What is "alpha"? Returns vs. a benchmark

The core idea

Three terms first, because everything below leans on them:

• A benchmark — the passive alternative you compare your strategy against. Usually a broad market index like the S&P 500.
• SPY — the biggest ETF (a fund that trades like a stock) tracking the S&P 500. When traders say "the market," they often mean SPY.
• Beta ($\beta$, the Greek letter beta) — a number measuring how much your strategy moves when the market moves. Beta 1 means you move one-for-one with the market. Beta 0.5 means about half as much.

You can always earn returns by simply taking risk. Buy SPY, hold it, take full market beta. Over time you collect the equity risk premium — the extra return stocks pay over safe assets as compensation for their risk. That's not skill; it's exposure.

Alpha ($\alpha$, the Greek letter alpha) is the part of your return that exists beyond what your benchmark exposure explains. It's the return attributable to skill. It separates real signal from "I just took beta."

This single concept is what splits the finance industry into:

• Passive (sell beta cheaply via ETFs — no skill claimed, low fees)
• Active (claim to generate alpha, charge fees for it)

The cleanest definition (Jensen's alpha)

Run a regression — a statistical procedure that finds the straight line best fitting a cloud of data points — of your strategy's excess returns on the benchmark's excess returns. ("Excess return" means your return minus the risk-free rate — what a safe government bill would have paid you anyway.)

$r_t - r_f = \alpha + \beta (r_{m,t} - r_f) + \varepsilon_t$

Read the symbols aloud first. $r_t$ ("r sub t") is your strategy's return in period $t$ — the subscript $t$ indexes time, like a loop variable over days. $r_f$ ("r sub f") is the risk-free rate. $r_{m,t}$ ("r sub m, t") is the market's return in the same period. $\varepsilon_t$ ("epsilon sub t," a Greek letter) is that period's leftover noise.

In words: your return above the risk-free rate equals a fixed amount (alpha), plus beta times the market's return above the risk-free rate, plus random noise. The regression finds the alpha and beta that best fit your data.

Then:

• $\alpha$ (the intercept — where the fitted line crosses the vertical axis) = your return when the market returned only the risk-free rate. The "free" part of your return.
• $\beta$ = your sensitivity to market moves (the slope of the line).
• $\varepsilon_t$ = residual noise — whatever's left over that the line doesn't explain.

If you ran this on daily data, the alpha comes out per day. Annualize it — convert it to a per-year figure — by multiplying by 252, the number of trading days in a year.

This was introduced by Michael Jensen in 1968 and remains the standard definition.

Example interpretation

A strategy with:

• Annual return 18%
• $\beta = 0.7$ (less market-sensitive than SPY)
• Market return 12%
• Risk-free rate 4%

Expected return from beta exposure alone: $4\% + 0.7 \cdot (12\% - 4\%) = 9.6\%$. (The $\cdot$ dot means "times." Read it: 4% plus 0.7 times the gap between 12% and 4%, which is 9.6%.)

Actual return: 18%. Alpha = 18% − 9.6% = 8.4%. A meaningful positive alpha.

By contrast, if you ran a strategy with:

• Annual return 18%
• $\beta = 1.5$ (leveraged market exposure)

Expected from beta: $4\% + 1.5 \cdot (12\% - 4\%) = 16\%$. Actual: 18%. Alpha = 2%. The same 18% return is much less impressive once you adjust for beta.

The information ratio

Alpha alone can be misleading, because you can earn 10% alpha with extreme noise around it. The right risk-adjusted version is the information ratio:

$\text{IR} = \frac{\alpha}{\sigma(\varepsilon)}$

In words: the information ratio equals alpha divided by "sigma of epsilon" — your annualized alpha divided by how much the leftover noise spreads around, measured as the standard deviation of the regression residuals. That spread has a name: your tracking error — how wobbly your returns are around what beta alone predicts. IR is the Sharpe ratio of your alpha, and it's what serious active managers report.

Benchmarks:

• IR ≈ 0.5: respectable active manager.
• IR ≈ 1.0: top decile (top 10%) of active managers globally.
• IR > 1.5 sustained: extraordinary.

Why is positive alpha so hard?

In an efficient market, the price already reflects everything publicly known. Every trade has two sides. If one side has positive expected alpha, the other side must have negative expected alpha. Alpha is approximately zero-sum — one player's gain is another's loss — before fees. After fees and transaction costs it's strictly negative-sum: the players as a group lose money to costs.

This isn't a theorem. It follows from efficient markets plus simple arithmetic: all investors together are the market, so their combined return equals the market return. Half of all active dollars beat the benchmark; half don't. After fees, meaningfully less than half.

This is why passive index funds are the rational default for retail investors with no specific edge. It's also why active management has been losing AUM — assets under management, the money investors have placed with them — to passive for two decades.

What edges produce real alpha?

Realistic categories of retail edge:

1. Behavioral biases — big institutional investors can't take positions retail can. They face size limits, mandate restrictions (rules on what a fund is allowed to hold), and end-of-quarter window-dressing (dressing up holdings to look good in reports). Slow-money behavioral patterns persist because of this.
2. Long-horizon discipline — most institutions can't tolerate 3 years of underperformance, even on strategies that work over 10 years. You can.
3. Specific niches — small-caps, microcaps, certain emerging markets, illiquid corners (assets that trade rarely) with little institutional attention.
4. Tax-efficient structuring — direct indexing, harvesting losses to reduce taxes.

What does NOT produce retail alpha:

• Reading the news faster than HFTs — high-frequency trading firms whose machines react in microseconds (you can't).
• Picking stocks based on public earnings announcements (priced in within seconds).
• Following the same indicators as 10 million other retail traders.

We'll see specific strategies in Track D5. The framing here: for any strategy, ask "what's the source of the edge?" and "why hasn't it been arbed away?" — meaning, why haven't professionals already traded the profit out of existence? If you can't answer both, you don't have edge.

Try it

Reproduce the lesson's worked example in code:

Explain it back

Two sentences in your own words: what's the difference between "returns" and "alpha," and why does the distinction matter for evaluating a strategy?