Day Trading Strategies Backtested: 26 Strategies, 34 Tests, One Survivor

00The short answer

• Most day trading strategies did not survive. Of 139 versions, 99 made money before costs, 44 after costs, and 2 passed all six of our gates: both are the noise area breakout on NQ futures in the New York session, now being tested on new data.
• Costs were the biggest killer. 55 versions made money before costs and lost it to them; 40 lost even before costs.
• Beating random entries was not enough. 38 versions timed their entries better than chance, and 14 of those still lost money after costs.

01What counts as day trading here

A test is in this guide when every trade is closed within the same session or day: at the session close, at a fixed time, at the end of the day, or with no position held overnight, as written in its frozen specification. That gives 34 tests of 26 strategies: some rules were tested separately on several markets, the noise area breakout on NQ, ES, gold, Bitcoin and YM, and the initial balance breakout, intraday momentum, previous day high/low reversal and TheStrat on two markets each. Each market counts as its own test, with its own result. The rules are the popular, published or widely taught versions, written down before any data was run; each test page has its full rules, code and every trade.

How each was judged: six gates, the same for every test: a profit factor of at least 1.2 in the validation years, at least 50 trades in each period, a clear result in the first period, better than the same number of trades at random times, still positive without the 10 best trades, and positive in the final period. Results are in each test's own unit (dollars per contract, R, pips or percent), so this guide compares them by what is unit-free: gates passed, and whether a version made money before and after costs. Read the protocol.

02The ranking: 34 day-trading tests, from closest to furthest

Each test, one strategy on one market, is ranked by its best version's gates, then by the share of its versions that made money after costs. A higher rank means it got closer, not that it works.

03What came closest

The noise area breakout on NQ, New York session, is the only day trading strategy that passed every gate, long and short and long only. It comes from a published paper (Zarattini, Aziz and Barbon, 2024): trade NQ only when it leaves the range of its usual intraday moves, with a trailing stop. The same rule did not pass on ES, gold, Bitcoin or, retested on 7 November, YM, the Dow. Our holdout had been used by earlier tests on the same file, so it is now tested forward on data no one has seen.

Four more came close, with five or four gates: the previous day high and low reversal on YM (five gates), and on NQ the initial balance breakout, the EMA ribbon 5/8/13 and the MACD + RSI + 200 EMA strategy (four gates each). Three of them are also followed forward. Of the five strategies with four or more gates, four trade NQ.

04What killed the rest

Costs

55 of 139 versions made money before costs and lost after them. Day trading means many small trades, and a fixed cost per trade takes a large share of each. The clearest case is the VWAP bands on gold: positive before costs in both versions, negative after, because the stops were often under a dollar. The opening range breakout on gold beat random entries in all 12 of its versions and still lost after costs in most of them. Our study of costs across all our tests has the details.

Losing even before costs

40 versions lost money before any cost. Among them, all the versions of the first 5-minute candle and fair value gap strategy from a popular YouTube video, and the intraday momentum rules of a published paper on ES and NQ.

No better than random timing

The 5-minute opening range breakout on NQ made money in all 4 versions, but the same trades placed with the direction of other days did as well: the profit came from NQ's big trend days, not from the first candle. The ICT Silver Bullet did not beat random entries in any version.

05By market

NQ, the Nasdaq-100 future, did best: 61% of its versions made money after costs, against 24% on ES and 14% on gold. Gold is the opposite of NQ: 78% of its versions made money before costs, because its moves are small next to a $0.30 round trip on short-term stops.

06By style and by chart

Breakouts did better than mean reversion in day trading: 35% of breakout versions were positive after costs, against 21% for mean reversion, and they beat random entries more often (33% against 26%). Mean reversion also lost the largest share of its profits before costs to costs: 8 of its 12 versions that made money before costs lost after them. Our comparison of trend following and mean reversion across all our tests, not only day trading, found the same lean.

07Before you trade any day trading strategy: a checklist

• Count the cost per trade in R. Divide the round-trip cost by the distance to your stop. On the gold VWAP bands it came to about one R a trade, which no signal could pay for.
• Compare with random entries. On an index that trends, almost any long entry makes money in a backtest; ask whether the signal does better than the same trades at other times.
• Look at the worst years and the period you did not optimise on. Many strategies here looked good in one period and gave it back in another.
• Be suspicious of one beautiful day. The fair value gap strategy matched its video's example day almost to the dollar, and lost money over its 1,527 trades.
• Prefer rules you can see. Every strategy here is published with its rules and code; a result you cannot reproduce is a claim, not evidence.

08Intraday charts, held overnight

These tests use intraday charts, from 5-minute to 4-hour, but hold positions overnight, so they are not day trading and are not ranked above: Golden Cross (1H), Golden Cross (EUR/USD 1H), EMA 9/21 (1H), EMA 9/20 gold (M15, H1), Coppock Curve (1H), TTM Squeeze (gold, 1-hour), RSI 30/70 (gold, 5-min and 1-hour), WaveTrend (gold, 1-hour), Head and shoulders (gold, 15-minute), Engulfing candle (gold, 1-hour), Supertrend (gold, 1-hour), Supertrend (ES), The RSI(2) Dip Buyer (ES, 1-hour), Bollinger Band Reversal (ES, 1-hour), Donchian Channel Breakout (ES, 1-hour), RSI Divergence (ES, 1-hour), MACD Crossover (ES, 1-hour), Six Strategies on Four Currency Pairs, Parabolic SAR (EUR/USD, 1-hour), ADX + DI (4 forex majors, 1-hour), Fibonacci 61.8% (EUR/USD 1H), Stochastic + RSI + MACD (EUR/USD), Order block + fair value gap, SMC (EURUSD, 5-min), Anchored VWAP pullback (EURUSD, 30-min), Trendline breakout and bounce (YM), TDI, Traders Dynamic Index (EUR/USD, 4-hour), Trading Made Simple, TDI + Heikin Ashi (EUR/USD).

09What this guide cannot tell you

• 26 strategies, not all of them. Others may do better; we add them as we test them.
• Our markets and our costs. A trader with lower costs, or on another market, could see different results.
• The future. The one strategy that passed everything is being tested on new data; a backtest is not a promise.

FAQCommon questions

Do day trading strategies work?

Rarely, in our tests: 32% of 139 versions of 26 popular day trading strategies, in 34 tests, made money after costs, and one strategy passed all six of our gates.

What is the best day trading strategy?

In our tests, the noise area breakout on NQ futures in the New York session, the only day trading strategy that passed every gate. It is now being tested forward on new data, so it is not proven yet.

Why do most day trading strategies fail?

In our tests mostly because of costs: 55 versions made money before costs and lost it to them, and 40 lost even before costs. Many small trades make a fixed cost per trade a large share of each.

Which market is best for day trading strategies?

In our tests, NQ (Nasdaq-100 futures): most of the versions that made money after costs, and the one survivor, traded NQ. Gold was the opposite: most versions made money before costs and lost it to them.

How were these day trading strategies tested?

Each on 1-minute to 1-hour data, every trade closed the same session or day, with rules written and dated before the data was run, costs included on every trade, results split into development, validation and holdout periods, a random-entry control as the benchmark, and every variant reported. Read the protocol.

10Check our arithmetic

11Corrections

11 Nov 2026, data correction: the London breakout test on GBP/USD was rerun on the same data (to 28 September 2026) with its clock corrected (the time stamps were one hour early in most weeks of 2004 to 2014 and in a few weeks a year since). Two of its versions no longer beat the random-day control and one no longer made money after costs, so the counts changed: positive after costs 45 to 44 (33% to 32% of versions; breakouts 38% to 36%), lost to costs 53 to 54, timed better than chance 40 to 38, of which still lost after costs 15 to 14 (breakouts beat random entries 38% to 35%); the London breakout moved from 7th to 13th in the ranking (best 3 of 6 gates to 2 of 6, positive after costs 2 of 3 versions to 1 of 3), and the GBP/USD, breakout and 1-minute rows changed with it. 7 Nov 2026, wording: the guide now counts 29 tests of 22 strategies, because some rules were tested separately on several markets (the noise area breakout on four); it first called them 29 strategies. No result changed. 7 Nov 2026, correction: the chart column now gives 15-minute bars for the previous day high/low reversal on YM and 5-minute bars for the ICT Silver Bullet on ES, as in their specifications; it had said 5-minute and 1-minute. Only the table by chart changed with them. 7 Nov 2026, addition: the noise area breakout on YM futures (6 versions, none passed) was added; the guide now counts 30 tests and 127 versions. 8 Nov 2026, addition: the DAX 15-minute opening range breakout (4 versions, none passed) was added; the guide now counts 31 tests of 23 strategies and 131 versions. 9 Nov 2026, addition: the DAX trend pullback to the Fibonacci 61.8% level (2 versions, none passed) was added; the guide now counts 32 tests of 24 strategies and 133 versions. 9 Nov 2026, addition: the Standard Error Bands reversal on ES 2-minute bars (3 versions, none passed) was added; the guide now counts 33 tests of 25 strategies and 136 versions. 11 Nov 2026, addition: classic pivot points on GBP/USD 15-minute bars (3 versions, none passed; listed under breakout, since two of the three trade in the direction of the move) were added; the guide now counts 34 tests of 26 strategies and 139 versions, and the breakout row 35% positive after costs and 33% beating random entries (was 36% and 35%).

+More tests and guides

• Noise area breakout nq
• Trading costs
• Trend following vs mean reversion
• Opening range breakout strategy
• Free prop firm challenge calculator
• What all the backtests found so far
• The backtesting protocol
• How to backtest a trading strategy without fooling yourself
• Backtesting glossary