jaxperro
2f4f2ccbbe
research/ is a hard silo (README rules): read-only tape, no bot imports, own launchd (com.jaxperro.research-nightly 09:15, after daily ingest). - tape.py: proxy-resolution (the 742/742-validated method), niche + crypto strike/expiry/sprint parsers, tick loaders - sim.py: FAK execution replayer; hold_s=3 fitted on 29 real labeled live attempts (79% fill/miss classification), price noise 2-4c, measured OPTIMISM BIAS -2c/fill carried into every verdict threshold - requote.py: crater refill timing per niche (crypto 94% <4s, esports 83% <10s, sports needs ~25s, geo/politics minutes) -> params/requote_timing.json - study_flow.py + robustness: in-play surge momentum. Identity NULL result: 10 pooled controls +23.85/fill == informed +23.68 -> hypothesis revised at freeze, surge-EV primary, identity secondary (#16) - study_oracle.py: oracle digital fair value. 86% craters, winner's-curse inversion at big edges, nothing frozen (no cell at 30 fills) (#17) - forward.py + nightly.sh: re-scores frozen studies on last 3 tape days, appends forward_ledger.jsonl; verdicts ONLY from post-freeze rows Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
1.7 KiB
research/ — the edge factory (SILO)
Standing rules (user directive 2026-07-20: "built in a silo to not affect anything on the live bot"):
- Nothing here is imported by, or imports, the bot (
copybot.py,copytrade.py, their configs). The Fly workers run pinned entrypoints; this directory is inert to them. - Tape access is read-only (
duckdb.connect(..., read_only=True)). - The ONE shared write:
live/cache.duckdb::resolutionsvialive/payouts.py— append-only immutable chain facts, the same store the daily pipeline already feeds. Nothing else inlive/is touched. - The existing paper bot is the live test's CONTROL — graduated edges get their own paper harness here, never that one.
- Studies are pre-registered (GitHub issue per study: hypothesis, params,
verdict + kill criteria) BEFORE their forward window opens. Exploration
happens on already-collected tape; verdicts only come from
forward_ledger.jsonlrows dated after the params freeze commit.
Layout: tape.py read-only loaders · tape proxy-resolution (terminal-VWAP + sibling veto, the 742/742 chain-validated method) · title parsers (niche, crypto strike/expiry/sprint) sim.py execution replayer calibrated on OUR live fills ledger (lag, FAK no-match, protected band, 3% taker fee) study_flow.py Study A — informed-flow state signal study_oracle.py Study B — crypto oracle fair value vs the book requote.py crater→requote timing measurement (retry tuning) forward.py scores frozen studies on new tape days → forward_ledger.jsonl params/ frozen study parameters (committed = frozen) nightly.sh manual/launchd runner (separate from daily.sh)