Honest question: Is this backtest result worth pursuing further?
Hey everyone,
Long-time lurker, first time posting something like this. I've been building an EA based on BBMA methodology (Bollinger Bands + Multiple Moving Averages) applied on XAUUSD, and I've finally reached a point where the backtest numbers look... surprisingly good. Maybe too good, which is exactly why I'm here asking for a reality check from people who actually know what they're doing.
Setup:
- Symbol: XAUUSD
- Chart: M10 (main), M5 + M1 for confirmation
- Period: Jan 1 – May 14, 2026
- Initial deposit: $100 / Leverage 1:1000
- History quality: 100% (tick data)
Results:
| Metric | Value |
|---|---|
| Net Profit | $2,020.64 (+2,020%) |
| Profit Factor | 1.52 |
| Sharpe Ratio | 5.81 |
| Recovery Factor | 1.74 |
| Max Balance DD | 31.21% |
| Max Equity DD | 46.16% |
| Win Rate | 68.31% |
| Total Trades | 284 |
| LR Correlation | 0.70 |
To be fully transparent: this is backtest only. No forward test yet. I'm aware that backtest results can be misleading — curve fitting, broker spread differences, slippage in live conditions, etc. I'm not here to sell anything.
What I genuinely want to know from the more experienced builders here:
- Are there any obvious red flags in these numbers I should investigate before moving forward?
- The max consecutive loss is 13 trades — I've already identified the root cause (a 3-day sharp correction in April where the EA kept buying into a downtrend). Is adding a MaxConsecutiveLoss cooldown parameter a reasonable fix, or is there a better approach?
- At what point would you personally consider a backtest "good enough" to start a forward test?
I know the only real test is live/demo performance. Just trying to learn from people who've been through this process before.
Thanks in advance.
Are there any obvious red flags in these numbers I should investigate before moving forward?
If you want usefull feedback test in points mode with a balance of 100k and 1 lot fixed and post the result. There is no way to judge the true performance of the strategy if money managment is applied.
Hey guys, thanks for the feedback — really useful directions.
@Enrique — ran the retest with $100k balance, fixed 1.0 lot, MM completely off. Here are the clean results:
| Metric | Value |
|---|---|
| Net Profit | $292,335 (+292%) |
| Profit Factor | 1.72 |
| Sharpe Ratio | 4.54 |
| Recovery Factor | 1.96 |
| Max Balance DD | 23.37% ($111,295) |
| Max Equity DD | 29.26% ($149,344) |
| Win Rate | 68.31% (284 trades) |
| Avg Profit Trade | $3,610 |
| Avg Loss Trade | -$4,534 |
| Expected Payoff | $1,029 per trade |
Same period, same symbol, same everything — just stripped out the risk management. PF still holds at 1.72, so the edge seems to be coming from the strategy itself, not from compounding or dynamic lot sizing.
@Dominik — the chart bias point is interesting. I dug into the long/short breakdown:
- Long trades: 192 (win rate 65.62%)
- Short trades: 92 (win rate 73.91%)
If the strategy were simply riding the bullish trend, I'd expect the long win rate to dominate. The fact that shorts are actually outperforming despite the overall bullish context makes me think the bias isn't the main driver here — but I'm open to being wrong on this. Your virtual position approach is something I haven't tried yet, will look into it properly.
One thing I'm not hiding: the max consecutive loss is still 13 trades, translating to -$37,792 on the $100k account — about 37% of balance from a single streak. I've traced it back to a 3-day sharp correction in late April where the EA kept entering longs against the trend.
Next step is implementing a circuit breaker (pause after N consecutive losses + cooldown period) before moving to forward testing. Will post an update once that's done.
Appreciate the engagement — this is exactly the kind of feedback I was looking for
Quick update — found something interesting while digging deeper into the drawdown data.
The two largest single losses both occurred during weekly gap opens in April. The EA entered a position late Friday, market closed, then Monday opened with a significant gap against the position — SL was hit immediately at open with heavy slippage.
This explains a big chunk of the worst drawdown period. It's not a strategy failure per se, it's a market structure issue — weekend gaps on XAUUSD can be brutal, especially during geopolitically active periods like we had in April 2026.
The fix is straightforward: add a Friday close filter — no new entries after, say, 21:00–22:00 server time on Fridays, and force-close any open positions before weekend. This is a pretty standard practice for gold EAs.
Will implement this alongside the circuit breaker and report back. Curious if anyone here has a preferred approach for handling weekend gap risk on XAUUSD specifically — session-based filter or time-based hard close?
I meant profit in pips mode. That is, there is only points, no money, swaps, commisions. You should take them into account just like money managment, but technically are not a part of the strategy.
Overal it does not look too bad, however, from trade 118 it goes sideways. It does not ruin the account, that is good, but you should ask yourself if such periods are mentally sustainable. Little less then 2/3 of the period it does not perform well.
I would suggest making the test longer, and check for a periods with different market dynamics, in 2025 there was a long trending period up, you want to take note of that. You also want to check how it performs in a choppy regime and prolonged downtrend for example. This gives a better insight in rebustness. Also ideally you should exclude outliers, in Januari this year there where a few big moves, not a recurring event. If your strategy made profit on that moves, nice, but these non-systemic events skew the core performance.
Of course, this is in context of trading it yourself and want to avoid selfdeception. If for the market, then i would say you are good to go.
I meant profit in pips mode. That is, there is only points, no money, swaps, commisions. You should take them into account just like money managment, but technically are not a part of the strategy.
Overal it does not look too bad, however, from trade 118 it goes sideways. It does not ruin the account, that is good, but you should ask yourself if such periods are mentally sustainable. Little less then 2/3 of the period it does not perform well.
I would suggest making the test longer, and check for a periods with different market dynamics, in 2025 there was a long trending period up, you want to take note of that. You also want to check how it performs in a choppy regime and prolonged downtrend for example. This gives a better insight in rebustness. Also ideally you should exclude outliers, in Januari this year there where a few big moves, not a recurring event. If your strategy made profit on that moves, nice, but these non-systemic events skew the core performance.
Of course, this is in context of trading it yourself and want to avoid selfdeception. If for the market, then i would say you are good to go.
Thanks for the detailed breakdown, Enrique — really appreciate it.
On the points mode: understood, will rerun with pure points to strip out swap/commission effects entirely.
The trade 118 observation is spot on. Looking at the equity curve more carefully, the EA does go into a consolidation phase from around mid-March onward — it holds ground but doesn't push forward aggressively. I suspect this ties to the market shifting from a clean trending regime to more choppy price action. The MA confluence signals become less decisive in those conditions, which is probably the root cause.
Your suggestion on extending the test period is something I've been thinking about too. I'll run it across 2025 as well to capture the long uptrend phase and see how it holds up — and separately test against a choppy/ranging period to stress-test the signal logic.
On the January outliers — valid concern. Those big moves in early January did contribute disproportionately to the early profit spike. I'll flag those separately and look at the "ex-outlier" performance to get a cleaner read on the core edge.
Appreciate the honest take. Will post an update once I have the extended backtest results.
Hey everyone,
Long-time lurker, first time posting something like this. I've been building an EA based on BBMA methodology (Bollinger Bands + Multiple Moving Averages) applied on XAUUSD, and I've finally reached a point where the backtest numbers look... surprisingly good. Maybe too good, which is exactly why I'm here asking for a reality check from people who actually know what they're doing.
Setup:
- Symbol: XAUUSD
- Chart: M10 (main), M5 + M1 for confirmation
- Period: Jan 1 – May 14, 2026
- Initial deposit: $100 / Leverage 1:1000
- History quality: 100% (tick data)
Results:
| Metric | Value |
|---|---|
| Net Profit | $2,020.64 (+2,020%) |
| Profit Factor | 1.52 |
| Sharpe Ratio | 5.81 |
| Recovery Factor | 1.74 |
| Max Balance DD | 31.21% |
| Max Equity DD | 46.16% |
| Win Rate | 68.31% |
| Total Trades | 284 |
| LR Correlation | 0.70 |
To be fully transparent: this is backtest only. No forward test yet. I'm aware that backtest results can be misleading — curve fitting, broker spread differences, slippage in live conditions, etc. I'm not here to sell anything.
What I genuinely want to know from the more experienced builders here:
- Are there any obvious red flags in these numbers I should investigate before moving forward?
- The max consecutive loss is 13 trades — I've already identified the root cause (a 3-day sharp correction in April where the EA kept buying into a downtrend). Is adding a MaxConsecutiveLoss cooldown parameter a reasonable fix, or is there a better approach?
- At what point would you personally consider a backtest "good enough" to start a forward test?
I know the only real test is live/demo performance. Just trying to learn from people who've been through this process before.
Thanks in advance.
Hey everyone,
Long-time lurker, first time posting something like this. I've been building an EA based on BBMA methodology (Bollinger Bands + Multiple Moving Averages) applied on XAUUSD, and I've finally reached a point where the backtest numbers look... surprisingly good. Maybe too good, which is exactly why I'm here asking for a reality check from people who actually know what they're doing.
Setup:
- Symbol: XAUUSD
- Chart: M10 (main), M5 + M1 for confirmation
- Period: Jan 1 – May 14, 2026
- Initial deposit: $100 / Leverage 1:1000
- History quality: 100% (tick data)
Results:
| Metric | Value |
|---|---|
| Net Profit | $2,020.64 (+2,020%) |
| Profit Factor | 1.52 |
| Sharpe Ratio | 5.81 |
| Recovery Factor | 1.74 |
| Max Balance DD | 31.21% |
| Max Equity DD | 46.16% |
| Win Rate | 68.31% |
| Total Trades | 284 |
| LR Correlation | 0.70 |
To be fully transparent: this is backtest only. No forward test yet. I'm aware that backtest results can be misleading — curve fitting, broker spread differences, slippage in live conditions, etc. I'm not here to sell anything.
What I genuinely want to know from the more experienced builders here:
- Are there any obvious red flags in these numbers I should investigate before moving forward?
- The max consecutive loss is 13 trades — I've already identified the root cause (a 3-day sharp correction in April where the EA kept buying into a downtrend). Is adding a MaxConsecutiveLoss cooldown parameter a reasonable fix, or is there a better approach?
- At what point would you personally consider a backtest "good enough" to start a forward test?
I meant profit in pips mode. That is, there is only points, no money, swaps, commisions. You should take them into account just like money managment, but technically are not a part of the strategy.
Overal it does not look too bad, however, from trade 118 it goes sideways. It does not ruin the account, that is good, but you should ask yourself if such periods are mentally sustainable. Little less then 2/3 of the period it does not perform well.
I would suggest making the test longer, and check for a periods with different market dynamics, in 2025 there was a long trending period up, you want to take note of that. You also want to check how it performs in a choppy regime and prolonged downtrend for example. This gives a better insight in rebustness. Also ideally you should exclude outliers, in Januari this year there where a few big moves, not a recurring event. If your strategy made profit on that moves, nice, but these non-systemic events skew the core performance.
Of course, this is in context of trading it yourself and want to avoid selfdeception. If for the market, then i would say you are good to go.
Yeah go on. Show us 1y, 2025 t'il now. This one seems to be great
Ran the extended backtest as suggested — 17 months (Jan 2025 – May 2026), profit in pips mode, $100k fixed lot 1.0:
PF: 1.39 | Sharpe: 2.57 | Recovery: 4.23 | 928 trades
You were right about the outliers — PF drops from 1.98 to 1.39 in pips mode, which confirms Jan–Feb 2026 skewed the core numbers. The real test was Nov–Dec 2025 (bearish period): -$20k and -$52k respectively, win rate dropped to 31% in December. EA survived but it was painful. Consecutive loss peaked at 14 trades.
The core edge across choppy/ranging months (Apr, Jun–Aug 2025) averages around +$8k/month — modest but consistently positive. I'd say the honest PF without outliers is closer to 1.2–1.3.
Summary 29 bulan (Jan 2024 – Mei 2026) — Pips Mode:
| Metrik | Value |
|---|---|
| Net Profit | $576,843 |
| Profit Factor | 1.34 |
| Recovery Factor | 4.82 |
| LR Correlation | 0.85 |
| Max Balance DD | 32.03% |
| Total Trades | 1,631 |
| Monthly Win Rate | 22/29 = 75.9% |
Extended to 29 months (Jan 2024 – May 2026), profit in pips mode, $100k fixed lot:
PF: 1.34 | Recovery: 4.82 | LR Corr: 0.85 | 1,631 trades | 22/29 months profitable
2024 is the most honest year — 5 loss months out of 12, no outliers, net +$73k. Core PF without the Jan–Mar 2026 big moves is closer to 1.2. Worst consecutive stretch: Nov–Dec 2025 bearish (-$72k combined, win rate dropped to 31% in Dec). EA survived and recovered fully in Jan 2026. Max balance DD 32%.
The strategy has edge but it's modest and regime-dependent.
Bambang Christianto #:
A few suggestions:
1. Expand the Backtest Timeframe: Data from January – May 2026 is too short. It is recommended to go back at least 2-3 years (eg: 1 Jan 2024 – now) to see the EA's resilience in various market conditions (Trending, Sideways, & High Volatility).
2. Use Every Tick Based on Real Ticks: Make sure the backtest method uses the "Every tick based on real ticks" option. Without real tick data, thousands of percent profit results on XAUUSD are often artificial (biased) because they do not take into account the smallest price fluctuations and variable spreads.
3. Beware of Equity Drawdown (DD): The figure 46.16% is a "yellow light". In instruments like XAUUSD, a drawdown of that size is very vulnerable to Margin Call (MC) if slippage or price anomalies occur in the live market.
4. Logical Optimization vs. Optimization Cooldown: Instead of just adding a Cooldown (stop trading) feature, it's better to improve the entry logic filter. Try integrating Market Structure or Price Action as an additional filter so that the EA does not continually make counter-trend entries when sharp corrections occur.
5. Cent Account Trial: Before entering a Standard account, it is highly recommended to carry out a forward test on a Cent account first to see the performance of the EA logic with actual broker execution.
Thanks Bambang, really appreciate the detailed feedback. Let me address each point:
1. Backtest timeframe — Already extended! Just posted 17-month data for Icham above. I also ran 29 months (Jan 2024 – May 2026): PF 1.34, Recovery 4.82, 1,631 trades, 22/29 months profitable. 2024 had 5 loss months — which I think is the most honest representation of core performance.
2. Every tick based on real ticks — Confirmed, all tests use "Every tick based on real ticks" at 99% history quality, not just "Every tick." 160 million ticks for the 29-month run.
3. Equity DD 46.16% — Valid concern. That figure is from an earlier version (v2.9). Current version (v3.1) with circuit breaker + weekend gap filter brought max balance DD down to 32% on the 29-month test. Still not ideal, but significantly better. Working on it.
4. Logic filter vs cooldown — This is the most valuable suggestion. You're right that cooldown alone is a band-aid. The consecutive loss streak in April 2026 was caused by the EA repeatedly entering longs during a 3-day sharp correction — the H1 EMA was still bullish but price was clearly bearish. Adding Market Structure or Price Action filter (e.g. recent swing high/low break confirmation) would be a more surgical fix. This is on the roadmap for v3.2.
5. Cent account forward test — Planned. Will run on cent account before any standard account deployment. No live trading until forward test shows consistency.
Again, appreciate the honest critique — this is exactly the kind of feedback that makes a difference.
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
You agree to website policy and terms of use
Hey everyone,
Long-time lurker, first time posting something like this. I've been building an EA based on BBMA methodology (Bollinger Bands + Multiple Moving Averages) applied on XAUUSD, and I've finally reached a point where the backtest numbers look... surprisingly good. Maybe too good, which is exactly why I'm here asking for a reality check from people who actually know what they're doing.
Setup:
Results:
To be fully transparent: this is backtest only. No forward test yet. I'm aware that backtest results can be misleading — curve fitting, broker spread differences, slippage in live conditions, etc. I'm not here to sell anything.
What I genuinely want to know from the more experienced builders here:
I know the only real test is live/demo performance. Just trying to learn from people who've been through this process before.
Thanks in advance.