Strategies

BTC Coin-Margined vs USD-Margined Funding-Basis Relative Value, Delta-Neutral Cross-Venue (Short the Higher-Funding BTC Perp / Long the Lower-Funding BTC Perp Across BINANCE_CM and BINANCE USD-M, Harvest the Persistent Inter-Margin Funding Spread, 2-Parameter)

Outcome: Abandoned

BtcCoinMUsdmFundingBasisRelativeValueNeutral

Outcome Summary

BtcCoinMUsdmFundingBasisRelativeValueNeutral sought to harvest a persistent funding spread between Binance's COIN-M and USD-M BTC perps as a delta-neutral pair. Across three iterations it hit a hard verification wall: the USD-M leg's per-instrument funding is never generated on the inline Layer-3 supplementary path, so the pair could never open, and threshold loosening (the literal fix requested) could not reach a series that wasn't there. The developer then measured the real data directly and refuted the premise — both contracts' funding derive from the same index premium and formula, leaving a near-zero spread 55-200x underwater against the ~0.20% pair round trip. It was ultimately abandoned as 'verifying_stuck_3_timeouts', never reaching a backtest, with both an infrastructure blocker and a falsified edge documented.

Strategy report

Hypothesis

A MARKET-NEUTRAL relative-value strategy that harvests the persistent FUNDING-RATE SPREAD between Binance's two BTC perpetuals with structurally different collateral bases: the COIN-margined inverse perp (BTCUSD_PERP.BINANCE_CM, margined and settled in BTC — dominated by miners and long-term holders) and the USD-margined linear perp (BTCUSDT.BINANCE — dominated by leveraged USDT traders). Because the two contracts track the same BTC index but have a different participant mix and a different collateral/PnL currency, their 8-hour funding rates persistently diverge. When one perp's funding runs materially richer than the other's, you SHORT the richer-funding perp and go LONG an equal-USD-notional of the cheaper-funding perp; the two BTC-price exposures cancel (delta ≈ 0) while you collect the funding differential 3x/day until it normalizes. This is NOT a momentum/OHLCV directional bet (the dead BTC-perp classes L9/L38/L41), NOT a single-leg option VRP short (which just timed out in verification), and NOT the spot-perp funding carry already in the pipeline — the edge is the CROSS-CONTRACT funding SPREAD, a distinct structural cash flow. Fills three of the most under-represented portfolio buckets at once: BINANCE_CM venue (3.2% vs ≥5% target), cross-venue scope (6.8% vs 15%), and market-neutral direction (13.3% long_short). Deliberately low-parameter (spread entry + exit thresholds) to resist the overfit/deflated-Sharpe deaths.

verifying_stuck_3_timeouts

Iteration History

Verification failed (Layer 3 — sandbox backtest): No trades produced Bar type used: BTCUSD_PERP.BINANCE_CM-8-HOUR-LAST-EXTERNAL, Bars processed: 6384 Diagnostics: should_enter() returned a side 0 times over 6380 evaluated bars -> your ENTRY CONDITION never triggered. Loosen the entry logic / thresholds. Ensure your strategy produces trades with the given data and parameters.

Outcome Summary

Two perpetuals whose funding is computed from the same index premium and formula have no harvestable funding spread regardless of differing collateral currency — verify the economic input actually diverges before building a relative-value book; and a cross-leg edge is untestable if the second leg's per-instrument funding series isn't provided on the sandbox's supplementary-data path.

It was abandoned for 'verifying_stuck_3_timeouts': the verification stage failed to complete after three timeouts across three iterations, so it never cleared to a valid backtest, review, optimization, or analysis. Substantively, the premise was also refuted by the developer's direct measurement — Binance computes both contracts' funding from the same BTC index premium with the same formula, so no persistent inter-margin differential exists to harvest and the tiny spread is far below the two-leg round-trip cost.

A market-neutral, cross-venue relative-value strategy that shorted the richer-funding and longed the cheaper-funding of Binance's two BTC perpetuals — the COIN-margined inverse (BTCUSD_PERP.BINANCE_CM) and the USD-margined linear (BTCUSDT.BINANCE) — in equal USD notional (delta ≈ 0) on the 8h funding grid, aiming to harvest the persistent inter-margin funding-rate spread until it normalized.

The developer's own measurement over 305 matched 8h prints (2025-05-25 to 2026-05-02) showed the spread mean was +0.0038 bps — indistinguishable from zero — and simulating the exact rules harvested only ~0.0022-0.0036% gross per trade against a ~0.20% pair round trip (55-200x underwater), with total accrual of ~0.23% over 11 months (about one round trip). It also could not run in the sandbox: the USD-M leg's funding (funding_rates_by_instrument) is structurally absent from the inline Layer-3 supplementary path, so the pair never opened.

Backtest and paper results are hypothetical. Trading involves risk of loss.