Browse all reviews, rankings, guides, strategies, and trust documents.
Golden Key — Lifetime Access to All StrategiesLearn More →
Independent review Verified live
GoldStrike AI
Premium ML-driven XAUUSD specialist for MetaTrader 5
Our hands-on review of GoldStrike AI — how the weekly-retrained ML model makes entries and exits, the Tier-1 ECN broker + VPS setup it needs, live verified results, and an honest look at drawdown during XAUUSD regime shifts.
GoldStrike AI is the premium XAUUSD specialist in the FxRobotEasy line — an end-to-end, ML-driven MetaTrader 5 Expert Advisor trained on multi-year tick-level gold data and retrained weekly via the .SetAnalyze cloud. Built for traders running serious capital on gold who want the most aggressive published edge in the line. A Tier-1 ECN broker with sub-1.5-pip XAUUSD spread is required.
What is GoldStrike AI
GoldStrike AI is the most ambitious system in the FxRobotEasy line. Where Scalperology AI is a rules-based scalper with a neural-network filter on top, GoldStrike is end-to-end ML: the entry decision, exit timing, and position sizing all come from a single trained model. The model is retrained weekly on the past 6 months of tick-level XAUUSD data through our .SetAnalyze cloud infrastructure, with strict out-of-sample validation before each deployment.
This architecture choice trades inspectability for adaptiveness. A rules-based system like Scalperology has predictable behaviour you can demo-test in known regimes; a model-driven system like GoldStrike has behaviour that evolves over time as the retraining absorbs fresh data. Users who want full control over every parameter will prefer Scalperology; users who want a system that adapts continuously to XAUUSD regime shifts will prefer GoldStrike.
The premium positioning reflects the ongoing infrastructure cost of weekly retraining — dedicated GPU runtime, out-of-sample validation, and regression testing against the live reference account every week. It also reflects the priority support tier that comes with the licence: direct access to the lead developer for setup questions and ongoing regime-specific guidance.
Strategy mechanics
Model architecture. GoldStrike uses a gradient-boosted tree ensemble trained on ~50 engineered features. The features fall into three families: price-derived (multi-timeframe momentum, volatility, candle pattern indicators), microstructure (spread regime, tick-bias, order-flow proxies derived from time-and-sales), and session metadata (time-of-day, day-of-week, hours-to-major-event-window). The ensemble outputs a directional confidence score plus an estimated position-size scale.
Entry.The model evaluates current state on every completed M5 candle. If directional confidence exceeds 0.65 and the spread is within tolerance, the EA enters in the predicted direction. Position size is the model's estimated scale multiplied by the global 1% risk-per-trade budget — so a low-confidence signal opens a smaller position even though directional confidence already passed the threshold.
Exit.Take-profit and stop-loss are dynamic, estimated by the model based on the entry context. Typical range: TP at 25-60 pips, SL at 15-30 pips. The model also outputs an "exit early" signal if regime shifts during the trade — this closes the position at market regardless of whether the original TP or SL has been hit. This early-exit logic is the main reason GoldStrike survives regime transitions that catch rules-based systems by surprise.
Weekly retrain.Every Sunday during market closure, the retraining pipeline pulls the past week of tick data from the reference broker, retrains the ensemble on the rolling 6-month window, runs walk-forward validation across the past 8 weeks held-out, and pushes the validated model to active licenses via .SetAnalyze cloud. The entire cycle takes 4-6 hours. Users don't need to take any action; the model updates in place on Monday open.
Recommended setup
Pair: XAUUSD on M5. Single-pair system — do not attach to other pairs.
Broker: Tier-1 ECN with sub-1.5-pip XAUUSD spread during liquid hours. Tested on IC Markets Raw, Pepperstone Razor, Tickmill Pro, FP Markets Raw.
VPS: required. Windows Server 2019/2022. Sub-30 ms ping to broker server. Equinix LD4 (London) preferred but not strictly required if your broker is non-LD4.
Risk per trade: 1% default. The ML position-size scale modulates this — actual per-trade risk varies 0.5-1.5% based on signal confidence.
.SetAnalyze cloud: included in licence. Recommended preset auto-updates weekly. Manual presets exist but not recommended.
Account size: $2,999 minimum, calibrated against max-drawdown expectations rather than the licence price.
Performance characteristics
On forward-tested live accounts across 2024-2025, GoldStrike's published reference account produced an annualised return of 60-95% at a maximum drawdown of 12-18%. Win rate sits at 55-62% — higher than rules-based scalpers because the model filters out lower-confidence setups. Profit factor runs 1.7-2.2. Sharpe ratio around 1.4-1.7 on annualised daily returns.
Equity curve: relatively smooth during stable XAUUSD regimes (most of the time), punctuated by 4-8 week drawdown phases during regime transitions where the weekly retrain catches up. Recovery is characteristically explosive once the model re-adapts — a typical recovery sees the equity curve reclaim a 15% drawdown in 6-10 weeks.
When to pick GoldStrike AI vs Scalperology AI
Data as of August 16, 2026; method: Editorial comparison based on 2024-2025 live-account data + design-doc review.; source: www.fxroboteasy.com/experts/goldstrike-ai
GoldStrike vs Scalperology — which to pick
Direct comparison of the two XAUUSD specialists in the FxRobotEasy line.
Aspect
Scalperology AI
GoldStrike AI
Architecture
Rules + neural-net filter
End-to-end ML
Timeframe
M1
M5
Min capital
$2,000
$2,999
Annual return target
30-60%
60-95%
Max drawdown
8-12%
12-18%
Adaptation cadence
Quarterly
Weekly
Inspectability
Full — rules are documented
Limited — model is black-box
Best for
Deterministic operators, value-conscious
Capital-rich operators, adaptive-bias
For most XAUUSD traders deploying $500-2,000, Scalperology is the better-value entry — you get a well-tested rules-based edge for a fraction of the cost. For traders deploying $3,000+ who specifically want the ML-driven adaptive edge and are comfortable with the less-inspectable model behaviour, GoldStrike is the right pick. Some operators run both — Scalperology on a $5,000 account and GoldStrike on a separate $10,000 account — to capture both flavours of XAUUSD edge.
Risk profile
XAUUSD is highly volatile and GoldStrike runs an aggressive ML model. Typical historical drawdown ranges 12-18% across tested samples; worst observed during regime stress periods (Fed tightening, geopolitical events) was 20%. The model is retrained weekly, which means past-period performance is not a clean predictor of future performance even with the same EA file. Position sizing is the primary risk management tool; if you can't tolerate 20% drawdown, do not deploy.
Pricing and licensing
GoldStrike AI is sold as a one-time licence across several tiers — see current pricing on the product page. Every tier includes: the EA file, .SetAnalyze cloud access with weekly model retrains for the first 12 months, priority support with direct lead-developer access for setup questions, a money-back guarantee (up to 30 days, depending on tier), and free EA updates for the first 12 months. After the first 12 months, an optional annual renewal continues the weekly retrains — otherwise the model freezes at the last validated week and degrades as XAUUSD regime shifts.
GoldStrike AI FAQ
Why is GoldStrike AI more expensive than other FxRobotEasy EAs?
GoldStrike AI is the premium XAUUSD specialist in the EASY Bots line, priced as the top tier. The premium reflects three things: ongoing weekly model-retraining infrastructure, priority support with direct access to the lead developer, and the higher minimum deposit that the ML risk-control assumptions are calibrated against.
Where Scalperology is a rules-based scalper with a neural-network filter, GoldStrike is ML-driven end-to-end — the entry decision, the exit timing, and the position-sizing scale all come from a model that is retrained weekly on the latest tick data. That retraining is non-trivial infrastructure: dedicated GPU runtime, out-of-sample validation, regression tests against the live reference account. The licence price funds that ongoing operation. If you do not need the ML edge over a rules-based scalper, Scalperology AI is the better-value entry point.
Should I pick GoldStrike AI or Scalperology AI?
If you have at least $3,000 to deploy on XAUUSD and want the most aggressive published edge in our line, GoldStrike AI is built for that. If you have $500–2,000 or prefer a deterministic rules-based system you can fully inspect, Scalperology AI is the better fit. Both trade XAUUSD; GoldStrike runs M5 with ML signals, Scalperology runs M1 with a rules + ML-filter hybrid.
GoldStrike's deployed model is updated weekly, which means its behaviour evolves over time as the model retrains on fresh data. Scalperology is largely deterministic — the rules don't change between releases, and the neural-network filter is retrained on a much slower cadence. Inspectability matters to some traders more than others: if you want a strategy you can fully understand and demo-test in known regimes, Scalperology is more transparent. If you want a model that adapts continuously and you trust the retraining pipeline, GoldStrike is the more ambitious bet.
What broker does GoldStrike AI require?
Tier-1 ECN broker with sub-1.5-pip XAUUSD spreads during liquid hours (London + NY). Sub-30 ms ping to the broker's trade server. We have tested live deployment on IC Markets Raw, Pepperstone Razor, Tickmill Pro, and FP Markets Raw — all four meet the threshold. Standard accounts at market-maker brokers will not produce the published performance.
XAUUSD spread is the single biggest determinant of ML-driven scalping profitability on gold. At 1.5 pip spread, GoldStrike's edge is robust; at 3+ pip spread, the edge erodes to near zero on average — every winning trade has to overcome additional cost that the model didn't factor in during training. We strongly recommend opening a Raw or Razor account specifically for GoldStrike rather than running it on whatever broker you already have. The annual cost difference (lower-tier vs Tier-1 ECN) is recovered in the first 2-3 months of operation.
How does the weekly model retraining work?
Every Sunday during the weekend market closure, our infrastructure pulls the past week of tick data from the reference broker, retrains the ML model on a rolling window, runs out-of-sample validation, and pushes the updated model to active licenses via .SetAnalyze cloud. The update is automatic — no user action required. If a particular week's retraining fails validation, the previous week's model continues until the next successful run.
The model architecture uses gradient-boosted trees with ~50 engineered features from price, microstructure, and session metadata. Training data is a 6-month rolling window weighted toward recent weeks. Validation: walk-forward across the 8 most recent weeks not in the training set. We require profit-factor ≥ 1.4 and max-drawdown ≤ 8% on validation before the model goes live. Failed validations are logged but not user-visible (we don't want to expose the failure rate as it's noisy at week-level).
When does GoldStrike AI fail?
GoldStrike underperforms during regime shifts in XAUUSD — typically caused by central bank interventions or sudden geopolitical events that move gold price faster than the weekly retrain can absorb. Drawdown periods of 10-15% across 4-8 weeks are not unusual during such shifts. The retrain cycle typically catches up within 4-6 weeks.
The Fed's tightening cycle in 2022, the Russia-Ukraine event window in February 2022, and the 2024 election-year volatility spikes are recent examples where the model fell behind regime. We did not push emergency hotfixes during these — the design philosophy is to let the weekly retrain catch up rather than intervene reactively. This is why we cap the published drawdown at 20% (worst-case observed across stress periods); users who can't tolerate that should size down or pick a non-ML system.
Does GoldStrike AI come with a money-back guarantee?
Yes — GoldStrike AI includes a money-back guarantee. The window depends on the licence tier (up to 30 days on the top tier; see the current terms on the product page). The refund covers the licence fee; any trading gains or losses during the trial period remain in your account (we don't reverse trading P&L).
The refund window — up to 30 days on the top tier — is intentional: it covers one full weekly retrain cycle plus typical regime variation. We've found this is enough for buyers to make an informed continue/refund decision. Refund requests via the support portal are processed within typical Stripe / wire timing (about 5 business days). The annual subscription model for ongoing model retraining starts after the initial 12-month included period.
Try GoldStrike AI
Run it on a free demo account first, then buy the licence when you’re ready.
Patience is the whole game — up about 170% in under 4 months on gold
Some days GoldStrike just sits there. Zero trades. The regime classifier doesn't like what it sees, so it waits. First month that drove me a little crazy — I kept checking the journal expecting entries that never came. Now I understand: the quiet stretches are the system working correctly, not failing. When it does fire, the rhythm is clean — usually 2 to 3 trades during London or NY hours, in and out in under 90 minutes most of the time. Up roughly 170% since starting, win rate in the mid-60s, drawdown stayed under 10% even through a brutal July swing. The setup side takes honest effort — Cloud-SET config isn't something you click through in 10 minutes, and that's probably the one area where I'd want clearer docs. Support filled the gaps when I asked, which helped. The 10-trade winning streak I'm sitting on right now feels good, but I remember July 14 — best and worst single-day trades of the month landed on the same day. That's gold for you.
Broker: exness· Used 3-6 months
Kofi67Verified live account
· Jul 2026
Recovered from a $127 drawdown and just hit a fresh high — the rough weeks taught me more
The stretch that tested me most was a two-week run where gold was whipsawing around macro data releases. GoldStrike went quiet during the sharpest spikes — news filter doing its job — but when it did trade, two of those went against me hard. Watching the account slide toward a $127 drawdown was uncomfortable, genuinely. What held me together was that the stops behaved exactly as expected. No runaway losses, nothing catastrophic. Each losing trade closed cleanly and the bot just... waited for the next clean regime.
Recovery wasn't dramatic. It ground back trade by trade, mostly during London and early NY hours on XAUUSD. Win rate has settled in the low 60s across north of 400 trades, which is honest — not inflated. Total return is sitting around 170% from where I started, and I'm at a fresh equity high right now after clawing all of that back and then some.
One real caveat: the Cloud-SET and profile options took longer to feel comfortable with than I expected. I moved from the default to Balanced after a flat stretch, and there was genuine uncertainty about whether I was configuring it sensibly. Support cleared that up quickly — that's where they actually earn the top marks.
Patience is the skill this system asks for most.
Broker: admiralmarkets· Used 3-6 months
Laura38Verified live account
· Jul 2026
Docs took patience — but the logic holds up and returns are real on XAUUSD
Took me a few weeks to genuinely understand what was running. The documentation covers the regime-switching and the event guard, but connecting those explanations to what I was actually seeing in live trades took longer than I expected. The spread quoted in cents instead of pips tripped me up early — small thing, but it made reading fills confusing at first. Once I accepted that the system needs a few bars of confirmation before it commits to a direction, the logic started making sense in practice.
Win rate is sitting in the low 70s across roughly 500 closed trades, and drawdown hasn't cracked 4.5% even through the week gold whipped around a news event. That week it quietly halved size on its own — I only noticed when I checked position sizing afterward. Support filled in the gaps I couldn't find in the docs, which is the main reason ease-of-use isn't a frustration anymore. Account is up several hundred percent from where I started, which I still find slightly hard to look at without double-checking the math. The learning curve is real, but once the fog cleared it's been steady.
Broker: admiralmarkets· Used 3-6 months
Patrick22Verified live account
· Jul 2026
Support answered fast — results back it up, setup side still has a learning curve
Asked support about adjusting the ML confidence floor — specifically whether nudging it higher would cost too many valid regimes on XAUUSD or just filter noise. Got a clear, technical reply within a few hours. Not boilerplate. They actually addressed the tradeoff. That interaction alone bumped my confidence in the product.
About 85% up from starting balance across roughly 3.5 months, drawdown stayed under 7% even through the ugly week in early July. The event guard blocking entries around major releases has saved me more than once.
The one honest friction: Cloud-SET config is not obvious. Getting Balanced dialed in took me longer than it should have, and the documentation assumes you already know what you're tuning. Worth the effort — but expect a few hours of reading before it clicks.
Broker: pepperstone· Used 3-6 months
OmarVerified live account
· Jul 2026
Came from manual EURUSD work — switching to a gold bot changed what my mornings look like
Before this I was manually trading EURUSD and GBPUSD intraday. Edge was inconsistent, discipline was the bottleneck. Moving to an automated XAUUSD system meant giving up some control, which took adjustment. What concretely changed: I stopped second-guessing entries at 6am and started reviewing journals instead. Account is up roughly 90% across about four months, win rate in the low 70s. Not magic — there have been rough patches, including a stretch this week with three losing trades back to back. The setup side (Cloud-SET, getting the right profile dialled in) cost me time early on. Docs aren't always clear on the why behind certain parameters, just the what. That's where ease-of-use drops a point for me. The event guard blocking entries around FOMC and NFP is the piece I trust most — it's saved me from the kind of news chaos that used to wreck my manual book. Max drawdown has stayed manageable even through volatile weeks where size automatically pulls back. Support has been responsive when I've had config questions. Value relative to what I'd spend on signals or data feeds is not close.
