Source: https://forexiapro.com/signature-trade/why-trading-setups-fail/ # Why Do Trading Setups Fail? Our Own Backtest Says, and Two Real Days Show It Dylan Forexia14 November 20266 min read Why do trading setups fail? We ran our own and watched it lose. The setup we teach, run as a fixed rule with real costs, lost money in every period from 2015 to 2026: 0.104R a trade in 2015 to 2020, 0.122R in 2021 to 2024 and 0.128R in 2025 to 2026. R is the amount risked on one trade. This page is the short answer to why, with two real days that show it. [![Why do trading setups fail: every order past the level is a buy, and once they are filled, who is left to buy](/seo/why-do-breakouts-fail/why-do-breakouts-fail-every-order-past-the-level-is-a-buy-diagram.webp)](/seo/why-do-breakouts-fail/why-do-breakouts-fail-every-order-past-the-level-is-a-buy-diagram.webp) Diagram: a break fills every order past the level, and then nobody is left to carry it. The real days are below. The full study, with its tables and its limits, is [the backtest](/research/mechanical-setup-backtest-2015-2026/). This page keeps to the question. It is also the page that sits behind everything else we say about [the #SignatureTrade](/signature-trade/): the setup is a description of where to look, and it did not make money as a rule. ## Why do trading setups fail in our test? Because the moment has no lean, and a trade with no lean pays its costs. After 18,884 triggers in the fitting years, with no stop, no target and no costs, price had moved -0.2% of a daily range the trade's way after an hour and -0.5% by 17:00. That is zero. [![Why do trading setups fail: bar chart of the result per trade after costs, in R, for four periods, all below zero](/seo/mechanical-setup-backtest-2015-2026/forex-strategy-backtest-2015-2026-loss-per-trade.webp)](/seo/mechanical-setup-backtest-2015-2026/forex-strategy-backtest-2015-2026-loss-per-trade.webp) One fixed version of the setup as a rule, seven dollar majors and gold, 2012 to October 2026. Every period lost after costs. The first bar, 2012-14, comes from data added later for another test. - **No lean.** With no costs the setup made -0.03R, with the spread -0.10R, and with commission as well -0.16R. - **Costs do the rest.** In the fixed version, 2021 to 2024, spread and commission were about 821R of a 1,093R loss. Take them away and about -0.03R a trade is left, which is break-even and not a profit waiting to be freed. - **Real breaks.** The biggest group of stop-outs was a day that kept going. On 8,931 trades, 23% were stopped on a real breakout, for -2,012R. The test is a program with no opinion about the day. It does not describe a person reading the chart, and we do not claim it does. ## How many levels were real breaks? Not a lot, and enough. We counted every time a level, an edge of the Asian Box or the previous day's high or low, was traded through by at least 3% of the average daily range after the London open, on seven dollar majors and gold, from 6 April to 1 October 2026. That happened 1,601 times. On 1,310 a candle closed back inside the level that day. On 291, 18%, nothing did by 17:00, and those ran a median of 65% of an average day past it. The 291 look just like the rest until the day ends. A setup that is right on the 1,310 and wrong on the 291 has to survive its own wrong days, and the wrong days are the ones that go furthest. That is a count of candles, with no entry, stop or cost. ## What does a real break look like? AUD/USD on 10 September 2026 On Thursday 10 September, Wednesday's low was 0.72095. The 05:30 candle closed under it, the 05:45 candle closed back at 0.72110, and the 06:00 candle closed under at 0.72064. No candle closed back after that. By 08:30 on Friday price was at 0.71496, 60 pips under the line. There was no reversal to mark up. [![Why do trading setups fail: AUD/USD, 15-minute chart, 10 September 2026, one candle closes back at 0.72110, the next closes under Wednesday's low at 0.72064, and price keeps falling](/seo/previous-days-high-and-low/audusd-previous-day-low-real-break-10-september-2026.webp)](/seo/previous-days-high-and-low/audusd-previous-day-low-real-break-10-september-2026.webp) AUD/USD, Thursday 10 September 2026. One candle closed back above Wednesday's low at 05:45 New York, and the next closed under it for good. There was no reversal to mark up. Real 15-minute candles. A trader who read the 05:45 close back as the hunt was wrong, and the setup was not at fault in any way the chart showed at the time. Some breaks are real, and a close back does not rule that out. ## What does a close back that does not hold look like? USD/JPY on 28 September 2026 On Monday 28 September, Friday's low was 156.938. The 04:00 candle closed at 156.697, 24 pips under it, and the 04:15 candle reached 156.506, 43 pips under. That candle and the next two also closed under the line. The 05:00 candle closed at 156.991, back above, and by 11:00 price was at 157.583, 108 pips off the low. This one turned, and it is the other side of the coin: for an hour it looked like a real break, with four closes under the line. A trader who judged it on those first closes judged it wrongly, as one who judged AUD/USD on its close back did. The day is on [the previous day's high and low page](/glossary/previous-days-high-and-low/), and [the previous day high breakout](/signature-trade/previous-day-high-breakout-london-open/) has the counts. ## Who is on each side when a setup fails? The retail seat reads a setup as a rule that should work and treats a loss as the rule breaking. The market maker's seat sees the same level as a pool of orders and expects it to be used, whichever way the next candle goes. The observer's seat, ours, reads a day backwards, from where the stops were. That read is useful for explaining a day. As the test shows, it is not by itself a way to trade one. That is our reading of the seats and cannot be proved from candles. ## Why do the chart markups look so much better than the test? Because a markup is chosen after the fact. We mark up a day only once its extreme has held and price has come back at least 0.4 of the daily range, which is close to the definition of a good trade. The program does not know that when it fires. In the six marked-up months it fired 2,205 triggers, and only 347 of those were on a marked extreme. Trades on the days we never marked up lost 0.56R each, and there were far more of them. Across the fitting years, the extreme that set off the trigger was still the day's extreme at 17:00 New York only about two times in five. Every day shown on our site was picked afterwards. They show what happened. They do not show what the setup earns. [Textbook or Partial](/glossary/textbook-or-partial/) says how days are filed, and what that filing leaves out. ## Did we find a fix? Nothing that held. Each attempt moved along a trade-off and none lifted the result above zero. - **More confirmations.** Setups with six of the method's confirmations lost 0.04R a trade against 0.18R for those with two. The extra confirmations come later in the move. [How many confirmations a setup needs](/signature-trade/how-many-confirmations-forex-setup/) has the counts. - **Risking more on the best setups.** It made things worse: over 2015 to 2020 the loss went from 1,069R at level risk to 1,331R. - **Wider stops, other exits, the best pairs and hours.** Wider stops lost less only because costs shrink against a larger risk, and a slice that held in one period lost in the next. [Trade management in forex](/signature-trade/trade-management-forex/) goes through them. ## What does this not prove? It does not prove that a person reading the chart loses, or wins. In the six test months about one trigger in six was on a marked extreme, and a trader's picks would have to find nearly twice that share just to break even. Nothing in the candles, the calendar or the confirmations closed that gap. That is a reason to learn the read. It is not evidence that it pays, and we would rather you heard that from us. The way to find out is a log of every setup you see, taken or skipped, and why, then a test of whether the taken ones beat the skipped ones. Course 5 of ForexiaPRO, The #SignatureTrade, teaches the days it failed, in lesson 5.3.3. It is on [ForexiaPRO](/pro). ## Frequently asked questions ### Why do forex setups fail? In our test the moment the setup fires has no lean: after 18,884 triggers in the fitting years, price had moved -0.2% of a daily range the trade's way after an hour and -0.5% by 17:00, which is zero. A trade with no lean pays costs, and in the fixed version, 2021 to 2024, spread and commission were about 821R of a 1,093R loss. ### Did more confirmations stop a setup failing? No. Setups with six of the method's confirmations lost 0.04R a trade against 0.18R for those with two, and both lost. The extra confirmations arrive later in the move, so a setup that is right has less left and one that is wrong costs more. ### Does this mean the setup does not work for a person? No, and we do not claim it. The test is a program with no judgement. In the six test months about one trigger in six was on a marked extreme, and a trader's picks would have to find nearly twice that share to break even. Nothing we measured closed that gap, and it is not shown that a person cannot. ## More from the blog ### [Confluence Trading Forex: Why One Reason Is a Guess, and What a Stack Is](/signature-trade/confluence-trading-forex/) Confluence trading forex means more than one reason on a chart. What counts as one reason, what counts twice, and one day that stacked six of eight. ### [Best Forex Trading Books: Seven Worth Reading, and What Each Gets Wrong](/learn-forex/best-forex-trading-books/) Best forex trading books: seven worth reading, from psychology to candles, and what each gets wrong. A reading order for beginners. No returns promised. ### [Average Daily Range on Tuesday: What ADR Met Tuesday Means](/weekly-cycle/average-daily-range-tuesday/) What it means when the average daily range is spent by Tuesday, how often it happens (28% of Tuesdays), and why it is one reading and not a signal.