Automated trading systems are everywhere in 2026, and the gap between a professionally engineered Expert Advisor (EA) and a repackaged martingale gambling script is often invisible until you look at the right numbers. Most buyers evaluate a trading robot the way they'd evaluate a course or an app: they read the sales page, watch a testimonial video, and check the price. That approach gets people burned. This guide walks through the specific, checkable metrics that separate a legitimate rules-based system from a marketing wrapper, with worked numeric examples so you can apply the framework to any product you're considering, not just gold-focused EAs like this one.

Why Metrics Beat Marketing Claims

Every EA vendor can write "consistent profits" or "low risk" on a landing page. Almost none of that language is independently checkable. Metrics are different: a drawdown figure, a trade count, and a verification badge are either present and auditable, or they're absent. That distinction is the entire point of due diligence.

The CFTC's advisory on trading system fraud specifically warns that promoters of automated systems often show cherry-picked results or simulated performance dressed up as real trading. The regulator's guidance is blunt: if a system's track record can't be traced to an actual funded account with a broker, treat the numbers as unverified. This is the single most important filter you can apply before spending any money.

Separately, the FTC's overview of investment scams notes that urgency ("only 20 licenses left today"), guaranteed-return language, and pressure to buy before you can verify claims are consistent hallmarks of scams across asset classes, not just crypto or forex. If a seller's page leans on urgency instead of data, that itself is a metric worth recording.

The Difference Between a Backtest and a Live Result

A backtest tells you how a strategy would have performed historically, using an MQL5-based Expert Advisor or similar tool run against past price data. It's a useful engineering step, but it can be fitted to the data after the fact. A live, verified track record tells you how a strategy actually performed with real order execution, spread, and slippage. Both matter, but only one of them is proof. If you want to understand how backtests are built correctly, our guide on how to backtest an EA on MT5 walks through the process and its limitations in detail.

Metric 1: Verified Track Record (Not Self-Reported Screenshots)

The single highest-leverage check you can do is confirming that performance numbers come from a third-party verification service rather than a screenshot the vendor controls. Services like Myfxbook connect directly to a live trading account via read-only investor access, pull the broker's actual statement data, and publish it without the vendor being able to edit the numbers after the fact. The platform's own verification documentation explains exactly how that read-only connection works and what "verified" actually certifies.

A second independent layer is the MQL5 Signals service, which similarly mirrors a real account's trade history and lets you subscribe to copy trades directly, with the platform itself tracking every fill. When a vendor offers both a Myfxbook-verified account and an MQL5 signal, you have two independent third parties confirming the same underlying trade history, which is meaningfully stronger evidence than either alone.

Concretely, when you land on a verification page, check for: an account number you can look up (not just a vendor-chosen display name), a "real" or "verified" account type tag rather than "demo," a start date old enough to include at least one meaningful market pullback, and a link that lets you inspect individual closed trades, not just an equity curve image.

Golden Viper EA, for example, publishes its live results on a verified Myfxbook account (account 11943038) alongside an MQL5 copy signal, so both the equity curve and the underlying trade list are open to inspection rather than asserted in marketing copy.

Metric 2: Maximum Drawdown Relative to Your Account Size

Drawdown is the peak-to-trough decline in an account's equity, expressed as a percentage. It's arguably the most important single number in this entire evaluation, because it tells you what you would have needed to sit through, emotionally and financially, to stay invested through the strategy's worst stretch. Investopedia's explainer on drawdown is a good primer if the concept is new to you, and our own deep dive on how drawdown works and why it matters covers the mechanics in more depth.

Here's the worked math that most buyers skip: if a system shows a maximum historical drawdown of 20% and you fund it with $10,000, you should expect that at some point your equity could dip to roughly $8,000 before recovering, assuming future drawdowns are in the same range as past ones (which is never guaranteed). If your risk tolerance can't survive watching $2,000 evaporate on paper, that system is not right for your account size, no matter how good its long-term return looks.

Worked Example: Two Robots With the Same Return, Different Risk

Imagine two systems both show a 40% annual return over the past 12 months.

• Robot A achieved 40% with a maximum drawdown of 12%.
• Robot B achieved 40% with a maximum drawdown of 45%.

On the surface, the return line looks identical. But Robot B needed a nearly 82% gain just to recover from its 45% drawdown (a loss of 45% requires a subsequent gain of about 1/(1-0.45) - 1 = 81.8% to get back to breakeven), while Robot A only needed about 13.6% to recover from its 12% dip. Robot A's path is dramatically less punishing to sit through, and it implies a much more conservative position-sizing approach behind the scenes. This is why comparing return numbers alone, without drawdown, is close to meaningless.

Metric 3: Risk-Adjusted Return, Not Just Raw Profit Percentage

Raw return percentages are easy to market and easy to misread. A more honest picture comes from risk-adjusted metrics that weigh the return against how much volatility or drawdown was required to earn it. The core principles of risk management apply directly here: a strategy's quality is a function of the return it generates per unit of risk taken, not the return in isolation.

Two simple ratios you can calculate yourself from any published track record:

• Return-to-drawdown ratio: annual return divided by maximum drawdown. A system returning 30% with an 8% max drawdown has a ratio of 3.75. A system returning 60% with a 40% max drawdown has a ratio of 1.5 — the first system is doing more with less risk, even though its headline number is smaller.
• Average win versus average loss: if a system's average winning trade is $180 and its average losing trade is $90, the reward-to-risk ratio is 2:1, which means it can be profitable even with a win rate below 50%.

This is also where trade selectivity matters. A system that takes one carefully filtered setup per day on a single instrument and a single higher timeframe behaves very differently, statistically, from one that fires dozens of trades across multiple pairs. Selectivity generally means fewer data points but each one carries more signal; over-trading systems can look busy without being better.

Metric 4: Sample Size — How Many Trades Back the Statistics

A win rate or average return calculated from 15 trades tells you almost nothing statistically. A win rate calculated from 400+ trades across multiple market regimes tells you considerably more. This is basic statistical significance, and it's routinely ignored by buyers dazzled by a short, hot streak.

As a rough practitioner's rule of thumb: treat any published statistic based on fewer than 50 trades as very low confidence, essentially noise you shouldn't use to judge edge yet. Treat 50-150 trades as directionally useful but still fragile, and 150-300 trades as a reasonable basis for a cautious decision. Anything above 300 trades, spanning at least two to three different market conditions (a trending phase, a ranging phase, a high-volatility news period), is about as solid as published evidence gets short of running your own live test. If a vendor's marketing leads with "500% return" but the trade history only has 22 entries, that return figure is not statistically meaningful yet, regardless of whether it's verified.

Metric 5: Platform, Broker, and Execution Compatibility

A strategy's live numbers are only reproducible if you can run the exact same setup: platform version, broker execution model, and account type. Check three things before buying.

Platform Support

Confirm whether the robot runs on MetaTrader 4, MetaTrader 5, or both, and whether the license covers one platform or is transferable between them. Some vendors charge separately per platform; others bundle both under a single purchase. Confirm this explicitly before paying, since re-purchasing a second license for the other platform can double your effective cost.

Broker Execution Quality

The same EA can perform differently across brokers because of spread, slippage, and execution speed differences. Wider average spreads on gold quietly erode a strategy's edge, especially for systems with tighter targets. Our comparison of broker spreads on gold and our breakdown of the best brokers for running a gold EA are useful starting points.

Hosting and Uptime

An EA that isn't running when the market moves can't execute its logic. Confirm whether the vendor recommends or requires a VPS, and budget for it separately, since this is a real recurring cost most buyers forget to model. Our VPS setup guide for forex trading covers the practical setup steps.

Metric 6: Total Cost Structure — Not Just the Sticker Price

Compare pricing models carefully, because "cheap" and "expensive" are meaningless without looking at the structure. A one-time license fee has a very different long-term cost profile than a recurring monthly subscription, and a copy-signal service has different implications than a downloadable file you install yourself.

As a concrete example, Golden Viper EA is sold as a one-time $199 lifetime license that covers both MT4 and MT5 under a single purchase, with no subscription and no recurring renewal fee; it's also offered as a separate $30/month MQL5 copy signal for traders who'd rather mirror trades than install the file themselves. Running the math over three years, the lifetime license works out to roughly $5.50 per month amortized, versus $30/month for the signal service, which is the kind of comparison you should run for any product before deciding which format fits your setup. Related to cost, our guide on how much capital you need to start EA trading covers the full budget picture including VPS and broker minimums, not just the license fee.

Metric 7: Strategy Transparency and Risk Controls

You are not entitled to a vendor's exact entry logic — that's the intellectual property they're selling — but you are entitled to know the risk framework wrapped around it. Look for clear answers to these questions before buying:

• What instrument(s) and timeframe does it trade? A system that trades one instrument on one higher timeframe is easier to evaluate and reason about than one that trades everything on every timeframe.
• Does it use martingale, grid, or averaging-down logic to "recover" losing trades? These approaches can produce smooth-looking equity curves right up until a single large loss wipes out months of gains. This is one of the most important questions to ask, and a vendor should answer it directly, not evasively.
• Is position sizing risk-based (calculated from account equity) or fixed-lot (the same size regardless of account balance)? Risk-based sizing scales more sensibly as an account grows or shrinks.
• Are there multiple risk-mode settings (e.g., conservative, normal, aggressive) so you can match the system's aggressiveness to your own tolerance, rather than being locked into one fixed risk level?

Golden Viper EA, for instance, uses risk-based lot sizing with three selectable modes (Conservative, Normal, Aggressive), a profit-lock mechanism on winning trades, and an optional safety stop — and explicitly does not use martingale, grid, or averaging strategies. Whatever product you're evaluating, ask the vendor these same four questions directly and be skeptical of vague or deflecting answers.

Metric 8: Red Flags That Override Every Other Metric

Some claims are disqualifying regardless of how good the surrounding numbers look. Both the CFTC's guidance on forex fraud and the FTC's investment scam resources consistently flag the same handful of patterns.

If a product exhibits two or more of these simultaneously, treat that as a strong signal to walk away, even if the vendor's return figures look attractive.

Putting It Together: A Step-by-Step Checklist

When you're actually evaluating a specific robot, work through this sequence rather than jumping straight to the price:

1. Find the verified track record (Myfxbook, MQL5 signal, or equivalent third party) and confirm the account is marked "real," not demo.
2. Note the maximum drawdown and mentally apply it to your own account size, as in the worked example above.
3. Calculate a rough return-to-drawdown ratio and compare it against alternatives you're considering.
4. Check the trade count and history length; discount anything under roughly 100 trades or three months of live data.
5. Confirm platform compatibility (MT4/MT5), and check broker and VPS requirements and their added cost.
6. Add up the true multi-year cost: license fee, any subscription, VPS hosting, and realistic spread costs at your chosen broker.
7. Ask directly about martingale, grid, or averaging logic, and about position-sizing method.
8. Screen for the red flags in the table above — guarantees, pressure tactics, missing drawdown data.

If you'd rather see how this framework applies to a specific product category first, our guide on whether automated gold trading is actually profitable applies these same metrics to real numbers rather than marketing claims.

How This Applies Beyond Your First Purchase

Metrics evaluation isn't a one-time gate you clear before buying; it's an ongoing practice. Once you own a system, keep tracking the same numbers on your own live account, because your results depend on your broker's execution, your risk-mode selection, and market conditions going forward — none of which are identical to the vendor's published history. Our broader framework for capital preservation is a useful next step for anyone running an automated strategy long term.

It's also worth understanding what drives the instrument itself. If the robot trades gold specifically, background context on how economic news and macro events move gold prices will help you interpret why a verified track record looks the way it does during specific stretches, such as a high-volatility news week versus a quiet summer range.

A Short, Honest Risk Disclosure

Trading, including trading through an automated system, carries real risk. Past performance, even when independently verified, does not guarantee future results. Drawdowns can exceed historical maximums, and no legitimate vendor can promise otherwise. Only trade with capital you can afford to lose, size your risk mode to your actual tolerance, and treat every metric in this guide as a filter for reducing uncertainty, not as a promise of profit.

Frequently Asked Questions

What is the single most important metric to check before buying a trading robot?

Verification of the track record. A drawdown or return figure is only meaningful if it comes from a third-party service like Myfxbook or an MQL5 signal connected directly to a real account, rather than a screenshot or chart the vendor controls and could edit.

What counts as an acceptable maximum drawdown?

There's no universal number, but many conservative traders look for a system with a historical maximum drawdown under 20-25%, since recovering from anything deeper requires a disproportionately larger subsequent gain. Your comfort level should also factor in how that percentage translates to real dollars on your account.

How many live trades should a track record include before I trust it?

Treat anything under roughly 100 trades as preliminary. A history of 300 or more trades spanning at least a couple of different market conditions gives you a meaningfully more reliable picture of how the system behaves.

Is a higher win rate always better?

No. Win rate means little without knowing the average size of wins versus losses. A system with a 40% win rate and a 3:1 reward-to-risk ratio can be far more profitable than one with a 70% win rate and a 1:2 reward-to-risk ratio.

Should I be worried if a robot uses martingale or grid trading?

Yes, or at minimum you should understand it fully before buying. These techniques increase position size after losses to average down, which can produce a smooth-looking equity curve for a long time before a single adverse move causes a large drawdown or account blowup. Always ask directly whether a system uses this logic.

What's the difference between a one-time license and a monthly subscription for an EA?

A one-time license is a single upfront payment with no recurring cost, which tends to be cheaper over a multi-year horizon. A subscription lowers the barrier to entry but accumulates cost over time — always calculate the 12-24 month total before comparing the two models directly.

Can backtest results alone tell me if a trading robot is good?

Not on their own. Backtests are useful for engineering and refinement, but they can be fitted to historical data after the fact. Live, verified trading results carry far more weight because they reflect real execution, spread, and slippage conditions.

Do I need a VPS to run a trading robot?

Most vendors recommend or effectively require one, since an EA needs to be running continuously to catch its setups, and a home computer that sleeps or loses connection can miss trades or leave open positions unmanaged. Budget for VPS hosting as a separate recurring cost when comparing total pricing.

What should I do if a seller guarantees profits or claims their system is risk-free?

Treat it as a serious red flag and stop evaluating that vendor. Regulators including the CFTC and FTC specifically warn that guaranteed-return language is a hallmark of investment fraud, since no legitimate trading strategy, automated or manual, can promise outcomes.

Does platform choice (MT4 vs. MT5) affect which metrics matter?

The core metrics — verification, drawdown, sample size, cost, and risk controls — matter identically on both platforms. What changes is execution mechanics and available broker options, so confirm the specific platform compatibility and whether one license covers both before you buy.