# Alpha, beta and R²: is the edge real?

> Learn how alpha, beta, R² and the alpha t-stat tell skill from luck, worked step by step on real AVUV and IWN scorecard data. Free, on real ETF data.

Data as of Nov 6, 2026. Rebuilt every trading night.

## Alpha, beta and R²: is the edge real?

Data as of Nov 6, 2026. Rebuilt every trading night.

From Oct 4, 2019 to Nov 6, 2026 , AVUV's alpha was 4.2% a year, but its t-stat of 1.85 fell short of the 2.

Advanced Portfolio Design · Lesson 1 of 5 1 2 3 4 5

Set an amount. See the one-year pay AVUV earned above bills: what beta predicts, alpha, and the total. Amount (hypothetical) $10,000 On $10,000: What beta alone predicts $1,095, Alpha: beyond beta (4.2% a year, t-stat 1.85) $415, AVUV above bills $1,511. Hypothetical amount. Beta, alpha and returns are real marks against IWN, Oct 4, 2019 to Nov 6, 2026, 370 weekly returns. Step 1: beta. AVUV's beta to IWN is 1.04 , a little more than its benchmark. Step 2: what beta predicts. If AVUV were just IWN with 1.04 times the risk, it would have earned 1.0442 × 10.49 = 10.95 points above bills. Step 3: alpha. It earned 15.11 . So 15.11 − 10.95 = 4.16 points a year came from something beyond beta. The scorecard, fitted weekly, prints +4.2% . Most of the 4.7 -point lead is alpha. Step 4: does the yardstick fit? R² is 0.95 : IWN explains 95% of AVUV's weekly moves. Step 5: chance? The t-stat is 1.85 , under 2. With the scorecard's unrounded alpha of 4.15 , the margin is 4.15 ÷ 1.85 = 2.24 points . Two margins is 2 × 2.24 = 4.48 , so the range is 4.15 − 4.48 = −0.3 to 4.15 + 4.48 = +8.6 points . The scorecard prints −0.25% to +8.56% . The range includes zero : "Not significant".

### The question

My fund beat its benchmark by almost 5 points a year. Is that skill or luck?

### The idea

- Three scorecard numbers answer it together. - Beta: how far the fund moves for each 1% its benchmark moves. A beta of 1.2 means it tends to rise 1.2% when the benchmark rises 1%, and fall 1.2% when it falls.

- Alpha: the yearly return beyond what beta predicts. Some of a lead is just extra market risk. Alpha is what remains.

- R²: the share of the fund's weekly moves the benchmark explains, 0 to 1. Near 1, it is a good yardstick.

- Alpha is an estimate with a margin of error, its standard error . The t-stat is alpha divided by that margin. Below 2, the alpha could be chance. The scorecard also prints a 95% range for alpha. If it includes zero, zero alpha cannot be ruled out.

- All three use excess return : the return above Treasury bills, the near-riskless rate ( Intermediate lesson 2 ). Only that part pays for risk.

### Worked example

The 5 steps beside the drawing at the top work this through on ETFIQ data. Point at a step to light the part of the drawing it describes.

AVUV is an actively managed Avantis fund of cheap-looking small US companies. Its benchmark is the Russell 2000 Value Index, measured through IWN, a fund that tracks it. Window: Oct 4, 2019 to Nov 6, 2026 , 370 weekly returns. (Not a pick.)

| | AVUV | IWN |
| --- | --- | --- |
| Return a year | 15.28% | 10.61% |
| Above Treasury bills | 15.11 points | 10.49 points |

What that means: over this window the data cannot tell skill from luck yet. Another year or so of the same lead would push the t-stat past 2.

### The common mistake

"A big, significant alpha proves skill." Check R² first. SMH, a semiconductor fund, is measured against world stocks (ACWI). From Dec 30, 2011 to Nov 5, 2026 , its alpha was +11.7% a year with a t-stat of 2.71 : significant. But its R² was 0.63 . World stocks explain under two thirds of a chip fund's moves.

So that "alpha" is mostly a one-industry bet. Real return, but not proof of skill.

A second mistake: ignoring the window. A short one makes the range wide.

### Check your understanding

Pick an answer to see why. Two of three right passes the lesson.

1. A fund's beta to its benchmark is 1.2. The benchmark earned 8 points above bills; the fund earned 11. Roughly what is its alpha? A 3 points B 11 points C 1.4 points D 9.6 points Answer: C. 1.4 points. beta predicts 1.2 × 8 = 9.6; 11 − 9.6 = 1.4. Not A: 3 ignores the extra risk. 2. A fund shows alpha of +3% a year with a t-stat of 2.4, but R² against its benchmark is 0.40. What is the fair reading? A Proven skill B The alpha is chance C The fund has no risk D The alpha passes the chance test, but the benchmark may be the wrong yardstick Answer: D. The alpha passes the chance test, but the benchmark may be the wrong yardstick. 2.4 passes the test, but R² of 0.40 means the benchmark explains little. Not A: a poor fit mostly measures the bet itself. 3. An alpha's 95% range is −1% to +5%. What can you say? A An alpha of zero cannot be ruled out B The alpha is positive C The fund lost money D Beta is above 1 Answer: A. An alpha of zero cannot be ruled out. the range crosses zero. Not B: the middle is positive, but zero is allowed.

### Takeaways

- Alpha is the return left after beta, not the whole lead.

- Check R² first: alpha means little against a poor-fit benchmark.

- A t-stat under 2, or a range crossing zero, means the edge is unproven.

### Do it on ETFIQ

Test the alpha of a fund you hold Your fund

### Where to next

---

Source: ETFIQ, independent ETF data. Recomputed nightly from public filings and prices.
Cite: ETFIQ, Alpha, beta and R²: is the edge real?, data as of Nov 6, 2026. https://etfiq.com/academy/advanced-portfolio-design/alpha-beta-r-squared/
Licence: https://etfiq.com/license/ Data ETFIQ computes is free to use with attribution: ETFIQ (etfiq.com). Issuer data stays under its owner's terms.
HTML: https://etfiq.com/academy/advanced-portfolio-design/alpha-beta-r-squared/