RESEARCH
GoldPulse exists because we tested the strategies most gold robots sell, and threw away almost all of them. Here is the method, the full list, and the live log.
Rules, settings and the pass mark are saved to a file before any data is looked at.
Settings are picked on 2012–2021 (news data from mid-2017).
The choice is run a single time on 2022–2026. No re-tuning afterwards.
$0.35/oz per trade, $0.65 at news, and again at double cost.
Survivors are coded in MT5 and tested with real ticks and random delay.
Three months on demo accounts with the rules frozen, published monthly.
THE FULL LIST
PF = profit factor (gross profit ÷ gross loss) after costs. Above 1 makes money, below 1 loses. Grouped where several variants were tested together.
| Strategy tested | Result | Verdict |
|---|---|---|
| Trend-pullback scalper (M1 / M5 / M15) | PF 0.98 / 0.89 / 0.73 with real ticks | REJECTED |
| Pullback scalper + ADX trend filter | No improvement | REJECTED |
| Machine learning, 145 indicators (M5) | PF 0.96 on unseen data | REJECTED |
| Machine learning, 27 H1 setups | Best setup PF 0.66 on unseen data | REJECTED |
| AI as a trade filter (442 signals) | No reliable difference between accepted and rejected trades | REJECTED |
| US news surprise on gold | PF 1.20 / 1.70 / 2.06 by period · MT5 real ticks PF 2.20 | KEPT |
| Spike fade after big H1 candles | PF 1.15 but 35% drawdown | REJECTED |
| Momentum stop-order scalper | +$16,900 in default tester → −$995 with real ticks | REJECTED |
| News strategy on silver | PF 0.77 after costs | REJECTED |
| US news on 7 forex pairs | PF 0.81–1.07 | REJECTED |
| Daily trend following, 8 markets | −4.6% a year after costs and swap | REJECTED |
| AI reading 80 Fed statements | PF 0.60 | REJECTED |
| Foreign news on their own currency | EUR 0.65 · GBP 0.79 · JPY 0.44 | REJECTED |
| US news on indices & US stocks | PF 0.85–0.91 | REJECTED |
| News trades: let winners run / size by surprise / wait for confirmation | No real improvement | REJECTED |
| H1 channel (Donchian) breakout | PF 0.83–0.92 | REJECTED |
| Daily volatility breakout on gold | PF 1.18 in both periods | KEPT |
| + skip extreme-range days and NFP / CPI / Fed days | PF 1.24 / 1.31 · MT5 real ticks PF 1.29 | KEPT |
| Breakout: 1% risk sizing, trend filter, trailing or 2R exits | No improvement on unseen data | REJECTED |
| Previous-day, wide-level and weekly breakouts; silver breakout | Failed the pass rule | REJECTED |
| Failed-breakout reversal | PF 1.50 (2012–21) → 0.75 (2022–26) | REJECTED |
| Tight-stop, 7× target session breakout | PF 0.57–1.00 | REJECTED |
| Asian-range fade | PF 0.72–0.76 | REJECTED |
| H1 EMA swing trend, 1:2 target | PF 0.77–0.95 | REJECTED |
| Breakout rules on 14 other markets | None passed in both periods | REJECTED |
| Forex intraday fade (5 pairs) | PF 0.77–0.89 | REJECTED |
| London breakout of the Asian range (8 markets) | Combined PF 0.95 | REJECTED |
| "Smart money" structure break + retest (9 markets) | PF 0.78–0.88 | REJECTED |
Research on historical data. Simulated results do not guarantee future results. Details: how we backtest.
PROOF, NOT PROMISES
Most bots are tuned until the backtest looks perfect. We did the opposite: rules were written down before each test, chosen on older data (2012–2021), checked once on newer data (2022–2026), with real spreads and again at double costs. Everything that failed was thrown away, even ideas that looked great for ten years.
Only gold stays above 1.15 in both periods. That's why GoldPulse is sold for XAUUSD only.
Updated: November 2026
RESEARCH LOG
Every month we post what the bots did on the demo accounts and how it compares with the backtest, including the losing months.
Catalyst runs on a $1,000 demo account and Breakout on its own $3,000 demo account, both at 0.01 lot. AI Observer records a forecast every trading morning. The rules are frozen for three months: no tweaks, no manual trades.
What we expect: about 2–3 Catalyst trades and about 9 Breakout trades a month, with losing streaks along the way. First results at the start of December.
LAUNCHING EARLY 2027
Monthly demo results, the AI Observer scorecard and the launch price first. About one email a month.
No spam, unsubscribe any time. Privacy policy.