BNB Perp Discount — 1h

Status: Regime entry: a record only — worked in one test window, not the other · Regime #11 · Listed 2026‑10‑14 Instruments: BNBUSDT perpetual only · Timeframe: 1h · Direction: long only Frequency: ~27 trades/yr · Fees: 0.05%/side plus real funding

Why it is in the regime section, not the library

It failed the six‑check bar. It is listed under the regime bar, which asks four things of one of the two test windows: a net profit after fees and funding, still a profit without the best trade, a result above 95% of 400 random‑entry runs with the same exit, and at least 20 trades. It clears all four from May 2021 to February 2024 (68 trades), narrowly, and in a way to understand before anything else: one trade made nearly all of it. From February 2024 to October 2026 it lost money. Everything below the setup is disclosure, not a pass.

Where it fails the full bar:

• Two windows: −16.9% from February 2024 to October 2026 (PF 0.74), at the 20.5th percentile of random entries.
• Strip‑best, only just: one trade, bought on 23 May 2021, made +46.1% of the window's +49.1%. Without it the window made +3.0%.
• Cross‑asset: BTC lost in both windows (−28.3%, −8.1%), and so did ETH (−50.8%, −20.8%), ETH by more than random entries would have (1.8th and 14th percentiles).
• Drawdown: 23.1 summed points in 2021–24 and 22.4 in 2024–26, against a bar of 20.

It came out of sweep #15, pre‑registered on 2026‑10‑11: three free positioning and fear channels in front of RSI(14)'s turn up from 30 (the perpetual's discount to spot on BTC, ETH and BNB; options‑implied volatility and Bitfinex's margin shorts on BTC and ETH) in two windows. That is 14 window tests. With a 95th‑percentile bar, about one window test in twenty clears by luck alone. Three cleared the percentile: this one (98.5th), whale shorts on BTC in 2024–26 (99.5th), and options panic on BTC in 2024–26 (98th) on 13 trades, under the 20‑trade minimum.

Why the app can't run it

Its gate reads Binance's premium index, the perpetual's price against its spot index hour by hour, ranked against the last 28 days. The strategy language reads the funding rate, an eight‑hour average of that premium clamped near zero, but not the premium itself, and has no rolling percentile, so the gate can't be written; the RSI trigger and the trailing exit are ordinary.

The setup

Market: crypto BNBUSDT perpetual, 1h trading timeframe. Longs only. Gate: Binance's premium index (the perpetual against its spot index, hourly close) at or below its trailing 28‑day 10th percentile, the perpetual at an unusual discount. Entry (1h): RSI(14) crosses above 30 while the gate is open. Exit: trailing stop 2×ATR(14) below the highest price since entry. Fees: 0.05%/side (Binance USDT‑M taker) · funding charged and credited at every settlement held through.

Evidence: two windows on Binance perpetual candles and the premium index

Returns are per‑trade percentages summed, like every return in the research log. Holding BNB lost 51.8% in the first window and made +156.0% in the second.

• Random entries, 2021–24: 400 runs of the same exit on random bars, with the real funding billed, had a median of −7.0% and a 95th percentile of +29.2%.
• Random entries, 2024–26: a median of −5.5% and a 95th percentile of +21.9%; the real −16.9% sat at the 20.5th percentile.
• Funding: 59 of its 145 trades were entered while funding was negative. They made +37.2%; the other 86 lost 5.0%.

The record by year

*2021 from May, 2026 through September. Each trade counts in the year it was entered.

• One trade is most of it: bought on 23 May 2021, it made +46.1%. Without it, the two windows together lost 13.9%.
• It has lost money since 2022: −24.4% over 96 trades from January 2023.
• Last high: November 2022, at +58.6% summed since May 2021. It ended September 2026 at +32.2%, 26.4 points below.
• Last twelve months (October 2025 to September 2026): 38 trades, −16.1%.

What the result depends on

Nothing else was run: sweep #15 ran its fee stress, its channel‑lag check and its neighbors only for a strategy that cleared gates 1–5 on two assets, and none did (research log, sweep #15, “Results”).

Before you rely on it

• One trade of +46.1%, bought on 23 May 2021, is nearly all of its 2021–24 result: without it the window made +3.0%.
• It has lost money since: −16.9% in 2024–26, and −24.4% over 96 trades since January 2023.
• Drawdowns of 23.1 and 22.4 summed points, against the library's 20.
• Its trades with funding negative made +37.2% and the other 86 lost 5.0%. The research log's reading: the eight‑hour averaging that turns the premium into funding looks like it does real work, filtering out discounts that do not last.

Every number here can be re‑run: npx tsx scripts/research/sweep15-rsi-crypto-channels.ts for the windows and the null. The record by year, the last twelve months and the last high are the same runs' trades, grouped by the date they were entered.