Forex Reversal Indicators Tested: 7,181 Arrows From Four Rules, None Beat a Coin at Calling the Next 10 Pips, and Engulfing Did Worst

A “reversal indicator” makes one promise: the arrow marks the place where price turns. Not “a signal that might pay”, not “a setup with a good risk-reward”. A turn. So instead of asking only whether the arrows made money, I asked the question the name implies: did the price actually reverse where the arrow was?

The short answer: about as often as it would have anyway. The arrows sat after moves, not before them.

Which reversal indicators did I test, and how?

Most reversal indicators you can download for MT4 are one of four rules with a new name and a new arrow. I coded the rules themselves:

• Fractals (Bill Williams, 2 bars each side): a high with two lower highs on each side is a sell arrow, a low with two higher lows on each side is a buy arrow. It is the core of most “swing point” and “reversal point” tools.
• RSI 14 leaving 30/70: RSI crosses back above 30 = buy, back below 70 = sell.
• Bollinger Bands 20/2 re-entry: a close outside the band followed by a close back inside = an arrow the other way.
• Engulfing candle: a candle whose body swallows the previous body, after 6 bars (30 minutes) of movement the other way.

Every arrow went through three checks:

• Did it mark the real turn? The arrow bar is within 2 bars of the highest high (for a sell) or lowest low (for a buy) of the window from 1 hour before to 2 hours after.
• Which way did price go first? From the close where the arrow became visible, did price reach 5, 10 or 15 pips in the arrow’s direction before the same distance against it? A coin scores about 50%.
• Did it pay? Enter at that close, exit after 1, 2 or 4 hours, 1 pip of cost from my spread and swap measurement, one trade at a time.

Every rule was compared with the same number of random arrows. The data was split at July 15 into two halves.

Did price reverse where the arrows were?

“At the real turn” = the arrow bar is within 2 bars of the extreme of the 3-hour window around it. “N pips first” = share of arrows where price reached N pips the arrow’s way before N pips against it.

Two columns tell different stories. In the first, every rule beat random: arrows land near real turns more often than chance, and Bollinger did it three times as often. In the race columns, nothing beat random. Price after an arrow was a coin flip, and after the RSI and engulfing arrows it was a slightly loaded coin pointing the wrong way.

Why does an arrow near the top still not pay?

Because being at the top is something you learn two hours later. I split every rule’s arrows by whether they turned out to sit at the real turn:

When the arrow was at the real top, it won the race 80.2% of the time. When it was not, it won 35.3%. Nothing on the chart at the moment of the arrow separates the two groups. The screenshot you see on a download page shows the first group. Your account trades both.

This is the same thing I found with support and resistance levels: a line that held looks obvious afterwards, and the ones that broke are not in the picture.

What happened before and after each arrow?

I measured the move in the hour before the arrow and the hour after it, both in the arrow’s direction:

Average pips, before costs. Negative “before” = price had been moving against the arrow.

Every rule fired after a move against it, which is how a reversal rule is built. None of them saw the move come back. A Bollinger buy arrived after a 6.68-pip drop and was followed by exactly nothing. RSI and engulfing arrows were followed by a little more of the same move. The RSI divergence test found the same shape: most of the move was over before the signal.

Did any reversal rule make money?

Pips per trade, 1 pip of cost in “net”.

No. Holding for 2 hours, the best was Bollinger at -0.81 net (+0.19 gross), and that small plus came entirely from the first half: +0.03 net in June and early July, -1.61 after. Held 4 hours, RSI reached +0.25 gross and -0.75 net. Nothing survived a normal spread.

What about the fractal arrows that look perfect on the chart?

They look perfect because of where they are drawn. A fractal needs two more bars to close before it exists, but the arrow is placed on the pivot bar. I ran the same 4,194 arrows twice, once from the pivot bar and once from the bar where the fractal could first be seen:

Same arrows, same price, 10 minutes apart. On gold, in a $1 race, the gap was wider: 84.2% from the drawn bar, 49.5% from the visible one. I measured this lag in the arrow indicators repaint test. Here it answers the reversal question directly: the fractal was a real turn, and it was only known after the turn had happened.

So should you use a reversal indicator?

• Not as an entry on EURUSD M5. 7,181 arrows from four rules, and none beat random arrows at calling the next 10 pips.
• Be most careful with the engulfing candle. After 56.1% of its arrows, the move it was supposed to end went on for another 10 pips first. My candlestick pattern test found the same continuation.
• If you want one, Bollinger re-entry was the least bad. It sat at real turns 33.8% of the time, three times the random rate, and lost the least.
• Judge any reversal indicator from the bar where you could first see the arrow, not the bar it is drawn on. That one change turned the fractal from +0.71 to -1.11 pips per trade.
• How these rules compare with simply trading against the last hour’s move is in reversal indicators vs fading the last hour. The settings grid and the gold results are in the MT4 reversal indicator guide.

Frequently asked questions

What is the most accurate reversal indicator for forex?

In this test, the Bollinger Bands re-entry. 33.8% of its arrows were at the real top or bottom of the surrounding 3 hours, against about 11% for random bars. It still lost 1.04 pips per trade after a 1-pip cost, because it did not predict which way price went next any better than a coin.

Do reversal indicators work on the 5-minute chart?

Not as entries. On 16,926 EURUSD M5 bars, the four rules won the 10-pip race 43.9% to 50.5% of the time. Random arrows won 50.0% to 51.2%.

Is the engulfing candle a good reversal signal?

Not on EURUSD M5. After an engulfing candle that followed a 30-minute move, price went another 10 pips in the old direction before 10 pips in the new one 56.1% of the time.

Do fractal indicators repaint?

A fractal does not move once it is fixed, but it is drawn two bars before it can exist. Read from the bar it is drawn on, fractal arrows won the 5-pip race 65.1% of the time. Read from the bar where they became visible, 49.8%.

Why do reversal arrows look so good on charts?

Because the good ones are easy to see afterwards. Bollinger arrows that turned out to be at the real turn won the 10-pip race 80.2% of the time. The other two-thirds won 35.3%, and nothing at the time of the arrow told the two apart.