Price Hits the Old High: Ride It or Fade It? 5 Years of Gold Tests
Does price break or reverse at the previous swing's extreme? We measured gold M1 2022–2026, tested ride and fade systems, and show why fading lost in trend years.
- On gold M1, price that returns to the previous swing's extreme breaks through more often than it bounces back.
- Stop losses pile up just above old highs and below old lows. When they get swept, they push price further.
- A fade system can look great in some years, then lose so much in a trend year that the account is wiped out.
- The ride-side edge is tiny. On small swings it loses to the spread, and the system we built from it made little, mostly in 2026.
- If you ride, give your Stop loss room to breathe and don't set your profit target too close.
When price runs back to an old high or an old low, a trader has to choose: ride it (enter in the direction price is moving and hope it breaks through) or fade it (enter the opposite way and hope it bounces back). In our backtest on five years of gold M1, price that reached an old extreme kept going more often than it reversed, and fade systems lost heavily in trend years. But riding is no shortcut either. The edge is very small, and the system we built from it made little money. This article lays out the numbers for both sides, plainly.
Terms used in this article
- Ride / Breakout Price climbs to an old high, and you open a Buy hoping it breaks through.
- Fade / Reversal Price climbs to an old high, and you open a Sell hoping it bounces down.
- Swing A stretch of price from one turning point to the next. We find swings with the ZigZag indicator (a line joining the key highs and lows) using fixed settings, not by eye.
- Extension A level beyond the swing start, such as 127.2 or 161.8. For the basics, see How to use Fibonacci.
What we measured
Place Fibonacci on the latest swing, with 0 at the latest turning point and 100 at the swing start, which is the previous swing’s extreme. In the picture, price falls from a high (100) to a low (0), then bounces back up to touch the old high. We logged every first touch of 100, then followed price for another 300 M1 bars, or 5 hours.
Conditions: gold (XAUUSD) M1, 2022 to September 2026. 2026 uses real tick data; earlier years are not real ticks. 715 touches in total. The median swing length is about $10. This part is just counting statistics; no real trades were opened.
The race: dip to 90 first, or break 105 first?
The most direct question is: after touching 100, which side does price reach first? We set two lines: a dip of 10 levels down to 90, and a break of 5 levels up to 105.
Before looking at the real result, we need to know what random price movement would give. The nearer line naturally gets hit first more often. For a random walk, the chance of reaching one line first = distance to the opposite line ÷ total distance.
- Distance to 90 = 100 − 90 = 10 levels
- Distance to 105 = 105 − 100 = 5 levels
- Chance of reaching 90 first, if random = 5 ÷ (10 + 5) = 33.3%
The real result was 23.7%, almost 10 points below random. Fading at 100 to target 90 should work about 1 time in 3 if price leaned neither way, but it actually worked only about 1 time in 4.
Split by year, the dip-back never beat random. Four years were below it, and 2025 matched it exactly.
Careful not to misread this. Price doesn’t shoot through without any pullback. Within those 5 hours, price dipped to 90 in as many as 89% of cases. The difference is the order: most of the time price breaks through first, then pulls back later.
Why old extremes tend to break
Above an old high, two piles of orders are waiting. Traders who sold put their Stop loss there, because “if price goes past it, I’m wrong.” Traders waiting for a breakout put Buy stops in the same place. A seller’s Stop loss is a Buy order once price touches it, so both piles turn into buying at the same moment and push price further. At an old low, it’s the mirror image. So a trader who fades right at the old extreme is standing in front of a force that’s about to be released. For more on these clusters of Stop losses, see ICT and Liquidity sweeps.
Example: an edge, but still beaten by the spread
Take the median swing of $10 and a 0.01 lot on gold (a $1 price move = $1 profit or loss). Each Fibonacci level is 10 × 1% = $0.10.
Fade side Sell at 100, target 90, Stop loss 105
- A win gains 10 levels = $1.00, a loss costs 5 levels = $0.50
- Expected profit = 0.237 × 1.00 − 0.763 × 0.50 = 0.237 − 0.382 = −$0.14
- After the $0.26 spread = −$0.40 per trade
Ride side Buy at 100, target 105, Stop loss 90
- Expected profit = 0.763 × 0.50 − 0.237 × 1.00 = 0.382 − 0.237 = +$0.14
- After the $0.26 spread = −$0.12 per trade
Double-check with the break-even win rate. The ride side risks 1.00 to make 0.50. With no costs, it needs to win 1.00 ÷ 1.50 = 66.7%. It actually won 76.3%, so it has an edge. But add the spread and it needs to win (1.00 + 0.26) ÷ 1.50 = 84%, which it doesn’t reach. (For why costs eat profit on small swings, see Small timeframes and spread cost and Win rate vs break-even.)
The picture changes with bigger swings. On swings larger than $50 (107 touches, using wider race levels: a dip to 61.8 first or a break to 161.8 first), the ride side had an expected profit of +$5.47 per trade at 0.01 lot after spread, while the fade side had −$5.99. That said, 107 touches in five years is not many. Backtest results do not predict the future and are not a return you should expect.
Turning it into a real system: only a little
We built a system that rides a break of 100, only on swings at least $25 long, with the Stop loss at 78.6 and the profit target at 127.2. We ran it in the MT5 Strategy Tester on gold M1 from 2022 to September 2026 (2026 real ticks, earlier years not), with a fixed lot size.
| Year | Trades | Net result ($) |
|---|---|---|
| 2022 | 27 | +82 |
| 2023 | 21 | −2 |
| 2024 | 43 | +42 |
| 2025 | 100 | −73 |
| 2026 | 77 | +384 |
| Total | 268 | +433 |
Over five years it made $433, an average of 433 ÷ 268 = $1.6 per trade, and 384 ÷ 433 = 89% came from 2026 alone. The direction is right, but trades are few (about one a week) and the profit depends on years when the market runs hard. Backtest results do not predict the future and are not a return you should expect.
A statistic that beats random is still a long way from a system that makes money. Costs, trade count and changing market years eat almost all of the gap.
The fade side: pretty in some years, then broken in trend years
We have built 11 fade systems: bounce trades at Fibonacci levels, order grids, and fades at extensions. Looked at over several years, the ride-side systems won and every fade-side system lost.
Waiting to fade at a farther extension didn’t help either. On 2026 data, bounces from 127.2, 161.8 or 200 were no better than random at any level. For example, after reaching 200, price bounced to 161.8 first 55% of the time, while random would give 62% (83 cases). A system fading right at 161.8 was tested in 10 variations, and all 10 lost money (Profit factor 0.51–0.89), because 70% of the time price ran on past 178.6.
The clearest example is a ladder fade system. It fades when price reaches an extension, and if price keeps going, it adds a trade at the next step, up to 9 steps, so the whole set can close in profit on a small pullback. Tested on gold M1:
Another version, which measures swings of different sizes, uses 3 steps and no Stop loss ($3,000 starting balance, 0.02 lot first trade). It tells the same story: profit in years when the market ranged, a blown account in years when it ran in long trends.
| Year | 2022 | 2023 | 2024 | 2025 | 2026 |
|---|---|---|---|---|---|
| Result ($) | +376 | +530 | Account blown | Account blown | +1,000 |
Fade systems like this win so often they look safe, until the day price doesn’t pull back and every added trade loses at once. See Grid, Martingale and averaging down and When the market changes character. Backtest results do not predict the future and are not a return you should expect.
If you ride: give the Stop loss room, and don’t cap the profit
From the numbers above, price that touched an old extreme dipped to 90 in 89% of cases. A Stop loss tucked just below the breakout point tends to get swept before price really moves. That’s why the ride system we tested puts the Stop loss deep, at 78.6. When the Stop loss is farther away, cut the lot size so the money lost per trade stays the same (see Position size from Stop loss distance and Wide or tight Stop loss).
The other side is the profit target. Price broke on to 127.2 in 79% of cases, but reached 200 in only 34%. Ride-side profit comes from the few times price runs far. If you take profit too early, you’re left with small winners that lose to the spread. For how far to set it, see Take profit near or far. And be ready for months of getting nothing, because profit clusters in periods when the market runs hard (Losing streaks are normal).
The fade side loses big on the day price doesn’t come back. The ride side has to let price pull back first, then wait for profit from the few times it runs far. Whichever side you pick, you have to live with its nature.
Common mistakes
- Seeing an old high as a wall. In our data, old extremes broke more often than they bounced, because Stop losses pile up behind them.
- Trusting one year. The ladder fade system made 11,791 in 2026, yet the same set lost the whole account in 2024. Always test a trend year first.
- Forgetting costs. The ride-side edge on small swings is smaller than the spread. A statistical edge doesn’t mean money.
- Stop loss too tight. A real breakout often pulls back to retest first. A Stop loss right below the breakout point gets hit often.
Frequently asked questions
Does this mean you should never trade reversals?
No. It means fading every touch of an old extreme or an extension level gave no edge in the gold M1 data we tested. Some fade systems made money in years when the market ranged, but you need a plan for the years when it trends.
If riding has a statistical edge, why doesn't it make much money?
The edge per trade is small. On a swing of about $10, the expected profit before costs is only about $0.14 per 0.01 lot, less than the spread of about $0.26. You have to pick only the big swings, so there are few trades, and the results still cluster in years when the market moves hard.
What are the 100 and 127.2 numbers in this article?
They are Fibonacci levels measured on the latest swing. 0 is the latest turning point, 100 is where the swing started, which is the previous swing's extreme. 127.2 is an extension that goes past the swing start by another 27.2% of the swing length.
This is general education, not investment advice. Trading is risky and you may lose all of your capital.