Definition
If you regress a strategy's returns against the market's (for Botty: BTC buy & hold), you get:
R_strategy = alpha + beta × R_market + residual
Alpha is the intercept: the return that remains when the market is flat (R_market = 0). It is the portion that cannot be explained by simply riding along with the market — genuine skill / structural edge rather than disguised beta.
Why this matters for strategy evaluation
The most common self-deception in crypto trading: a strategy looks profitable but is in truth just disguised long BTC (high beta, zero alpha). In a bull market almost everything makes money — that is beta, not alpha. The hard question: does the strategy still earn once you strip out the market exposure?
Real alpha is market-neutral: it pays off independently of BTC's direction. Delta-neutral funding carry and the variance risk premium (VRP) are alpha candidates; a trend-following strategy that only works in a bull market is usually beta.
How Botty uses it
The edge-correlation audit (backtesting/edge_correlation_audit.py) separates exactly this: PTJ trend turned out to be effectively BETA (correlation 0.91 to buy & hold) — no independent alpha, it belongs in a separate return/beta sleeve. Funding carry and VRP, by contrast, are market-neutral (near-zero correlation to everything) — genuine alpha that forms the market-neutral core. The cross-sectional approach confirmed market-neutrality via a BTC beta of 0.038.
Limits
- Alpha is relative to the chosen market model. What looks like alpha against BTC may be exposure to another factor (liquidity, vol, funding regime).
- Often vanishes out-of-sample. In-sample alpha is cheap; robust alpha survives walk-forward + cross-asset.
- Can be hidden tail risk. Some "alpha" is just a small premium collected before a rare large loss.