5.8 Intraday strategies: what the evidence says

Motivation

Most published intraday "edges" are small, and many are artefacts of costs that the authors' market did not charge. This section states the best evidence precisely and gives you a tool to check whether an edge survives Indian costs.

5.8.1Intraday momentum and breakouts#

Market intraday momentum Gao, Han, Li and Zhou ("Market intraday momentum", Journal of Financial Economics 129(2), 2018, 394–414) use SPY from 1993 to 2013. The return over the first half hour (measured from the previous close, so it includes the overnight gap) predicts the return over the last half hour. The predictive $R^2$ is about 1.6%; the effect is stronger on high-volatility days, high-volume days and macro-news days, and appears in 10 other ETFs. Li, Sakkas and Urquhart (Journal of Financial Markets, 2022) find it in 12 of 16 developed markets. Later out-of-sample checks report a weaker effect (secondary sources).

Opening range breakout (ORB) Define the opening range as the high and low of the first $m$ minutes after the open. Go long if price breaks above the high, short if it breaks below the low; place the stop at the other side of the range; exit at the close. Zarattini, Barbon and Aziz, "A Profitable Day Trading Strategy for the U.S. Equity Market" (Swiss Finance Institute Research Paper 24-98, 2024), apply a 5-minute ORB to "Stocks in Play" (stocks with unusually high relative volume, usually because of news) across more than 7,000 US stocks from 2016 to 2023 and report, for a top-20 portfolio, a net return above 1,600%, Sharpe 2.81 and annual alpha of 36%. Zarattini, Barbon and Aziz, "Beat the Market: An Effective Intraday Momentum Strategy for S&P500 ETF (SPY)" (SSRN 4824172, 2024; SFI RP 24-97) report 1,985% net total return, 19.6% a year and Sharpe 1.33 from 2007 to early 2024 for a "noise area" breakout with trailing stops. A VWAP-based paper by Zarattini and Aziz (SSRN 4631351) is summarised by secondary sources with large returns; those figures were not verified against the paper.

Caveats you must apply

These are authors' backtests, in working papers, not peer-reviewed at the time of reading; one author sells day-trading education.
US costs differ from Indian costs: no STT, cheaper stamp duty, and tighter spreads on SPY and US large caps. An independent replication on index CFDs reported that the gross edge reproduced but the net edge was about zero (a blog replication; treat as anecdotal).
The parameter space (range length, stop rule, filters, universe) is large, so the multiple-testing corrections of M04 apply. A reported Sharpe of 2.8 after searching many variants is evidence, not proof.
The Indian microstructure changed in 2026: the pre-open's random close and market-order window and the closing auction change the first and last minutes of the day, which is exactly where these strategies trade.

5.8.2From $R^2$ to basis points#

Proposition 5.6from $R^2$ to basis points

Let the predictor $x$ and the target return $y$ be jointly Gaussian with mean 0, correlation $\rho > 0$ and $\operatorname{sd}(y) = \sigma_y$ . A strategy that takes a unit position $\operatorname{sgn}(x)$ earns

E[\operatorname{sgn}(x)\,y] = \rho\,\sigma_y\sqrt{2/\pi}. \tag{5.24}

Proof

$E[y \mid x] = \rho\sigma_y x/\sigma_x$ . By the tower property, $E[\operatorname{sgn}(x)y] = E[\operatorname{sgn}(x)\rho\sigma_y x/\sigma_x] = \rho\sigma_y E|x|/\sigma_x$ , and for a centred Gaussian $E|x| = \sigma_x\sqrt{2/\pi}$ .∎

Worked example

$R^2 = 1.6\%$ means $\rho = \sqrt{0.016} = 0.126$ . If the last-half-hour return of an index future has standard deviation 0.25% (an illustrative value; measure it), (5.24) gives $0.126 \times 0.25\% \times 0.798 = 2.5$ bp per day before costs. A Nifty futures round trip costs 5.95 bp in statutory charges alone after April 2026 (M01 · 1.7), plus the spread. The trade loses money in expectation in India even if the US effect holds here with the same strength. On days with twice the volatility the gross edge doubles to about 5 bp, still below costs. This is a five-line calculation that should precede any backtest.

5.8.1Intraday momentum and breakouts#

5.8.2From R2R^2R2 to basis points#

5.8.2From $R^2$ to basis points#