—

Information Theory & Network Economics

A Random Walk Down Wall Street

Burton G. Malkiel · first published 1973

Malkiel's case is that prices absorb public information so fast that beating the market after costs is mostly luck. The useful part for investors is the test: how long a record must be before skill can be told apart from chance.

The big picture

Malkiel, a Princeton economist, first published the book in 1973 and has revised it ever since; the 13th edition marked its 50th year in 2023. His argument: in a market with many informed buyers and sellers, new public information is priced within minutes, so the next price move is close to unpredictable from what is already known. A stock picker can still win, but the average active fund starts every year behind the index by its own costs, and the winners of one period rarely repeat in the next. His practical answer, which he pushed years before the first retail index fund launched, is a low-cost, broadly diversified index holding as the core of a portfolio.

Why it matters now: index concentration is high, a handful of AI-linked names drive a large share of benchmark returns, and a record that beat the index for three years can be a single sector bet that happened to be right. The book gives the question to ask of any track record, including your own: how much of it is signal, and how much would a coin have delivered?

Skill or luck — the years a record needs, and the streaks chance alone produces t = IR × √years  ·  years needed for t = 2: (2 / IR)² YEARS OF RECORD BEFORE SKILL SHOWS 25 50 75 100 0 0.2 0.4 0.6 0.8 1.0 INFORMATION RATIO (excess return ÷ tracking error) YEARS IR 0.3 → 44 yrs IR 0.5 → 16 yrs IR 1.0 → 4 yrs a record needs t ≥ 2 to be told apart from luck the smaller the edge, the longer the wait 1,000 COIN-FLIP MANAGERS, 15 YEARS zero skill · 50/50 odds every year 676 125 3 372 62 4 183 31 5 85 16 6 39 8 7 18 4 8 WINNING STREAK, YEARS IN A ROW STREAK ANYWHERE FROM YEAR 1 FEE DRAG, 7% GROSS, 30 YEARS 1% a year: −24.5% · 0.05%: −1.4%
A modest edge takes decades of data to prove, while pure chance hands a few hundred of 1,000 managers a multi-year winning streak. Costs are the one gap you can measure in advance.

The 3 strategic pillars

  1. The efficient-market argument, in three strengths

    Prices reflect available information, so trading on that information should not earn more than the market after costs. The claim comes in three strengths, and each rules out a different kind of edge.

    Weak form: past prices and volumes contain no exploitable pattern, so chart reading alone should not beat the market. Semi-strong form: all public information (earnings, news, filings) is already in the price, so fundamental research on public data should not either. Strong form: even private information is priced, which almost nobody defends because insider trading does pay. Malkiel's position is close to the semi-strong form, held as a strong approximation rather than a law.

  2. Castles in the air vs. firm foundations

    Two ways to value a stock: by what a future buyer will pay for it (crowd psychology), or by the cash it will generate (intrinsic value). Both have explained real prices; neither lets an investor win reliably.

    The firm-foundation view (the line from Graham) values a share as the present value of expected dividends and earnings, so it depends on growth and discount-rate forecasts that are often wrong. The castle-in-the-air view (Keynes's beauty-contest idea) prices what the crowd will want next, which explains bubbles from tulips to the dot-com era but gives no timing rule. The book walks through both and concludes that forecasting either one well enough, and consistently, is rare.

  3. Costs, persistence and the index

    Before costs, active investors as a group earn the market return, because together they are the market. After costs, the average must trail it, and the gap compounds.

    A 1% annual cost on a 7% gross return removes 24.5% of ending wealth over 30 years; at 0.05% the loss is 1.4%. S&P's SPIVA scorecards have repeatedly found that most US large-cap active funds trail the S&P 500 over 10 and 15 years, and that top-quartile funds rarely stay there. Malkiel's own caveats: markets do make mistakes (bubbles, crashes), some patterns such as value, size and short-term momentum have shown up in the data, but they tend to be small, unstable and eaten by costs once widely known.

What a Closelooknet reader does with it

The working use is a test you run before believing any record, a fund's or your own: take monthly returns net of costs and a fitting benchmark, compute the average excess return and its volatility (tracking error), and ask whether the ratio of the two has been held long enough to rule out luck. With an information ratio (excess return per unit of tracking error) of 0.5, a t-statistic of 2 takes about 16 years; at 0.3 it takes about 44. The pack runs that on your own data, prints the fee drag for any two cost levels, and simulates 1,000 managers with zero skill, of whom about 183 still post a five-year winning streak somewhere in a 15-year career. The mistake it prevents: switching into last period's winner on a record that chance alone produces every year.

The bridge to the Closelooknet approach

The cost side is where Closelooknet measures the argument directly: the ETF section lists benchmark funds by tracking difference (fund total return minus index total return), and the Tracking Difference 101 explains why that number says more than the published expense ratio. The skill test uses the vocabulary of the glossary: tracking error and the information ratio. The house's own portfolios publish their results next to a benchmark, which is the record this pack is built to test. For the opposite reading, Misbehaving documents where investor behaviour pushes prices away from value, and The Intelligent Investor is the firm-foundation school in its original form. The Information covers the other half of the idea: news that is already expected carries almost no information.

Action-Kit — from theory to practice

Tooling & data

What you needWhere to get itCost
Monthly return history for YOUR portfolio, a fund, and its benchmark The input for the alpha test; use total returns (dividends included) and net of all costs Your broker's statement export, or the fund's factsheet data The pack ships no market data; the sample file holds EXAMPLE_ placeholder rows only. Free
SPIVA scorecards Published counts of active funds that trailed their benchmark over 1 to 20 years, by category and region S&P Dow Jones Indices Free
Factor and benchmark returns Market, size, value and momentum return series to use as benchmarks or to check whether an excess return is a factor tilt Kenneth R. French Data Library (Dartmouth) Free
Factor regression and fund analysis A second tool to cross-check alpha, beta and factor loadings on your own portfolio or a fund Portfolio Visualizer Freemium

The formulas

  • Information ratio

    IR = mean(e) × 12 / (stdev(e) × √12), e = r_portfolio − r_benchmark
    • r_portfolio — monthly return of the portfolio or fund, net of costs
    • r_benchmark — monthly return of the benchmark
    • e — monthly excess return

    Annualized from monthly data; use 52 instead of 12 for weekly returns.

  • t-statistic of alpha

    t = mean(e) / (stdev(e) / √n) = IR × √years
    • n — number of monthly observations
    • e — monthly excess return

    |t| of about 2 corresponds to the 5% two-sided level; the pack also prints the exact p-value from the Student-t distribution.

  • Years of data needed

    years ≈ (t / IR)²
    • t — the threshold you require (e.g. 2)
    • IR — annualized information ratio

    At IR 0.5 that is 16 years; at IR 0.3, 44 years.

  • Fee drag and coin-flip streaks

    drag = 1 − ((1 + g − f) / (1 + g))^T; streak holders from year 1 = N × p^k
    • g — gross annual return
    • f — annual cost
    • T — years
    • N — number of managers
    • p — chance of beating the market in one year
    • k — streak length in years

    Streaks that may start in any year of a career are far more common than N × p^k; the pack simulates them.

Applied Pack · free members

Malkiel Alpha Validator Pack

Your returns and a benchmark in, the skill-or-luck test out: information ratio, t-statistic of alpha, the years your record still needs, fee drag over decades and a coin-flip manager simulation. Software for your own research, never signals.

  • Malkiel_Alpha_Validator.xlsx — READ ME, a Returns sheet for up to 240 months, an Alpha Test sheet (excess return, tracking error, IR, t-statistic, p-value, beta, regression alpha, years needed), a Years Needed grid, a 40-year Fee Drag table and a Coin Flip sheet, all live formulas with amber input cells
  • malkiel_alpha_test.py — stdlib-only CLI: runs the alpha test on a CSV of monthly returns, prints the fee-drag table for two cost levels and a Monte-Carlo of zero-skill managers with winning streaks from year 1 and anywhere in a career
  • returns_sample.csv — 60 EXAMPLE_ months of made-up portfolio and benchmark returns in the expected format
  • README.txt — inputs, the math, how to run, limits of the test and the educational-use disclaimer

Pack security

Macro-free Excel · plain-text Python you can read before you run it · no installers, no network access — the code works only on files you provide. Served only from closelook.net; we never distribute through download portals or email attachments. How to verify in 30 seconds →

SHA-256 d84e121341a32fddb881bf351fbd841b1fa91aa4a7dd7f1e498638f0ae5fc406

Independent scan report — VirusTotal, 70+ engines ↗

The pack is a free-membership download — no card, free forever. Membership also delivers the Daily Pulse and Weekly Signal to your inbox.

Join the Look — free →

Closelooknet publishes a market diary, not investment advice. This condensed read restates the book's ideas in our own words for education — for the author's full argument, go to the source.