Three Viral Trading Strategies Re-Tested on 16+ Years of Data

Introduction

Short trading videos promise a lot: an "86% win rate", "+982% in backtest", "the most famous buy signal in the world". I took three of these popular rules, wrote each one down as exact code before running anything, and tested it once on long history with realistic costs. Every rule was tested exactly as stated; where a video left a detail open, the choice was written down before the run and is reported below.

The short version: one claim failed badly, one only worked in the years the video chose to show, and one was confirmed.

All three are historical results for these exact rules. They are not predictions and not trading advice.

1. The "86% win rate" MACD rule on gold

Rule as tested (30-minute XAUUSD bars, bid/ask tick data, 2010-2025):

• Long: the MACD(12,26,9) line crosses above its signal line while both are below zero, and the close is above EMA 200. Short: the mirror image.
• Entry at the next bar's open (ask for longs, bid for shorts).
• Stop-loss at the EMA 200 value of the signal bar ("stop behind the 200 EMA"); signals with a stop closer than $0.50 or wider than 3 × ATR(14) are skipped.
• Take-profit at 1.5 × the stop distance. Exit on the bid for longs, ask for shorts. A bar touching both levels counts as a stop. Time exit after 48 bars.
• Commission $0.07 per ounce round trip, plus the real spread from the data.

The rule loses before costs too, so this is not a spread problem: on gold the signal simply does not have the edge the video shows. A win rate also says little on its own. With a target of 1.5 times the risk, a strategy needs 40% winners just to break even; with a target ten times smaller than the stop, almost any entry "wins" 85-90% of the time and still loses money. Before trusting a win rate, ask what the target was.

2. The first 5-minute candle of New York and the EMA 12

Claim: on Nasdaq futures, M5, 2019-2026: 1,448 trades, 57% winners, profit factor 1.29, drawdown under 20%, +982% total.

Rule as stated: at the New York open, look at the first 5-minute candle. If it closes above the EMA 12, buy; below, sell. Then the stop becomes a trailing stop that follows price, and the trade stays open while the move continues.

The video does not say where the first stop goes or how the trailing stop moves, so these choices were fixed before the run:

• Data: NQ continuous futures, 1-minute bars resampled to 5 minutes, New York time. EMA 12 on 5-minute closes.
• Signal: the 09:30-09:35 bar. Entry at the open of the 09:35 bar.
• Initial stop: the other end of the signal candle. Trailing: after each closed bar the stop moves to that bar's low (long) or high (short), only in the trade's favour.
• Exit at the 15:55 bar close if the stop was not hit. One trade per day.
• Costs: one tick of slippage on each side plus $4.50 commission per contract = 0.725 NQ points per trade.

The key step: the video only showed 2019-2026. The same rule was also run on the nine years before it, which the video did not show.

Three things stand out.

1. The window matters. 2019-2026 was one of the strongest runs in Nasdaq history. Over the nine earlier years, the same rule lost money after costs in eight of them.
2. Costs are not a detail. Before costs the rule is mildly positive in both windows. A fixed cost of 0.725 points is a much larger share of a move when the index trades at 2,000-7,000 than at 20,000.
3. The published numbers could not be reproduced from the published rules. This version produced 36% winners, not 57%, because the stop and trailing rules are unspecified and every reasonable choice changes the result. A rule that cannot be written down exactly cannot be tested, and cannot be traded the same way twice.

3. The Golden Cross on the S&P 500

Claim: over 33 years, being in the index only while the 50-day average is above the 200-day average does not earn more than buy and hold, but it loses less on the way.

Rule as tested: S&P 500 index daily closes, January 1993 to October 2026. In the market while SMA 50 > SMA 200 at the previous close, otherwise in cash earning nothing. No costs and no dividends for either side, so the comparison is like for like.

The claim holds: about half a percentage point less return per year, and a maximum drawdown cut from −57% to −34%. The rule switched only 30 times in 33 years. With interest on the cash periods the return gap would be smaller still. On a CFD the opposite applies: a long position held for months pays a daily financing charge that this index test does not include.

4. What the three tests have in common

• Write the rule down before testing it. Every detail a video leaves open is a place where the result can be tuned after the fact.
• Test the years that were not shown. A strategy shown only on a strong trend has not been tested yet.
• Include real costs, and check the result without them. If a rule is flat before costs, the spread decides the outcome.
• A win rate means nothing without the target and stop.
• Report the confirmations too. Not every popular claim is wrong; the Golden Cross did what was claimed.

Files

The attached GoldenCross_Index.mq5 is the Golden Cross rule as an Expert Advisor for MetaTrader 5. It is meant for an index CFD on an H1 chart and makes one decision per day, a few minutes before the New York cash close:

• the daily close is taken as the close of the last H1 bar at or before the cash close; short US-holiday sessions are skipped;
• if SMA 50 > SMA 200 and there is no position, it buys a position worth a set percentage of the balance (input NotionalPct );
• if SMA 50 falls to or below SMA 200, it closes the position;
• no stop-loss and no take-profit, as in the tested rule; every decision is logged to a CSV file in MQL5\Files .

The server-time inputs ( DecisionHHMM , CloseHour ) are set for a broker whose server runs on New York time plus 7 hours; adjust them to your broker. Test it on a demo account first, and remember the financing charge on long CFD positions.

Historical results for these exact rules. They do not guarantee future results and are not financial advice.