Puria Method Forex Strategy: Rules, 2008–2026 Backtest and Puria 2.0 on Gold

The Puria Method is one of the best-known simple forex strategies: three moving averages and MACD, a small take profit and the promise of “at least 50 pips a day”. We published this article many years ago, and thousands of traders have read it since.

In 2026 we tested the strategy with a program over the whole history. First, strictly by the author’s rules: 13,002 trades on four currency pairs over 2008–2024, spread included. Then we improved it step by step and found a version that works on gold (XAUUSD): a profit over 2008–2025 and a profit in 2026, a year we did not use to choose the rules. Below are the original rules, the honest results of the classic version, Puria 2.0 on gold with every trade, and, separately, what happened on USDJPY.

Classic Puria MethodM30 · stop 14 · take profit 15–20−19,276 pips13,002 trades · 4 pairs · 2008–2024 · no pair in profitSee the classic test ↓Puria 2.0 on goldH1 · H4 and weekly trend · mirrored LWMAs · stop 1.5 ATR · take profit 2R+108RXAUUSD · 994 trades · 2008–2026 · 2026: +10.6RSee Puria 2.0 ↓

Puria Method rules (original)

Below are the rules exactly as they have always appeared in this article. These are the rules we tested.

Indicators

Add the following to an H1 or M30 chart (depending on the pair — see the table below):

• Moving Average 1: period 85, method Linear Weighted (LWMA), apply to Low, color red.
• Moving Average 2: period 75, method Linear Weighted, apply to Low, color red.
• Moving Average 3: period 5, method Exponential (EMA), apply to Close, color yellow.
• MACD: fast EMA 15, slow EMA 26, MACD SMA 1.

All indicators come with standard MetaTrader 4 and 5. A ready-made MT4 template is at the end of this section.

Entering a trade

• Buy: the yellow EMA 5 crosses both red LWMAs from below, and MACD confirms — the bar closed above the zero line.
• Sell: the yellow EMA 5 crosses both red LWMAs from above, and the MACD bar closed below zero.
• Stop Loss: 14 pips at most.
• Take Profit: fixed, different for each pair.

Currency pairs, timeframes and take profit

The author’s idea: “better to take a little every day”. A small take profit is hit often, and the stop, according to the author, “happens very rarely”. Let’s check whether that is true.

How we tested the strategy

We used to test the strategy by hand in Forex Tester — a few months on each pair. This time we ran it with a program over every bar, with no gaps and no picking of “convenient” trades.

• Data: MetaTrader 4 price history (Alpari), 2008–2024, terminal time.
• Pairs: GBPUSD, EURUSD, USDJPY (M30) and USDCAD (H1) — they have the longest and cleanest history. Timeframes and take profits strictly from the author’s table.
• Signal: on the closed bar EMA 5 is on the other side of both LWMAs, and on the previous bar it was not yet; MACD on the same bar is above zero (for a buy) or below zero (for a sell).
• Entry: at the open of the next bar. Stop 14 pips, take profit from the table. One trade per pair at a time.
• Spread: deducted from every trade — GBPUSD 1.5 pips, EURUSD 1.0, USDJPY 1.2, USDCAD 1.8.
• Accuracy: if one M30 bar touched both the stop and the take profit, we checked on M5 which came first.

Result: the classic Puria Method loses money

Not a single pair ended in profit. The equity curves go down almost in a straight line — these are not “bad years” but a steady loss over the whole history.

The promise that “the stop happens very rarely” did not hold either: 52–61% of trades closed at the stop. A 14-pip stop on M30 is smaller than the average range of a single GBPUSD candle — normal noise knocks it out.

Why it fails: the spread eats everything

We ran the same test without the spread. GBPUSD then gives +1,010 pips, EURUSD about zero (−146). So the signal itself gives almost no edge, and with a 15–20 pip take profit a 1–1.5 pip spread takes 7–10% of every winning trade and adds as much to every loss. Over thousands of trades this turns zero into a big loss.

What about the old +84% test?

In the manual test of 2018–2019 GBPUSD M30 showed +84% for the year, and losing trades there were “closed at breakeven”. We took the same period (August 2018 – June 2019) and tested two variants: strictly by the rules and with the stop moved to breakeven after +10 pips. Both gave about −100 pips after the spread. Moving to breakeven does not help over the whole history either: −4,234 pips on GBPUSD, −3,998 on EURUSD, −6,036 on USDJPY.

A manual test always risks mistakes: trades get skipped, entries get “adjusted”, the spread is ignored. That is why we now test strategies only with a program and show every trade.

Puria 2.0: the improved version on gold

The signal itself — EMA 5 crossing two LWMAs with MACD confirmation — is a trend signal by nature. A small fixed take profit forces you out before the trend has time to move. We kept the author’s indicators and changed five things. We found the first three earlier, on currency pairs, and the fourth and fifth while working out why the first improved version stopped working on the yen.

1. H1 timeframe instead of M30. Fewer false crossings, and the spread becomes a smaller share of the move.
2. H4 trend filter. Buy only if the last closed H4 candle closed above EMA 50 on H4. Sell only if it closed below.
3. Stop and target based on volatility. Stop Loss = 1.5 × ATR(14) on H1, Take Profit = 2 stops (2R). One trade at a time.
4. “Mirrored” LWMAs. For sells the red LWMA 85 and 75 are built on Low, as in the original. For buys — the same LWMAs, but on High.
5. Weekly trend filter. Buy only if the last closed week closed above EMA 20 on the weekly chart. Sell only if it closed below.

Why “mirrored” LWMAs

In the original both red lines are built on candle lows (Low). The close is almost always above the low, so EMA 5 crosses these lines upward too early — when MACD is usually still below zero. With the rule “MACD on the same candle” there are almost no buys: on gold over 2008–2025 it is 64 buys versus 727 sells. A strategy that almost only sells is doomed in a rising market.

We simply made the rule symmetrical: for buys EMA 5 must cross LWMAs built on highs (High), just as for sells it crosses LWMAs built on lows. Buys and sells become roughly equal (825 and 720), and the result on gold over 2008–2025 grows from +8R to +108R — even before the weekly filter.

The weekly filter removes trades against the big trend: there are fewer trades (994 instead of 1,600 over 2008–2026), the drawdown falls from −43R to −34R and the profit factor rises from 1.11 to 1.17.

How we made sure it is not curve fitting

Fitting parameters to history is easy: try a hundred variants and one of them will surely “make money”. So we tested in a way that kept the latest years out of the choice of rules.

• The rules were chosen on 2008–2023. Stop 1.5 ATR and take profit 2R are the same as in the first improved version for USDJPY; we did not tune them for gold.
• 2024–2026 is a “blind” test. 2024–2025 on Alpari history, 2026 (January–October) on fresh quotes that were not in the main history.
• Blind test result: 2024 −16R, 2025 +15.8R, 2026 +10.6R (40 trades, 17 take profits).
• Neighbouring settings work too. We tested 16 combinations of stop (1–2.5 ATR) and take profit (1.5–3R). All 16 are profitable on 2008–2023, and 15 of 16 are profitable in 2026. So the result does not rest on one “lucky” number.

Statistics: 994 gold trades over 2008–2026

Key numbers

R is the risk of one trade. +107.7R means: if you risk 1% of the account on every trade, the result is about +108% over 19 years without compounding, roughly 5.7% a year. On average +0.11R per trade after the spread. About 50 trades a year — fewer than one a week.

Results by year

14 years out of 19 are profitable. The best year is 2012 (+23.8R), the worst is 2024 (−16R): gold rose all year in a strong trend with sharp short pullbacks, and the strategy kept hitting stops — including 12 in a row. 2025 and 2026 covered this loss.

Buys and sells

Buys made +96.2R (609 trades), sells +11.5R (385 trades). Gold rose for a long time, and most of the profit came from trades with the uptrend. Sells are profitable too, and in 2026 they made more than buys: +8.7R versus +1.9R. We do not recommend switching sells off — that would be fitting to the past.

2026 on gold

The profit was made over several months, not by one lucky trade. A caveat: the Alpari history ends in December 2025, so 2026 was calculated on gold futures quotes (Yahoo Finance, UTC time). The futures price differs slightly from a broker’s spot price, and H4 and weekly candles are cut by UTC — individual trades at your broker may not match.

What it means in money

Gold has become much more expensive, so the stop in dollars has grown: in 2008 1.5 ATR on H1 was about $6, in 2025 about $16, in 2026 about $35 per ounce. Comparing years in dollars makes no sense, so we count in R and in percent of risk. At 1% risk per trade: 2008–2026 about +108%, 2026 (10 months) about +10.6%.

Puria 2.0 trade examples step by step

On every chart: the red lines are LWMA 85 and 75 (on High for buys, on Low for sells), the yellow one is EMA 5, the circle is the crossing, MACD is at the bottom. Top left shows the H4 and weekly trend direction. The blue line is the entry, red is the stop, green is the take profit.

Buy: August 2026 — +$47 per ounce

The week and H4 point up. EMA 5 crossed both LWMAs on High from below, MACD is above zero. Stop $24, take profit $47. Price went up right after the entry and closed at the take profit.

Sell: October 2026 — +$35

The week and H4 turned down. EMA 5 crossed the LWMAs on Low from above, MACD is below zero — a sell with the trend, take profit 2R.

Buy: December 2025 — +$36

Sell: December 2016 — +$7

An old trade for comparison: in 2016 gold cost about $1,150 and the stop was only $3.7. The rules are the same — only the size in dollars changes.

And this is what losses look like

August 2026, buy: there was a crossing, but the rise did not continue — price reversed and took the stop.

October 2026, sell: an entry with the downtrend, but price was thrown up above the stop.

Six trades out of ten close at the stop — this is normal for a strategy with a 2R take profit. The profit comes from the 39% where the take profit is twice the stop.

Gallery: all gold trades 2024–2026

Here are all 148 Puria 2.0 trades on gold for the blind test period — no selection, winners and losers alike. 2024–2025 are Alpari quotes, 2026 are GC futures quotes. Filter by result, direction and year, and click a picture to open the chart.

Trade table 2024–2026

How to calculate the lot size on gold

The stop in Puria 2.0 is different every time, so the lot has to be recalculated for each trade. On XAUUSD 1 lot = 100 ounces: a $1 price move gives $100 on 1 lot or $1 on 0.01 lot. Formula: lot = (account × risk %) / (stop in dollars × 100).

Example: account $5,000, 1% risk = $50, stop $35. Lot = 50 / (35 × 100) ≈ 0.014 → 0.01. With a $35 stop and a minimum lot of 0.01, a 1% risk is possible only from a $3,500 account. With a smaller account the risk per trade will be above 1% — keep this in mind.

USDJPY results: why the yen is at the end of the article

We found the first improved version on the yen, and until 2025 it looked better than any other pair. We are not removing these results — they honestly show how a strategy can stop working. But we moved them to the end because 2026 turned out negative on the yen, and the main reason is interventions and sharp moves of the yen.

First improved version (H1 + H4 trend + stop 1.5 ATR + take profit 2R)

This is Puria 2.0 without the last two steps: LWMAs on Low for both sides, no weekly filter. Test period 2008–2025, spread deducted.

On USDJPY: +2,692 pips on 818 trades over 2008–2025, profit factor 1.18, drawdown −430 pips, 14 profitable years out of 18. The rules were chosen on 2008–2017, and 2018–2025 was profitable too (+1,207 pips). Almost all the profit came from sells (+2,860), buys gave −168: with LWMAs on Low there were only 55 buys against 763 sells.

2026: −341 pips

On fresh USDJPY H1 quotes for January–October 2026 the same rules gave 17 trades, 3 take profits and −341 pips. 16 of the 17 trades were sells, together −303 pips.

What happened: in 2026 the yen kept weakening, USDJPY held around 155–162, and sharp drops in the rate — including those around currency interventions — turned the H4 trend down. The filter saw a “downtrend”, the strategy sold, and the market soon came back up and hit the stop. Conclusion: selling USDJPY on this signal is not tradable right now.

Puria 2.0 on USDJPY: buys work

With the Puria 2.0 rules (mirrored LWMAs + weekly trend) the yen is profitable over the long history too: +79.9R on 1,030 trades over 2008–2026, 11 profitable years out of 19. Over 2008–2025 buys made +43.8R and sells +41.7R. In the 2024–2025 blind test: +18R.

But 2026 is negative here as well: −5.6R on 51 trades. Split by direction, the picture is very clear:

The whole 2026 loss on the yen comes from sells against the weakening yen. All 8 sells closed at the stop. Buys with the trend stayed profitable, if only slightly.

Should you trade USDJPY with the Puria Method?

• Sells — no. As long as interventions and sharp yen reversals are possible, sells on this signal get stopped out again and again.
• Puria 2.0 buys — worth considering as an addition to gold, but with lower risk and on a demo first: their 2026 profit is small.
• The first improved version on the yen (LWMAs on Low, almost only sells) — do not trade it.

An example of a winning sell with the first version — May 2025, +81 pips, and a losing one — April 2025, a 101-pip stop during the market panic:

Gallery: all USDJPY trades 2018–2025 (first improved version)

Here are all 336 trades of the first improved version on the yen over 2018–2025 — the “blind” test period of that version, with no selection. Filter by result, direction and year, and click a picture to open the chart. The 17 trades of 2026 were calculated on other quotes (Yahoo) and are not in the gallery — their result is in the table above.

USDJPY trade table 2018–2025