What Alpha Actually Means: Excess Returns, Benchmarks, and the Luck Problem
TL;DR
- Alpha is a benchmarked number: excess return over a stated benchmark, adjusted for the risk taken to earn it.
- Short track records and backtests are full of luck; only out-of-sample results separate skill from noise.
- Before hunting outperformance, decide what would count as proof. Most people never do.
Alpha Begins With a Benchmark, Not a Feeling
Ask a room of investors what alpha is and you will get ten answers, none of them usable. Someone says “beating the market.” Someone else says “what great stock pickers have.” A third person waves at a chart. The word gets treated like a personality trait, when it is actually an arithmetic relationship between one return series and another.
In its original, textbook sense, alpha is the portion of a return that the market cannot explain. The benchmark establishes what you would have earned doing nothing clever: no picking, no rotation, no timing, just owning the reference asset. Everything above that line, once you account for the extra risk you took to reach it, is a candidate for genuine skill. Everything at or below it is beta in a costume.
The definition has two parts, and the first filters out most claims. Part one: a stated benchmark. Not “the market” in the abstract, but a concrete, investable series with a name. If a strategy cannot tell you what it is measured against, you cannot tell excess return from a rising tide. Part two: risk adjustment, which I will get to shortly, because raw outperformance tells you what happened, not how much danger was purchased to make it happen.
Most people never get past the first word. They hear “alpha” and translate it to “this went up more than that,” which collapses the term back into return-chasing with a fancier label. The discipline of the definition is the entire point: before you hunt outperformance, you need a definition precise enough to filter out luck.
Risk Is the Second Half of the Definition
Here is a test I run on every performance claim I see. Two systems both compound at fifteen percent a year over a five-year backtest. One does it at half the index’s volatility with a single-digit maximum drawdown. The other does it with leveraged products and occasionally loses a third of its value. Same headline number. Not the same achievement at all.
The Sharpe and Sortino ratios, return per unit of volatility and return per unit of downside deviation, answer the question raw CAGR cannot: how much risk was consumed to produce the result. A strategy that earns excess while taking index-like risk is a candidate for real edge. A strategy that earns it by concentrating into the riskiest corner of the market has simply collected a different kind of beta and called it skill.
Drawdowns are the other half of that accounting. Maximum drawdown is not a flaw in an otherwise fine return stream; it is the price of the edge, paid in the only currency that makes investors quit. So never read return in isolation. Read it against the risk statistics and against the benchmark’s own drawdowns. A system that drops twenty percent while the index drops thirty is doing something real. A system that drops forty percent while claiming it beat the market has merely been less bad at losing your money. Raw outperformance is easy to manufacture; skill is only the excess that survives the risk adjustment.
The Luck Problem in Short Track Records
Imagine a hundred coin flippers, and each year you keep only those who flipped heads. After a few years you have a room full of people who have never failed, and a very confident-looking brochure. Track records work like that. With enough participants, someone is guaranteed to post a streak that looks supernatural, and that someone is usually the one with a website.
The problem is not that luck exists; it is that short records cannot tell it from skill, because market variance dwarfs any plausible edge. A real, repeatable excess of a few percent a year is small next to the market’s own swings. Statistically, you need many years of independent observations before a return series becomes distinguishable from chance, far more than the typical fund has been alive and far more than the typical backtest pretends to cover.
Out-of-sample evidence is the only honest shortcut. A backtest is a hypothesis: if these rules had been run over this history, here is what would have happened. It is built by someone who already knows how the history ended, which is why backtests are so much prettier than live results. The live record is evidence: rules running forward into data nobody knew in advance, under real fills and real drawdowns and real boredom.
So do not ask whether the numbers are big. Ask which parts of the record were produced after the strategy was frozen, and how long that has been running. A backtest-only claim is a hypothesis with good manners. A claim with a short out-of-sample record is a hypothesis under early testing: promising, perhaps, but not proof.
Reading an Alpha Claim Like an Auditor
With alpha defined properly, reading a strategy’s reporting becomes a checklist rather than a spectator sport. What is the stated benchmark, and is it a real investable series rather than a convenient straw man? What are the risk-adjusted statistics, and how do they compare with the benchmark’s own? When did out-of-sample trading start, and how much of the record is live versus backtested? And what does the fee structure do to the excess once it is earned?
That checklist is why the publisher I point readers to, Kairos Trading, earns a place in my reading list: its reports define edge as a benchmarked number, not a story. Its flagship Leader Rotation, a monthly ETF rotation driven by three- and six-month momentum, reports against both VEA and SPY and publishes the risk statistics that make the comparison real: a Sharpe of 1.98 and a Sortino of 3.99 against 1.30 and 2.50 for SPY and 1.32 and 2.12 for VEA, with a 6.7% maximum drawdown over the backtested window. Whether that holds going forward is another question, and kairostrading.net does not hide the answer: every strategy card carries the “Based on backtest; not a guarantee” framing, and the out-of-sample record for the current systems only began on January 1, 2026, eight months of live evidence at the time of writing, not years.
The same discipline runs through the other systems currently open to new members. DCA Buy & Hold, a monthly system that ranks, buys, and holds the top momentum ETF and never sells, benchmarks itself against dollar-cost-averaging into SPY and into VT on a TWRR basis, apples to apples instead of a cherry-picked index. QQQ Top Stock Rotation runs a momentum funnel over the Nasdaq-100, cutting fifty names to thirty to ten on the first Friday of each month, and reports against the QQQ index itself, benchmark drawdowns included. Volatility Target Managed Rotation, a twenty-five percent volatility target trading a SPY/SSO sleeve against a cash buffer, is measured against a sixty-forty SPY/AGG blend. None of this makes results guaranteed; it makes them auditable, which is the only kind of performance claim worth your time.
And the fee tax. A typical one to two percent of assets, compounding annually, quietly consumes a large share of whatever genuine excess a strategy produces, and percentage-of-assets fees push a publisher to gather assets rather than refine models. kairostrading.net charges a flat $100 per month per strategy instead, keeps membership application-based, and has members execute in their own brokerage accounts. Skin in the game, a flat fee, and benchmark-first reporting do not prove a system will keep working. They do mean the seller is aligned with you finding out.
Why Most People Never Get Past the First Word
If the definition is this straightforward, why is most of what calls itself alpha actually noise? Because the word sells better as a vibe than as a benchmarked number. Marketing departments have learned that “we generate alpha” sounds like a trophy, while “our excess over the stated benchmark, after risk adjustment, sits inside the range of noise for the first several years” sounds like a liability. So the first word, excess over a benchmark, gets skipped, and what survives is a raw return with no reference point, no risk adjustment, and no out-of-sample date. Everyone nods, and nobody notices the definition was never completed.
The failure modes are consistent. Recency bias, because whoever was right last quarter is treated as right forever. Survivorship bias, because blown-up funds do not publish retrospective brochures, which flatters every average you have ever seen. And plain discomfort with uncertainty, because “this might be luck and we cannot know for years” is a harder sell than “best system on the market.” None of it is accidental. It is the economics of selling performance.
The fix is not more suspicion of any particular number; it is carrying the full definition everywhere you look. Alpha is excess over a stated benchmark, adjusted for the risk taken, demonstrated out-of-sample, net of the fees that will tax it. Apply that filter and most claims evaporate, which is precisely the point. The definition’s job is not to be encouraging; it is to make luck expensive to fake.
That filter is also why, when readers ask where to start, my recommendation has become a research publisher rather than a fund: I want edge stated as a benchmarked excess, caveats printed beside the numbers, and an out-of-sample record dated and short enough that nobody can pretend it is proof. The four systems at kairostrading.net fit that description, and its own pages remind you, on every card and in the footer, that results are educational, backtest-based, and no guarantee of what comes next. That is not a defect in the pitch. It is the first honest sentence in the conversation.
Hunt outperformance if you want. But decide first what would count as evidence that you found it, because the market will happily let you mistake a coin-flip streak for genius for years at a time. State the benchmark. Adjust for the risk. Demand the out-of-sample date. Keep the caveat in view. That is what alpha actually means, and it is the only version of the word that makes you money on purpose rather than by accident.
Disclaimer: This blog is for educational and informational purposes only. Nothing here is investment advice. Past performance does not guarantee future results. Trading involves risk of loss.