This article explains what alpha and beta mean and how they are measured. It is educational and general, not personalised investment advice, and not a performance claim about any specific strategy. Where it refers to historical results, past performance does not predict future results.
“We beat the market” is the easiest claim in finance to make and the hardest to mean. A strategy can finish a great year far ahead of an index and still have added nothing, because most of what looks like outperformance is just the market itself, dialled up. The words that separate the two are alpha and beta, and almost everyone uses them loosely. Here is what each one actually is, how the two are pulled apart with a single regression, and why the difference is the only thing worth paying an active manager for.
Beta is the part of a strategy’s return that comes simply from its exposure to the broad market, the return you could have bought cheaply through an index fund. Alpha is what is left over: the excess return that remains after accounting for that market exposure, the piece attributable to skill rather than to the market rising. Beta is the rising tide; alpha is whether you actually swam.
What alpha and beta are
Start with beta, because it is the easier of the two and the one that does most of the work. Beta is sensitivity to the market. A beta of 1 means a strategy tends to move one-for-one with the index: the market gains 10%, it gains roughly 10%. A beta of 1.5 means it amplifies the market, up more in rallies, down more in selloffs, which is, mechanically, what leverage does. A beta near 0 means the strategy moves more or less independently of the index, for better or worse. The crucial point is that beta is not skill. Anyone can buy beta, cheaply, by holding the index. Anyone can buy more beta by adding leverage. A return that is purely beta is a return you did not need a manager to produce.
Alpha is the residual. Once you have stripped out everything a strategy earned just by being exposed to the market, alpha is the return that is still standing, the value added (or destroyed) by the specific choices that strategy made. Positive alpha means it earned more than its market exposure alone would explain. Negative alpha means it earned less, which is the uncomfortable but common case once costs are included. Alpha is the part you are actually paying for, and it is far rarer than the marketing of the industry would suggest.
How a single regression pulls them apart
The reason these two ideas are precise rather than poetic is that they fall straight out of a line of best fit. Take a strategy’s returns over many periods and plot them against the market’s returns over the same periods. Fit a straight line through the cloud of points. That line has two numbers, and they are exactly alpha and beta.
The slope of the line is beta, how much the strategy moves for each unit the market moves. The intercept, where the line crosses zero market return, is alpha, the return the strategy delivered on average when the market did nothing, the part the market cannot explain. This is the framing behind the Capital Asset Pricing Model, and the intercept measured this way has a name: Jensen’s alpha. It is not a vibe or a marketing adjective. It is a coefficient you can compute, with a standard error attached, from a benchmark and a return series.
Two consequences follow immediately. First, a high raw return tells you almost nothing on its own, a strategy can post a huge number with a steep beta and a flat or negative intercept, meaning the market did the work and the manager arguably subtracted value. Second, alpha is defined relative to a benchmark. Change the index you regress against and the split between alpha and beta shifts. An honest decomposition therefore has to name its benchmark and defend it, not quietly pick the one that flatters the intercept.
Why separating them matters
Here is the practical version of the whole problem. A strategy soars 60% in a roaring bull market. Impressive, until you regress it and find a beta of 2 and an alpha of essentially zero. It did not beat the market; it was the market, levered twice. You could have replicated almost all of it with an index fund and a margin loan, for a fraction of the fee. The headline return was real, but it was beta in a costume, and in the next downturn that same beta of 2 will hand back the gains twice as fast.
This is why genuine alpha is the thing you actually pay an active manager for, and why it is so rare. Beta is a commodity now, a few basis points buys you all you want. Anything charging more than commodity prices is implicitly claiming alpha, and most of the time that claim does not survive a regression. The uncomfortable finding of decades of performance studies is that, after fees, the majority of active strategies show alpha indistinguishable from zero or worse. The exercise of separating alpha from beta is, more than anything, a way of checking whether there is anything there at all, or just an index fund with a story and a higher bill.
How alpha is measured, and how it fools you
The base case is the regression above: Jensen’s alpha, the intercept from regressing strategy returns on benchmark returns. But a plain straight line misses one thing it is worth knowing about, market-timing skill. A manager who is good at timing holds more market exposure before rallies and less before selloffs, so their effective beta changes with the market rather than staying fixed. A straight line cannot see that; it averages the changing beta into a single slope and can misattribute real timing skill to the intercept, or miss it entirely. To detect it you fit a curve, the Treynor-Mazuy market-timing regression adds a squared market-return term, and a positive coefficient on it is the statistical signature of a manager whose exposure bends in the right direction. It is the difference between measuring whether you swam and measuring whether you also caught the tide at the right moments.
And then the warning that hangs over every alpha figure ever published: torture enough strategies and one of them will show alpha by pure chance. An intercept that looks impressive in isolation can be a fluke of a short sample or the survivor of a thousand quietly discarded variants. A measured alpha is only as trustworthy as the discipline behind it, the length of the sample, whether the benchmark was fixed in advance, and whether the result clears a bar high enough to account for all the strategies that were tried and failed. Why that bar is far steeper than the textbook one is the subject of statistical significance, and why an in-sample alpha is the easiest number in the world to manufacture by accident is the argument of why backtests lie.
How Shishin frames alpha vs beta
Most published track records quote a headline return and let you assume it is skill. Shishin does the opposite: it runs the decomposition on its own track and shows the working. The live, paper-traded record is regressed against a market benchmark to separate the part that is genuine edge from the part that is simply exposure to a rising tape, a Jensen’s-alpha intercept for the skill component, and a Treynor-Mazuy curve to ask whether any of the edge comes from the regime-aware posture-switching architecture leaning into and out of the market at the right moments rather than from a fixed bet. The aim is not to advertise a flattering number. It is to publish whether the returns are alpha or beta, with the benchmark named, so the question can be checked rather than believed.
That is the differentiator worth being honest about. Plenty of sources can recite the textbook definition of an intercept. Far fewer turn the regression on their own results and tell you how much of their edge survives it. The discipline of reporting a return next to its risk and its drawdown, the argument of the return, Sharpe and drawdown trinity , extends naturally to reporting a return next to its beta: in both cases the second number is what tells you whether the first one means anything.
So, is it alpha or just beta?
The honest answer is that you cannot tell from the headline, and anyone who quotes only the headline is, intentionally or not, hiding the answer. Run the regression. If the return mostly disappears into the slope, it was beta, market exposure you could have bought cheaply, possibly levered, dressed up as skill. If a real, significant intercept survives after the market is accounted for, and survives the multiple-testing scrutiny, that is alpha, and it is the rare and valuable thing the whole exercise exists to find.
One final, deliberate caveat. A handsome alpha measured on the same history a strategy was built on is the weakest evidence there is, because that is precisely the number an over-fitted strategy is engineered to produce. In-sample alpha can be an artefact; the regression run forward, on a live track the strategy could not have seen in advance, is the only version that counts. That is why the decomposition that matters is the one performed out-of-sample, in public, not the one a backtest can always be coaxed into showing. Alpha is real, it is rare, and it is worth paying for. It is just almost never as large as the brochure, once you have subtracted the tide everyone was floating on.
Sources & further reading
- Sharpe, W. F. (1964). “Capital Asset Prices: A Theory of Market Equilibrium under Conditions of Risk.” Journal of Finance, 19(3), 425 to 442.
- Jensen, M. C. (1968). “The Performance of Mutual Funds in the Period 1945 to 1964.” Journal of Finance, 23(2), 389 to 416.