AlphaGold reads the same XAUUSD tape institutions read — sweeps, displacements, fair-value gaps — and executes ICT setups on your broker account. Every entry is graded, every position is risk-capped, every signal is logged.
AlphaGold doesn’t enter on hunches. It waits for a specific 5-step institutional footprint in real-time price action. If any step is missing, no position opens.
Price spikes through a recent swing high or low to harvest retail stops, then snaps back. The exact moment smart money loads its real position.
A strong, fast candle ≥1.5× ATR in the reversal direction proves the sweep wasn’t noise. Real money is now in.
The displacement creates an unfilled 3-bar void — a Fair Value Gap. Market makers reliably revisit these.
Price returns to the FVG and breaks the far side, inverting it. Support becomes resistance (or vice versa).
Price comes back to the flipped zone once more. Bot opens the trade with a tight stop — making 1:3 risk-reward achievable.
v5 runs 17 distinct setup configurations in parallel across four timeframes. A 64-feature machine-learning layer sizes every position by predicted edge. Six safety layers keep risk bounded.
Each gate is a distinct combination of timeframe, minimum grade (A, A+, A++), killzone window, and stop-loss cap. Any single gate firing in the absence of conflict opens a trade. Up to 6 positions concurrent.
distinct gates · 6 detectors · multi-position
Scores every candidate setup on win-probability using market structure, time-of-day, volatility regime, change-of-character signals, and cycle position. Sizes position 0.5×–1.5× by confidence.
features · CPCV AUC 0.807 · +0.326 over baseline
CPI, NFP, FOMC, PPI — auto-paused.
Daily −3%, weekly −7%, 30d −10%, absolute −20%, 5-loss streak.
Auto-failover if primary goes silent for 60s. Cross-validated every minute.
The brain knows where we are in gold’s typical cycle (peak / sell-zone / trough / accumulation) and what regime is active (trending up, trending down, range, breakout, quiet). Trade quality is sized against both.
Position size scales inversely to recent volatility so day-to-day equity swings stay predictable. Stop-loss is the deeper of structure-based or 1.5× ATR — never naively percentage-based.
Not a backtested CAGR. Four academic-grade validation tests on 15 years of real XAUUSD tick data. The numbers below are honest — in the paper above, you can read what makes them honest.
The machine-learning layer doesn’t replace the rule. It scores every rule-fired setup against 15 years of identical past setups, then sizes the position accordingly. It cannot trigger trades. It cannot veto them. It can only adjust risk between 50% and 150%.
Grade alone is a coin flip at predicting individual winners (CPCV AUC 0.48 — verified honestly). The strategy is profitable not because grade picks winners more often than chance, but because 1:3 risk-reward pays asymmetrically. Each winner pays 3R, each loser costs 1R. At 50% win rate the math is decisive.
The ML layer adds +0.326 AUC over that baseline by recognizing market context that the rule alone misses: vol regime, cycle phase, recent micro-structure, time-of-day patterns. That’s a real edge, and it’s the only number worth quoting to an investor.
Bring your own broker. We never touch your funds. Your credentials are encrypted in our backend and revocable from your dashboard in one click.
Connect your demo account. See the bot operate end-to-end before committing.
Live broker, full v5 engine, real bracket orders. The setup that just doubled $10K in a 1-year backtest.
For traders running multiple broker accounts or scaling above $25K equity.
No. AlphaGold connects to your broker account using encrypted, revokable credentials. We forward orders. You keep custody of your funds, your account, and your withdrawals. Revoke us anytime from your dashboard.
Any MetaTrader 4 or 5 broker that allows API trading on XAUUSD. We’ve tested with the major retail brokers (IC Markets, Pepperstone, OANDA, FxPro, Exness) and the institutional ones (Interactive Brokers via futures route for accounts above $25K).
Across the 15-year backtest, the bot averages 1-5 trades per week. Sometimes zero for days. Sometimes three in a single session. It’s patient by design — quality of setup over frequency of action.
Point estimate from the 15Y backtest: 19%. The Monte Carlo bootstrap p95 (worst 5% of alternative trade sequences): 28%. Plan psychologically for a 28% drawdown. It will happen at some point if you run the bot long enough.
Because the strategy isn’t trying to win every trade. It’s trying to capture the asymmetry of 1:3 risk-reward. At 50% win rate the math is decisive: every winner is worth 3 losses. We could engineer higher win rates by widening stops, but that breaks the math. We don’t.
AlphaGold pauses entries ±15 minutes around scheduled high-impact USD events (CPI, NFP, FOMC, PPI). The ForexFactory calendar is checked daily. Existing positions are allowed to run to TP/SL through news — the brackets were placed on your broker before the event.
Your stop-loss and take-profit orders sit on your broker, not on our infrastructure. If our backend fails, your positions still close at the levels you and the bot agreed on. We may miss new entries during downtime, but you don’t carry unprotected risk.
Yes. Every connected account gets a live dashboard showing the bot’s current pipeline state (which setup ingredient is detected, current grade, ML confidence score, open positions, equity curve). You can also receive a Telegram message on every open and close.
Connect a demo broker in 90 seconds. Watch the v5 engine work for 14 days. Decide if it’s for you.
Start the 14-day trial →