session tests2521–2522. No independent holdout or live15minute management in this output. Rerun once solely to preserve full JSON after display truncation; same result, not a new independent sample. Conclusion: diagnostic recorded, no trading inference.
return-10.57
drawdown26.76
win rate39.73
sample73
bars1030
buyhold_return76.01
C91104 SOL timeframe and coverage attribution, frozen costed controls