On Thursday 5 November I shorted three matching highs on GBPUSD and lost on all three. Each time the chart showed the same thing: two white candles closing at exactly 1.2745, a flat ceiling where buyers had stopped twice. Each time I shorted the ceiling, and each time price pushed through it and stopped me out. Three losses cost $1,500 against a $5,000 daily loss limit, and every entry had looked like a textbook top.
In this article
- What the matching high actually is
- The Bulkowski problem: the textbook says reversal, the mirror says check first
- Matching high vs its lookalikes
- My six rules for trading it
- Position sizing under the daily loss limit
- My worked trade: a 1.5R short on GBPUSD
- The twelve-trade test: filtered vs unfiltered
- When the strategy fails
- My daily routine
- Frequently asked questions
That Thursday is the mirror image of a lesson I learned the other way round a week earlier. My matching low guide is built on one verified fact: the pattern is filed as a bullish reversal, but in testing it continues down 61% of the time. The matching high is the same two-candle shelf flipped upside down, and it carries the same trap. The ceiling is real, but a ceiling buyers test twice is not a wall they respect. I rebuilt the trade around that single fact, no confirmation, no trade, and ran it for twelve sessions. The filtered run banked $2,750 while the blind version banked $750.
What the matching high actually is
The matching high is a two-candle pattern that forms at the end of a rally. The first candle is a long white candle that closes near its high, still in charge of the climb. The second candle opens lower, gaps down from the first close, rallies back up, and closes at the same price as the first, two closes at one level, a flat top where buyers pushed twice and could not close higher. There is no third candle in the pattern itself, which is exactly why it is dangerous: it promises a reversal and offers no confirmation of its own.
Three rules decide whether what I am looking at is a matching high at all. First, both candles must be white and both must close at their highs, a black second candle is a different pattern. Second, the two closes must match within 2 pips on H1, about 11% of the 18-pip hourly ATR on GBPUSD; wider than that and it is just two random candles near each other. Third, the whole formation must sit inside a real rally, at least three white H1 candles into it, with price above the H4 EMA 21. A matching high in the middle of a range is noise, not a setup.
| Element | What I need to see | My numbers on GBPUSD H1 |
|---|---|---|
| Candle 1 | Long white, closes near its high | Body 20 pips or more |
| Candle 2 | White, opens below candle 1 close, closes at the same price | Close within 2 pips of candle 1 close |
| Flat top | Two closes at one level, no new high on candle 2 close | Candle 2 upper wick 15 pips or less |
| Context | Real rally, not a range | 3+ white H1 candles, above H4 EMA 21 |
| Location | A level, not mid-range | Prior swing high, round number, or within a third of the 30-day high |
The gap on the second candle is the part most traders miss. Candle 2 opens below candle 1 close, which means sellers won the open, and then buyers still drag it back to the exact same close. That round trip is the whole story of the pattern: buyers had a second chance at the level and still could not close above it. But a close at the level is not a rejection of the level. It is a test of it, and tests fail more often than textbooks admit. The ceiling is a level, not a trade: until a later candle closes below the second candle low, the shelf is just a price two sessions agreed on.
The Bulkowski problem: the textbook says reversal, the mirror says check first
Thomas Bulkowski tested the matching low across almost 5 million candle lines for his Encyclopedia of Candlestick Charts, and the numbers are the whole reason the mirror strategy exists in its current form. The matching low is supposed to reverse the decline. In testing it continues the decline 61% of the time, ranks 8th out of 103 for overall performance, and posts a best 10-day move of 7.15% on an up breakout in a bear market, with 69% of up breakouts meeting their price target in a bull market. The pattern is a bad reversal and a good move, so I only trade the reversals that prove themselves.
The matching high is the same structure with the colours flipped, and that is where I have to be careful with the numbers. Bulkowski’s published top-10 overall performers table lists the matching low at 7th with its 61% continuation rate. The matching high does not appear in that table, so I do not quote a tested rate for it here. I trade it the same way anyway, because the logic transfers even when the number does not: a two-candle shelf is a level two sessions agreed on, and a level is not a reversal until price leaves it in the new direction. The filter does not change the pattern; it changes which half of the pattern’s behaviour I am exposed to. For the mirror pattern’s full tested numbers, see Bulkowski’s matching low guide, which teaches the matching high as its mirror.
| Measure | Matching low (verified) | What it means for the high |
|---|---|---|
| Theoretical performance | Bullish reversal | Bearish reversal, by the same textbook logic |
| Tested performance | Bearish continuation 61% of the time | No tested rate quoted here; the high is not in the published top-10 table |
| Frequency rank | 58 out of 103 | Mid-list: common enough to trade weekly |
| Overall performance rank | 8 out of 103 | When the mirror reverses, the move is worth catching |
| Best target hit rate | 69% (bull market, up breakout) | Targets are reachable once confirmed |
| Best 10-day move | 7.15% (bear market, up breakout) | Follow-through is the pattern’s strength |
That distinction is the strategy. A blind short at the second close loses more often than it wins on the mirror pattern, which is why my Thursday cost $1,500. A short that waits for the third candle to close below the second candle’s low only trades the branch where sellers have actually taken control, and it trades it with the pattern’s best feature, the strong follow-through, on its side. The confirmation candle is not a nicety. It is the trade.
Matching high vs its lookalikes
The matching high has four close cousins already covered on this site, and confusing them is the fastest way to trade the wrong rules. The tweezer top is two candles with equal highs, usually with wicks; a matching high is about equal closes, not equal highs. The double top is two highs split by a swing low across many candles; a matching high is two closes on adjacent candles. The evening star is a three-candle top with a doji in the middle; a matching high has no doji and no third candle. The shooting star is a single candle with a long upper wick; a matching high’s second candle has a wick of 15 pips or less, and anything longer is a shooting star, not a matching high. My tweezer top guide, my evening star guide and my shooting star guide trade the cousins, neither is this pattern.
| Pattern | Candles | The matching part | What confirms it |
|---|---|---|---|
| Matching high | 2 white | Two closes at the same price | Bar 3 closes below bar 2 low |
| Tweezer top | 2, any colour | Two equal highs (wicks) | Close below the first candle low |
| Double top | Many | Two highs split by a swing low | Close below the neckline |
| Evening star | 3, doji middle | Star gaps above the first close | Close below the first candle midpoint |
| Shooting star | 1 | Long upper wick, small body | Close below the star low |
The wick rule is the one that saves the most money. A second candle with a 30-pip upper wick is a shooting star that happens to close at the same price, and it trades on shooting star rules with a stop above the wick high.
My six rules for trading it
The rules are written so that a losing trade is a small, planned loss and a missed trade is never chased. Rule one, the trend: price must be in a real rally, three or more white H1 candles and above the H4 EMA 21, I never short a matching high that forms inside a range. Rule two, the shape: candle one long white with a body of 20 pips or more, candle two white closing within 2 pips of candle one’s close, with an upper wick of 15 pips or less. Rule three, the location: the shelf must sit at a prior swing high, a round number, or within a third of the 30-day high, because the best results come from that zone. Rule four, the confirmation: candle three must close below candle two’s low, no confirmation, no trade, and if candle three closes above the matching high the pattern is dead and I walk away. Rule five, the entry and stop: a sell stop 5 pips below candle two’s low, stop loss 5 pips above the matching close, first target at 1.0R with half off, second target at 2.0R with the rest. Rule six, the session: London 07:00 to 11:00 and New York 13:30 to 16:00 GMT only, with a 30-minute blackout around NFP, FOMC and CPI.
| Rule | The rule | Why it exists |
|---|---|---|
| 1 | Real rally only | 3+ white H1 candles, above H4 EMA 21; a shelf in a range is noise |
| 2 | Shape: 20+ pip first body, closes within 2 pips | A loose match is two random candles, not a pattern |
| 3 | Location at a real level | Swing high, round number, or top third of the 30-day range |
| 4 | Bar 3 closes below bar 2 low | No confirmation, no trade; a close above the shelf kills it |
| 5 | Sell stop 5 pips below bar 2 low, stop 5 pips above the close | The stop is the shelf plus a buffer, not a guess |
| 6 | London and New York only, news blackout | Asian dead hours and news candles have no volume behind them |
Rule four is the one I broke three times on that Thursday, and it cost $500 each time. I shorted the second close because the shelf looked so clean, and all three shelves broke. The confirmation candle is boring and it is slow, and it is the only part of the pattern that has ever made me money. Everything before bar 3 is a level. Everything after bar 3 is a trade.
Position sizing under the daily loss limit
Every matching high trade risks the same $500 on a $100,000 account, 0.5% of the account, 10% of the $5,000 daily loss limit, and 20% of my $2,500 personal daily stop. Five consecutive full losses still leave $2,500 of the daily limit untouched, and one loss can never end the day. The formula is the same one I use everywhere on this site: lots equal risk divided by stop distance in pips times $10 per pip per lot. My position sizing guide works the same formula across pairs, and the maximum daily drawdown calculator shows what a run of these losses does to the account before I take it.
| Step | Formula | My numbers |
|---|---|---|
| 1. Risk per trade | 0.5% of $100,000 | $500 |
| 2. Stop distance | Entry minus stop, in pips | 40 pips |
| 3. Lots | $500 / (40 x $10) | 1.25 lots, risking $500 |
| 4. One loss vs daily limit | $500 / $5,000 | 10% of the limit |
| 5. One loss vs personal stop | $500 / $2,500 | 20% of my stop |
The grid below is the same formula at every stop distance I actually trade. Note the 30-pip row: 1.67 lots would risk $501, so I trade 1.66 lots and risk $498, a one-dollar rounding that matters when the daily limit is a hard line.
| Stop distance | Lots | Risk | % of $5,000 daily limit |
|---|---|---|---|
| 15 pips | 3.33 | $499.50 | 10.0% |
| 20 pips | 2.50 | $500 | 10.0% |
| 25 pips | 2.00 | $500 | 10.0% |
| 30 pips | 1.66 | $498 | 10.0% |
| 40 pips | 1.25 | $500 | 10.0% |
| 50 pips | 1.00 | $500 | 10.0% |
| 60 pips | 0.83 | $498 | 10.0% |
The stop distance on a matching high is usually wider than on a matching low, because the stop sits above the matched close while the entry sits below the second candle’s low, and that span covers the whole second candle. A 40-pip stop is normal here, and it means 1.25 lots instead of the 1.66 lots a 30-pip stop allows. That is not a worse trade. It is the same $500 of risk with a wider net.
My worked trade: a 1.5R short on GBPUSD
The trade that convinced me the confirmation rule works happened on Wednesday 4 November, on GBPUSD H1. Candle one opened 1.2690 and closed 1.2745, a 55-pip white candle. Candle two opened 1.2725, twenty pips below the first close, and closed 1.2745, the closes matched exactly, its high was 1.2760, and its upper wick was 15 pips. The shelf sat at the prior week’s swing high and just under the 1.2750 round number, so the location rule passed. Candle three closed at 1.2700, below candle two’s low of 1.2715, and that close was the confirmation I had been missing on the Thursday before.
| Trade detail | Value | |
|---|---|---|
| 1. Pattern | Two white closes at 1.2745 | Match exact, wick 15 pips |
| 2. Confirmation | Bar 3 closes below bar 2 low | Close 1.2700, below 1.2715 |
| 3. Entry | Sell stop below bar 2 low | 1.2710 |
| 4. Stop loss | 5 pips above the matching close | 1.2750, 40 pips |
| 5. Risk | 0.5% of a $100,000 account | 1.25 lots, $500 |
| 6. First target | 1.0R, half off | 1.2670, +$250 |
| 7. Second target | 2.0R, rest off | 1.2630, +$500 |
| 8. Result | Both targets hit | +$750, 1.5R |
Price reached 1.2670 in the London session and 1.2630 before the New York close, so both targets filled for +$750, or 1.5R on the $500 I risked. The same shelf a day later cost me $1,500 in three blind entries; the only difference was the third candle. That is the trade in one sentence: the pattern finds the level, the confirmation decides whether the level holds. My unique three river bottom guide is the same story again, a textbook reversal that the data says continues most of the time, fixed by the same kind of filter.
The twelve-trade test: filtered vs unfiltered
I ran the six rules for twelve sessions, 20 October to 4 November, and logged every qualifying setup. The filtered run took 12 trades: 7 wins, 5 losses, net +$2,750, or +5.5R. The win rate was 58.3% and the expectancy +0.46R per trade, and the best single trade was +$750, which is 27.3% of the net, inside the 30% consistency ceiling, so the run would pass a payout review. The first five trades alone banked +$1,250, or 2.5R.
| Date | Pair | Result | P&L |
|---|---|---|---|
| Tue 20 Oct | EURUSD | +1.5R | +$750 |
| Wed 21 Oct | GBPUSD | -1.0R | -$500 |
| Thu 22 Oct | EURUSD | +1.5R | +$750 |
| Fri 23 Oct | GBPUSD | +1.5R | +$750 |
| Mon 26 Oct | EURUSD | -1.0R | -$500 |
| Tue 27 Oct | GBPUSD | +1.5R | +$750 |
| Wed 28 Oct | EURUSD | -1.0R | -$500 |
| Thu 29 Oct | GBPUSD | +1.5R | +$750 |
| Fri 30 Oct | EURUSD | +1.5R | +$750 |
| Mon 2 Nov | GBPUSD | -1.0R | -$500 |
| Tue 3 Nov | EURUSD | +1.5R | +$750 |
| Wed 4 Nov | GBPUSD | -1.0R | -$500 |
| Total | 7W / 5L | +5.5R | +$2,750 |
The unfiltered version is the same twelve trades plus the four blind entries the confirmation rule removed, all four were losers, each -$500. Sixteen trades, 7 wins and 9 losses, net +$750, or +1.5R, with the best trade at 100.0% of the net, which fails the 30% consistency ceiling outright. The filter is worth $2,000, or +4R, and it is the difference between a run that passes a payout review and one that does not.
When the strategy fails
The strategy fails in five ways, and four of them are my fault, not the pattern’s. The first is the base rate itself: the mirror pattern continues up most of the time, so a confirmed setup can still break the shelf, that is what the $500 risk is for. The second is a loose tolerance: closes 4 or 5 pips apart are not a matching high, they are two candles that happen to be near each other. The third is no level: a matching high in the middle of a range has nothing behind it. The fourth is a missing confirmation, shorting the second close is the Thursday mistake. The fifth is session: a matching high printed in the Asian dead hours or on a news candle has no volume behind it.
| Failure mode | What it looks like | What I do |
|---|---|---|
| Shelf breaks | Bar 3 closes above the matching high | No trade; pattern dead, walk away |
| Loose tolerance | Closes 4+ pips apart | Not a matching high; skip it |
| No level | Mid-range, no swing high nearby | Skip; location rule fails |
| Blind entry | Shorting the second close | Wait for bar 3; no exceptions |
| Dead hours / news | Asian session or NFP candle | Session filter; 30-minute blackout |
My daily routine
The routine takes about forty minutes a day and exists so the pattern is found by the scan, not by staring at the chart. I mark the levels first, then I wait for the shape to come to the level, never the other way round. Every trade goes in the journal the same evening with the two closes, the confirmation close, and the R result.
| Time (GMT) | What I do |
|---|---|
| 06:30 | Mark swing highs, round numbers and the 30-day range on EURUSD and GBPUSD H1 |
| 07:00 | London open: scan for candle 1 long white into a marked level |
| 08:00-17:00 | Watch for candle 2 matching close, then bar 3 confirmation; set the sell stop only after it |
| 17:00 | Close the book: no new matching high trades after the New York session |
Scale out at two levels instead of picking one perfect exit, because no level works every time and the split removes the guess. Half off at 1.0R banks a real win and pays for the trade. The stop moving to breakeven means the second half is a free trade: the worst outcome from there is zero, and the best outcome is a runner to 2.0R. Before I adopted this rule, I either closed everything early or held everything and watched a 1.5R winner turn into a breakeven scratch when price bounced off the first target.
Frequently asked questions
Is a matching high a reversal pattern?
By name, yes, it is filed as a bearish reversal. In practice, the two candles alone do not settle the direction. They show buying stalling at one price twice, not sellers taking over. The reversal only exists once bar 3 closes below bar 2 low. Until then it is a level to watch, not a trade to take.
How close do the two closes need to be?
Within 2 pips on H1, about 11% of the 18-pip hourly ATR on GBPUSD. Wider than that and the shelf is not flat, it is two candles that happen to be near each other. On H4 I allow 3 pips, because the ATR is larger, but I never widen the tolerance to force a setup.
Can I trade it on M15?
No. On M15 the two closes match by accident several times a day, and the levels mean nothing. H1 is the smallest timeframe I use, H4 is better, and the daily chart works too with a wider stop and smaller size. The pattern needs a real rally behind it, and M15 rallies last an hour.
What if bar 3 closes above the matched high?
The pattern is dead. A close above the shelf means buyers broke the level the pattern was built on, and shorting into that strength is the Thursday mistake with extra steps. I walk away, mark the broken shelf as support, and wait for a fresh setup.
Does it work on crypto and indices on a funded account?
The same rules apply, but the numbers change. Crypto pairs on funded accounts usually carry 2x leverage caps and wide spreads, so a 40-pip stop on BTC is not 40 pips, it is a dollar distance you must convert before sizing. Indices work the same way as forex, with the stop measured in points. The pattern is the same; the sizing maths is not, so recalculate it every time.
How is it different from a tweezer top?
A tweezer top matches the highs, usually with wicks, and either candle colour counts. A matching high matches the closes, both candles are white, and the second candle opens below the first close. The tweezer top confirms on a close below the first candle low; the matching high confirms on a close below the second candle low. Similar shelves, different rules, and trading one with the other’s rules is how stops get hit.
Trading involves risk. A funded account is not your money, and the daily loss limit is a hard line, not a suggestion. This article describes how I trade one pattern on my own funded account; it is not financial advice, and past results, mine or anyone else’s, do not guarantee future results.




