← REGISTRY
CASE FILE · REPLAY-LEGACY-BOTHSIDES-TWOSIDED
HISTORICAL REPLAYVERIFIED · ADAPTED8h capture · 33 windows · 6,312 events
NEGATIVE
NET −$4.33 · BOTH LEGS · DERIVED NO BOOK · NO % BASIS
Two-sided and reproducible: both legs filled against real depth and settled independently, and the strategy still lost. The engine models both sides faithfully; the loss is the strategy's.
PM-5M-BOTHSIDES-SCALP (BOTH LEGS, ADAPTED) — RECOVERED FROM A LEGACY BACKTEST
POLYMARKET BTC 15M UP/DOWN (UP + DOWN TOKENS; DOWN BOOK DERIVED)
POLYMARKET BTC 15M UP/DOWN (UP + DOWN TOKENS; DOWN BOOK DERIVED)
PROOFLINE
HISTORICAL REPLAY
RECORD SEALED
BOTH LEGS · 50/50 DEPTH-VERIFIEDDERIVED NO BOOK (ADAPTED)24 SETTLED · 0 OPENREPLAY BYTE-IDENTICAL
WHAT HAPPENED?
The M6 record verified only the up-side leg because the engine modeled one token at a time. With two-sided replay, the same buy-the-cheap-side scalp ran with BOTH legs — buying the up token when cheap and the down token when cheap, each taking profit at 56¢ or holding to the result. 50 trades (25 up, 25 down) all filled against book depth; 24 positions settled, none left open. It lost $4.33 net after $6.70 in fees.
WHAT WAS TESTED?
The same legacy bot, both sides at once: Polymarket BTC 15m up/down (up + down tokens; down book derived) · 8h capture · 33 windows · 6,312 events.
WHY TRUST IT?
Each leg filled against its own side's book depth (never the other side's), fees applied per trade, both YES and NO settled independently from the real venue outcome, and rerunning the sealed record produced the exact same report twice.
WHAT IS STILL UNKNOWN?
This does not mean the strategy works, and it is not a plain Verified Record: the source capture recorded only the up book, so the down (NO) book is DERIVED as the binary complement. Because Polymarket up/down are independent tokens, that complement is an approximation — a Verified ADAPTED Record. See the disclosures.
EVIDENCE · EXPAND TO INSPECT
PLAIN STORY FIRST · PROOF UNDERNEATH
Verbatim from the sealed record. The technical proof below lists the artifacts behind it.
How many times the rules fired during the tested period.
Every simulated trade was checked against the order book's displayed availability at that moment. “Partly” means the book showed less than the strategy wanted, so only the displayed amount filled.
Verbatim from the sealed record. The technical proof below lists the artifacts behind it.
Verbatim from the sealed record. The technical proof below lists the artifacts behind it.
Positions still awaiting a final result when the record ended. Anything pending stays pending; results after the recording stopped are never filled in.
The sealed two-sided spec re-run on the derived DatasetVersion produces a byte-identical report_hash (9baecdb8…) across independent runs; the ledger chain (8b35dcf1…) is stable regardless of engine code_version, so the economics are pinned even as provenance advances.
- ✓Strategy rules recovered from source
- ✓Whether trades could go through (both legs)
- ✓Fees
- ✓Final market results (YES and NO)
- ✓Profit/loss
- ✓Reproducibility
Events replayed6,312
Intents50 (25 YES + 25 NO)
Fills76 (50 entries + 26 take-profit sells)
depth_verified (YES / NO)50 (25 / 25)
depth_partial / depth_unverified0 / 0
Rejected — NO book absent0
Settlements applied32
Positions closed24
Open positions at end0
- This is a Verified ADAPTED Record, not a plain Verified Record: the down/NO book was never recorded by the source capture. It is DERIVED as the binary complement of the recorded up book (best_ask_no = 1 − best_bid_yes, and mirrored depth). Polymarket up/down are independent tokens, so this complement is an approximation, disclosed in the DatasetVersion's known_gaps. For a true binary (e.g. Kalshi) the complement would be exact and the same spec could produce a plain Verified Record.
- No percentage return is stated: the spec sizes a fixed $1.80 per leg with no declared capital basis, so a percentage would be invented.
- Timeframe adaptation (5-minute source → 15-minute settling dataset) and the dropped 60-second fill gate (no secs_elapsed recorded) are carried over from the M6 reading and remain declared assumptions.
TECHNICAL PROOF · FOR THOSE WHO WANT IT
NEVER SHOWN BEFORE THE STORY
FINGERPRINTS (HASHES)
SEMANTIC SPEC FINGERPRINT
15ac36f5211dcda17d9e676417f416e8a79361f7a4428b42fe42a50d49dfc732
SPEC HASH
7e74f7e4ebcadfc92c9f5696e706f1b9094df130809395c02b64aa8ff6cf281e
REPORT (TWO-SIDED V0.2)
9baecdb8b0ca00d30b0a044a43abf704702dbc6a6f516e5b943611d359ae3375
LEDGER CHAIN
8b35dcf1bfd9502d4388018500d4eceeba0437963f20f6b3247f53da53e4c9a2
DATASETVERSION (DERIVED, NORMALIZED)
40f9103a635b6564c9ad70d438bceec644fc719837363649fb1163c90609ee40
REPRODUCIBILITY
The sealed two-sided spec re-run on the derived DatasetVersion produces a byte-identical report_hash (9baecdb8…) across independent runs; the ledger chain (8b35dcf1…) is stable regardless of engine code_version, so the economics are pinned even as provenance advances.
EVIDENCE FILES · PROOFLINE REPOSITORY
compiler-benchmark/legacy/bothsides-twosided/spec.jsonSEALED
compiler-benchmark/legacy/bothsides-twosided/report_twosided.jsonSEALED
compiler-benchmark/legacy/bothsides-twosided/M7_NUMBERS.jsonSEALED
compiler-benchmark/replays/two-sided-0710/dataset/manifest.jsonSEALED
compiler-benchmark/replays/two-sided-0710/coverage_report.jsonSEALED
docs/integration/FIRST_TWO_SIDED_STRATEGY_SELECTION.mdSEALED
Written by the Proofline engine; the site only displays them. See the method for how a record is reproduced.
ENGINE RECORD
REPLAYREPLAY ONLYLEGACY IMPORTTWO-SIDEDP&L: COMPUTABLEBYTE-REPRODUCIBLE
Trust levelREPLAY ONLY
Run statusCOMPLETED
P&L statusCOMPUTABLE
P&L DETAIL
Net P&L−$4.33
Realized (gross)+$2.37
Fees$6.70
Max drawdown−$10.03
Two-sided spec: both YES and NO legs, max_concurrent 2, 5m→15m timeframe, 60-second fill gate dropped, fixed $1.80 per leg. The down/NO book is DERIVED as the binary complement of the recorded up book (Polymarket up/down are independent → approximation → Verified Adapted Record). Semantic hash 15ac36f5….
VERIFIED UNDER THE DATA AND ASSUMPTIONS SHOWN. NO GUARANTEE OF FUTURE RESULTS.