Unlock This Note

When shorts are paying longs just to stay in the trade, the crowd is the fuel for the next move — not the signal. This strategy fades extreme BTC funding rates: go long when shorts are crowded, go short when longs are crowded, and step aside once funding normalizes. It runs on BTCUSDT perpetual 8-hour bars, one bar per Binance funding settlement. Backtested 2025-09-30 to 2026-09-24 (about 12 months): +37.9% total return while BTC buy-and-hold lost 25.8%, Sharpe 1.73, max drawdown −14.5%. It's a community-listed paid strategy in the strategy library, priced at 20 TWD. Why would extreme funding be a contrarian signal? Perpetuals never expire, so funding is what tethers them to spot. Positive funding means longs pay shorts; negative means shorts pay longs. When funding hits an extreme, one side is crowded enough to keep paying for the privilege of holding. Once that side starts to unwind, the unwinding itself pushes price the other way. The strategy doesn't forecast direction — it only acts when positioning is lopsided. The rules: four thresholds, exits checked first Long: enter at funding ≤ −0.011% (crowded shorts), exit when it recovers to −0.001%Short: enter at funding ≥ 0.009% (crowded longs), exit when it falls back to 0.005% On every bar, exits are evaluated before entries, so a stale position is closed before a new signal can overwrite it. Every decision comes from funding alone; price patterns play no part. Backtest results (BTCUSDT perp, 8-hour bars, 2025-09-30 to 2026-09-24) MetricValueTotal return+37.9%BTC buy-and-hold, same period−25.8%Sharpe1.73Sortino1.39Max drawdown−14.5%Trades98Cumulative fees4.9%MCPT p-value0.0515 (2,000 permutations) Net of 0.05% per side (Binance VIP0 taker). Passed our look-ahead check: truncating the data leaves past signals unchanged. We re-ran these numbers on our own backtest engine. Backtest results, not a guarantee of future performance. Read this before you buy: four weaknesses We put it through our checks. Here's what didn't come out clean. 1. One year of data, all of it a downtrend. Funding history in our source starts 2025-09-30, so the whole backtest is those 12 months — and BTC fell 25.8% over them. Whether the contrarian logic, and the short leg in particular, holds up in a bull run or a range is untested. 2. The parameters sit on the peak. Shift each entry threshold up and down (long −0.007% to −0.015%, short 0.006% to 0.01%) and Sharpe ranges from 0.3 to 1.73 — the current setting is the single best cell. The surrounding 3×3 neighborhood averages about 1.45, or 0.84 of the peak. Solid neighbors, but the thresholds were tuned on the same data they're scored on, with no out-of-sample period. Expect live results closer to the neighborhood average, or lower. 3. Significance is borderline. Across 2,000 permutations, p = 0.0515 — just short of the conventional 0.05. Roughly 103 random shuffles of the signal did as well or better. Unlikely to be pure luck, but not proof either. 4. The short leg can't tell crowded from very crowded. Positive funding in the data source is capped at 0.01%, and the market never exceeded it during this period. The short leg sees "near the cap," not how extreme the crowding actually is. The sensible way to run it: small size, one contrarian leg within a broader book, reconciled against the backtest once it's live. Not a single bet. Buying the strategy is easy. Running it every 8 hours isn't. Funding settles three times a day; miss one check and you miss a signal. Then there's the exchange connection, order routing, position tracking, and staying up through the night. Blave Agent handles that chain — and there's no LLM in the order loop: signals follow the rules, orders follow the signals. 20 TWD buys the strategy. Your top-up buys the execution that keeps it live. Live in three steps 1. Install Blave Agent2. Top up from 300 TWD to start your Blave Agent3. Buy "BTC Funding Rate Contrarian" in the strategy library and tell your agent "deploy this strategy" → one-click deploy, runs on schedule Start with a 300 TWD top-up, then pay as you use: server & data fee from 1,440 TWD/month (30-day basis, 2.0 TWD/hr), plus AI usage. Run it on paper trading first and reconcile before going live. Install Blave Agent and put this strategy to work today →