Alpha is the intercept of the regression that gives you beta — the portion of your return that isn't explained by market exposure. If your strategy returned 15% with a beta of 1.0 against SPY, and SPY returned 10%, your alpha is 5%.
Alpha is the academic version of 'edge.' It's the answer to: 'after stripping out the part that's just market exposure, is there anything left?' Persistent positive alpha is rare and is what professional allocators actually pay for. Most retail strategies have positive raw returns and negative alpha — they're just long beta.
The honest framing: alpha is a one-period number, and on small samples it's noisy. A 12-month alpha of 5% with a t-statistic of 0.8 is statistically indistinguishable from zero. Real edge shows up in t-stats above 2 over multi-year windows — and almost no retail journal contains a long enough sample to confirm or deny edge at that confidence level.
Alpha is also factor-sensitive. The 'alpha' you see against SPY may disappear when you regress against a multi-factor model that includes size, value, momentum, and quality. Most traders who think they have alpha actually have undiversified factor exposure — useful information, but not the unique skill they assumed.