Mean Reversion
mean_reversionBasket-based mean reversion signal diversified across different windows to capture broad reversal dynamics.
How the factor is constructed
Mean Reversion is a basket-based signal diversified across different windows to capture reversal dynamics and statistical arbitrage opportunities in crypto assets.
The universe consists of the most liquid and actively traded assets, identified on a rolling basis and survivorship-bias free. Positions are scaled by the inverse of rolling volatility; the factor is available point-in-time with hourly updates.
Mean Reversion
| Period | Return | BTC | Ann. vol | Sharpe | Max DD |
|---|---|---|---|---|---|
| 1 month | -2.90% | +24.10% | — | — | -5.2% |
| 3 months | -7.66% | +17.86% | — | — | -10.3% |
| Year to date | -4.78% | -11.92% | — | — | -15.1% |
| 1 year | +7.98% | -32.28% | 14.5% | 0.52 | -15.1% |
| Since inception (CAGR) | +18.43% | +42.73% | 15.4% | 1.17 | -22.6% |
Putting Mean Reversion to work
There are two ways to take Mean Reversion from factsheet to live book: overlay it as a sleeve on the portfolio you already run, or use it as a building block in a standalone multi-factor portfolio. Both paths are documented in short, runnable notebooks that work out of the box against the public demo key — no signup required.
Add Mean Reversion to an existing portfolio
Size a Mean Reversion sleeve alongside your current book and quantify what it changes: correlation to your existing returns, then CAGR, volatility, Sharpe and maximum drawdown before and after the blend. One parameterized notebook runs for any factor — swap in Mean Reversion, then bring your own daily returns as a CSV or start from the built-in demo book.
Build a portfolio from scratch
Construct a market-neutral multi-factor portfolio from the ground up: ensemble the raw signals, apply inverse-volatility weights with a per-asset cap, and backtest net of transaction costs. The default factor list reconstructs the 7 Factor Composite — Mean Reversion is one of its seven constituents, so its contribution is in the blend from the first run.
New to the platform? The five-step guide covers data access, a proof of concept, and licensing.