Graveyard / Basis and spreads / Binance perps
Crypto pairs trading (statistical arbitrage)
Do pairs like BTC/ETH move together enough to trade the gap?
The idea
If two coins usually move together, buy the cheap one and short the rich one when the gap widens, and close when it shrinks. The relationship is called cointegration.
How we tested it
| Data | BTCUSDT and ETHUSDT perpetuals on 5-minute bars with three signal formulations. Follow-up screens: four altcoin pairs in shared cohorts (LTC/BCH, ATOM/DOT, AAVE/UNI, SOL/AVAX), and all 6 pairs among BTC, BNB, SOL and XRP using an Engle-Granger test on the full period, the first 80% and the last 20%. |
|---|---|
| Period | Two years of hourly data for the second screen. First test not stated. |
| Costs | Not stated. |
| Holdout | In-sample 80%, out-of-sample 20%. |
| Control | None. |
Result
| BTC/ETH signal formulations that reverted | 0 of 3 |
|---|---|
| Altcoin pairs that cointegrate | 0 of 4 |
| Major pairs that cointegrate (full period) | 0 of 6 (p 0.42 to 0.91) |
| Best pair | BTC-XRP, p 0.42 |
Why it failed
The relationship between ETH and BTC broke down over the test window. Altcoin pairs in the same narrative cohort were just as unstable, and no pair was close on all three slices. Major coins move together because of a shared market factor, not because of a stable pairing.
What would make it worth retesting
Only with a different pair that actually passes a cointegration test. For the altcoin pairs the answer is no.
Past results of a backtest say nothing certain about the future.