Trading Strategies: 41 Rule Sets Tested on Real Prices, Costs Included

Michael Whitman · Trading since 2017, full-time since 2020 · YouTube creator, 137K subscribers · Published 2026-11-07

This page is a scorecard rather than a catalog. Instead of describing ten styles in the abstract, it reports what happened when the rules behind them were written down, run on real prices and charged real costs. Most of them failed, and that is the useful part.

The scorecard: every rule set we tested

Each test below had its rules and its pass bar written down before a single result was computed, and none was adjusted after the run. “After costs” means 0.02% of the trade value per side on the daily tests and 1 cent a share each way on the intraday ones, with the ADX rows judged at 0.02% instead.

The short answer from 41 tests: 4 passed, and every one of those came from 56 trading sessions.

How it works

Every strategy answers three questions in advance. What event opens a trade, what event closes it, and how much money goes in. A trader who cannot write all three down does not have a strategy yet; they have a habit, and a habit cannot be tested.

Families differ by what they bet on. Trend-following rules, such as the golden cross and the Donchian breakout, bet that a move already under way keeps going. Mean reversion rules, such as the Keltner Channels rule we tested on the 15-minute chart, bet that a stretched price snaps back toward its average. Holding period then splits them again into scalping, day trading and swing trading.

A test is only honest if the rules come first. The channel’s tests followed one routine: write the rules and the pass bar, fetch the data, check the bars against a second source, run once, then have a separately written program reproduce the trades. Changing a setting after seeing the result is how a backtest turns into a story that fits the past and nothing else.

A worked example: one stock, two ways

Take Apple in the daily golden-cross test. Prices here are adjusted for splits and dividends, which is why they look low for the earlier years. The first signal bought $1,000 at the 31 Aug 2016 open of $24.16, which is 41.39 shares. The death cross sold them at the 24 Dec 2018 open of $35.12.

The arithmetic: 41.39 shares x ($35.12 - $24.16) = a gain of $453.50 before costs. The cost is 0.02% of $1,000 going in ($0.20) plus 0.02% of $1,453.50 coming out ($0.29), so the trade netted $453.01 over 582 sessions.

Apple produced six golden-cross trades in the window. They netted +$453.01, +$1,937.30, -$81.04, +$83.03, -$124.40 and +$431.21, the last one still open and valued at the 2 Nov 2026 close. That adds up to $2,699.10.

Now hold instead. The same $1,000 bought at the 2 Jan 2014 open of $17.19 and kept to the 2 Nov 2026 close of $338.32 made $18,678.19 after costs. Part of the gap is design, since every signal stakes a flat $1,000 and gains are not rolled into the next trade. The rest is time out of the market: the rule sat in cash through stretches when the stock kept climbing.

The original data

The long-run test covered 3,228 sessions, from 2 Jan 2014 to 2 Nov 2026. The universe was fixed in advance as the 20 largest S&P 500 companies at the end of 2013, ranked from SEC share counts and Yahoo closing prices, and every one stayed in whatever happened to it later. That choice avoids survivorship bias, the error of testing only on stocks that are still winners today.

All three daily rules made money, and all three still failed. The golden cross was in the market 64.1% of the time with a median hold of 156 sessions, and all 20 stocks ended up. The 20/55 breakout was in 78.3% of the time and the 20/10 version 46.1%, with a median hold of 21.5 sessions.

The failure shows up against random entry. For each rule, 2,000 simulated runs placed the same number of trades with the same holding lengths on random dates in the same stocks. The golden cross beat 62.6% of them, short of the 90% bar. The 20/55 breakout beat only 0.8%, against a random median of $31,847. The 20/10 version beat 3.6%, against a random median of $17,986.

The intraday tests ran on SPY, QQQ, NVDA and TSLA over 56 sessions, from 14 Aug to 2 Nov 2026, at $5,000 a trade. Each indicator had two rule sets, the plain default and the version taught in the channel’s own videos, on the 5-minute, 15-minute and 1-hour charts. A version passed only if it made money after costs, made money on at least 3 of the 4 symbols, and made money in both halves of the window.

Of those 24 versions, 11 finished positive at 1 cent a share and 7 at 0.02%. Together they lost $11,409.31 on 5,553 trades at the lighter cost and $21,146.56 at the heavier one. Supertrend lost on four of its six versions, and the channel’s own 1-hour Supertrend settings lost $3,196.16 on 49 trades. Summed over the three timeframes at 1 cent a share, Supertrend ended at -$4,161 on its default and -$3,882 on the channel’s settings, Hull Suite -$2,186 and -$2,249, and Keltner -$2,230 on its default against +$384 on the channel’s rules. ADX was the one indicator ahead on both rule sets, at +$1,438 and +$1,477.

The four passes: the channel’s Keltner Channels mean reversion on the 15-minute chart (+$390.55 on 230 trades) and its 1-hour pullback rule (+$383.70 on 20 trades), and the ADX crossover on the 15-minute chart, both the plain default (+$286.54 on 102 trades) and the channel’s version, which adds an exit when ADX drops below 20 (+$261.38 on 102 trades), those two at the 0.02% cost. The full rows are in the strategy test results file.

Why costs decide short-term strategies

The cost per trade is small; the number of trades is not. On a $5,000 position in a $50 stock, 1 cent a share each way is $2.00 a round trip, about 0.04% of the position. A rule that trades hundreds of times in 56 sessions pays that toll every time: the default Hull Suite on the 5-minute chart traded 832 times and lost $1,882.50 at 1 cent a share, or $3,337.45 at 0.02%.

The Parabolic SAR test shows it plainly. On the 1-minute chart, its every-flip version at the video’s setting made $130.45 before costs on 1,510 trades and lost $240.43 after them. The stricter 0.02% cost is closer to what a market order pays on wider-spread names, and it pushed the default Keltner breakout on the 5-minute chart from -$966.29 to -$1,711.51. Slippage adds to both.

When it fails

A pass on 56 sessions is a small sample, not a finding. The Keltner 1-hour pass rests on 20 trades, which a few lucky closes can swing. Testing 24 intraday versions and keeping the best few also invites luck: if enough rules are tried, some will clear any bar by chance. The four passes are leads for a longer test, not rules that are known to hold.

Pass bars can disagree with each other. The Keltner 15-minute rule passed at 1 cent a share but lost $16.64 at the 0.02% cost. A strategy that flips from profit to loss on a small change in costs has an edge too thin to rely on. Held to one cost for all 24 versions, the count moves: 6 pass at 1 cent a share, but only 3 at 0.02% (the Keltner 1-hour rule and both 15-minute ADX versions).

Beating zero is not the same as beating the alternative. Every daily rule made money from 2014 to 2026 because large US stocks rose. The honest comparison is holding the same stocks, or entering at random with the same exposure, and none of the three cleared either.

The regime can change under the rule. A trend rule that worked across a rising decade can whipsaw for years in a sideways market, and an intraday rule fitted to one autumn says nothing about a different volatility level. None of these results predicts how any rule will do from here.

Adjusted prices are a simplification. The daily test used prices adjusted for dividends, so it assumes dividends were reinvested at no cost and ignores the tax a taxable account would owe on each sale.

How to judge a strategy you find online

Ask for the rule in writing first. If a seller cannot state the entry, exit and size precisely enough to code, there is nothing to test. The VWAP strategies, RSI-2 strategy and 9 EMA strategy pages each pin a common rule down to that level.

Then ask four questions about any result shown. Were costs included, and at what rate? Was the test window chosen before or after the result? How many trades does it rest on? What did holding the same market do over the same dates? A result that cannot answer all four is an advertisement.

Questions people ask

What is the most profitable trading strategy? No test here supports naming one. In our daily test, holding the 20 stocks beat every rule, and the best intraday passes rest on 20 to 230 trades.

Do trading strategies work? Some make money in some periods. Across our 41 tests, 4 cleared a bar set in advance, and none cleared it over the nearly 13-year daily window.

What are the main types of trading strategies? Trend-following, mean reversion, breakout and momentum trading, each run at a holding period from minutes to months.

How do I test a trading strategy? Write the rule and the pass bar first, include costs, compare against holding and random entry, and run it once.

Related

The full testing routine, step by step, is in how to backtest a strategy, which shows how to set the pass bar before the run. For the long-run rule tested here, the golden cross page explains why the signal is rare, and turtle trading covers the original Donchian 20-day system. The intraday indicators each have their own page: Supertrend, Keltner Channels and the Hull moving average. If you trade crypto, how to trade crypto runs the same kind of test on bitcoin and ether.

This page is educational, not financial advice. Test every idea on your own charts before risking money.