Stream 03
Crypto Trend
BTC, ETH, and the top alts. Trend-following with order-flow filters.
Process
Three timeframes have to agree.
Crypto is the loudest market we trade — twenty-four hours a day, no circuit breakers, no closing print to anchor on. Loud markets reward filters, not opinions. Our entry rule is mechanical: the 1-hour, 4-hour, and 1-day trend must all point the same way before any size goes on.
The trend filter is a regime classifier built on log-returns and realized volatility. On top of that, a CVD-based aggression filter requires that aggressive market-buy flow confirms the trend before entry. If price is above the moving structure but the tape is being absorbed by passive sellers, we wait. We have learned not to fight cumulative volume delta.
Stop is 1.5R below the entry, where R is one ATR on the 4-hour. After +1R unrealised, the stop trails by ATR on every new high. We do not target a fixed multiple — we let the trend run until the trail is hit.
Universe
Six majors. Spot only. No leverage.
Spot only on Bybit Demo through walk-forward. No futures, no perps, no funding-rate games in v1. Once the live track is two quarters deep we will revisit the question of leverage — but only if walk-forward shows the unlevered Sharpe earns the right to it.
Status
In walk-forward.
Four scanners run in parallel — trend on price, trend on volume, CVD aggression, and a regime filter. They were ported from our prior platform (Investripe) where they ran on a 1-minute Binance feed for a year. The walk-forward at onecap re-validates them on Bybit data and on a fresh out-of-sample window starting 2026-01.
Expected behaviour, based on prior platform: roughly half the volatility of BTC with a similar long-run return, fewer drawdowns, and a long tail of small losers punctuated by occasional outsized winners. Hit-rate is in the low 40s — slugging average is what carries the strategy.
We will start with paper capital on a Bybit Demo account. Live capital is gated on a clean walk-forward through 2026-Q3 and a sign-off from the desk on the realised slippage versus the model.
Risk frame