Algorithmic Trading: Winning Strategies and Their Rationale (Chan, 2013)

Ernest P. Chan, John Wiley & Sons 2013, Wiley Trading series. Read here through a 15-page publisher’s excerpt — title pages, copyright, contents and the full preface. No chapter text. What that gives is unusually good for a book excerpt: the author describing, in his own words and at length, what the book argues and which techniques it uses.

The first source in this spoke that is a practitioner’s book rather than a repo, a catalog or a blog post, and the oldest thing here by a decade.

The structure

Eight chapters: backtesting and automated execution (1), then mean reversion across four chapters — basics, implementation, stocks and ETFs, currencies and futures (2–5) — then momentum across two, interday and intraday (6–7), then risk management (8). The shape is itself an argument: mean reversion gets four chapters and momentum gets two, and everything is downstream of chapter 1.

The thesis, in one sentence

The emphasis throughout is on simple and linear strategies, as an antidote to the overfitting and data-snooping biases that often plague complex strategies.

He restates it as a preference for “using simple mathematical models for trading” over “the usual ‘data-mining’ approach” of throwing indicators at a price series to see which is profitable. This is a methodological position, not a technical one, and it puts the book squarely against the direction the rest of this corpus runs in — see synthesis.

What the preface enumerates

Detecting mean reversion: augmented Dickey-Fuller (ADF), Hurst exponent, variance ratio test, half-life. Detecting cointegration: cointegrated ADF, Johansen test. He promises intuition rather than recipe — “an intuitive understanding of what they are really testing and the simple mathematical equations behind them.”

Trading it: linear, Bollinger band, and Kalman filter approaches, plus the question of whether raw prices, log prices or ratios belong in the tests. Instruments: interday and intraday stocks, ETF pairs and triplets, ETFs against their components, currency pairs, futures calendar and intermarket spreads, and the VX volatility future. Futures get the most theory — spot versus roll returns, backwardation and contango from a simple price model.

Why strategies stop working: he says some of these became hard “in recent years due to the rise of dark pools and high-frequency trading”, and that fundamental changes can unhinge a previously profitable ETF pair — and can also point at the repaired version.

Momentum is the thin half of the excerpt: statistical tests for time-series momentum are named, and the preface text cuts off mid-sentence on the main theme. The book’s own chapter count says the author had less to say here too.

The pitfall inventory

This is what carries over to backtesting:

These pitfalls can cause live trading results to diverge significantly from their backtests. As veterans of algorithmic trading will also agree, the same theoretical strategy can result in spectacular profits and abysmal losses, depending on the details of implementation.

Named: data-snooping bias, survivorship bias, primary versus consolidated quotes, the venue dependence of currency quotes, the nuances of short-sale constraints, the construction of futures continuous contracts, and the use of futures closing versus settlement prices in backtests. Plus “some instances of ‘regime shift’ historically when even the most correct backtest will fail to predict the future returns of a strategy” (market-regime-analysis).

Every item is a way a correctly implemented backtest still lies. That is a different failure from lookahead bias — the code is right and the data or the market has moved underneath it.

Four things he says he learned from running a fund

From the preface’s closing pages, presented as personal experience over four years:

Manually overriding a model is “seldom a good idea… no matter how treacherous the market is looking.”
“It is always better to be underleveraged than overleveraged, especially when managing other people’s money.”
“Strategy performance often mean-reverts.”
“Overconfidence in a strategy is the greatest danger to us all.”

If strategy performance itself mean-reverts, then a good backtest and a good recent live run are both weak evidence about next quarter — which is the same caution smolinski-h1-2026-aar earns from the other direction, a trader’s strong half-year reported without drawdown context.

Tooling, dated 2013

MATLAB is the author’s language and all code examples are MATLAB; he notes it “is no longer the only contender” and surveys integrated development environments for traders from Deltix down to open-source TradeLink, arguing the decisive feature is ease of switching from backtesting to live trading — the backtest/live parity point this spoke already holds from banbot, made thirteen years earlier. This section is a historical snapshot; the platform landscape it describes has turned over.

Tier

T2, with the scope stated. It is primary text — the author’s own preface, published by a university-grade trade publisher — not a summary of him. Against that: it is front matter only, so nothing here establishes that any strategy in the book works, and no result, table or backtest has been read. The book is also from 2013, which matters unevenly: the statistical apparatus (ADF, Johansen, Kalman) is durable, the market-structure observations are of their moment, and the software survey is obsolete.

ernest-chan · mean-reversion · backtesting · algorithmic-trading · strategy-optimization · market-regime-analysis · john-wiley-and-sons · synthesis