MT4/MT5 Report Analysis: Understand Metrics and AI Risk
MT4/MT5 report analysis turns an exported trading report into readable metrics and a contextual AI risk review. RBOT Evidence Lab parses the submitted HTML or HTM file, displays supported calculations, and separates reported results from unknowns. It does not confirm account ownership, rerun the strategy on independent market data, or predict future profit.
TL;DR: What can MT4/MT5 report analysis tell you?
MT4/MT5 report analysis can describe the submitted record and identify questions to investigate; it cannot authenticate the trading account or certify a strategy.
| Question | What RBOT provides | Important boundary |
|---|---|---|
| How did the submitted report perform? | Return, profit factor, drawdown and trade statistics when supported | Imported fields and parser coverage determine what can be calculated |
| What risks deserve attention? | An AI review, score rationale and missing-evidence notes | The 0–100 score is not a probability of profit or survival |
| Can I publish the review? | Private sharing or an eligible public report with an embeddable widget | A private link is accessible to anyone who holds it |
How are return, profit factor and drawdown calculated?
Return and profit factor compare reported profit with capital and losses; drawdown describes a decline from a historical peak and must be read with its balance or equity basis.
Three definitions you can use when reading a report
Report analysis is the interpretation of submitted trading records; independent strategy validation additionally requires a separate, documented test or account-verification process.
Profit factor is gross profit divided by the absolute value of gross loss over the same reporting period.
Maximum relative drawdown is the largest percentage decline from a preceding peak in the specified balance or equity series.
These profit-factor and drawdown definitions follow MetaTrader 5's official testing-report documentation. RBOT prefers available reported drawdown fields and may fall back to reconstructed closed-trade values; a closed-trade curve cannot reveal every intratrade equity low.
| Metric | Calculation or reading rule | Boundary |
|---|---|---|
| Return (%) | Net profit ÷ capital base × 100 | The parser uses initial capital, or a deposit-derived base when needed; cash flows require separate interpretation |
| Profit factor (PF) | Gross profit ÷ absolute gross loss | With no gross loss, PF is undefined; a displayed fallback zero is not evidence of a losing strategy |
| Win rate (%) | Profitable closed trades ÷ closed trades × 100 | Flat trades affect the denominator; high win rate alone does not establish positive expectancy |
| Average trade result | Net profit ÷ closed trades | Costs count only where the export and parsing expose them |
| Relative drawdown (%) | Peak-to-subsequent-trough decline ÷ peak × 100 | Balance drawdown and equity drawdown describe different series |
For example, gross profit of 1,500 and gross loss of 1,000 give PF = 1.5; net profit of 500 on initial capital of 10,000 gives a 5% simple return without cash flows. This is a worked example, not a result achieved by RBOT or an EA.
How do backtests, demo accounts and live histories differ?
Backtests model historical execution, demo histories record simulated-account activity, and live histories report real-account activity; none of these uploaded files alone proves ownership or completeness.
| Dimension | Strategy Tester backtest | Demo-account history | Live-account history |
|---|---|---|---|
| Data origin | A historical simulation | A broker's simulated account | An export described as a real account |
| Execution and costs | Depend on tester settings and source data | Depend on the demo environment | Reflect recorded activity, subject to export coverage |
| What it can support | Behavior under documented assumptions | Forward observations in a demo environment | Analysis of the submitted account record |
| What remains unknown | Future execution and out-of-sample robustness | Live fills, capital risk and ownership | Authenticity, ownership and omitted periods |
| Useful next evidence | Held-out tests with stated dates and costs | A longer, documented forward period | Account verification and reconciliation with source statements |
Record the symbol, timeframe, broker, dates, capital base, costs and closed-trade count before comparing two strategies. A PF from one window is not directly comparable with another window when costs, cash flows or market conditions differ.
What does the 0–100 AI risk score mean?
The 0–100 score is an AI-assisted assessment of the submitted evidence under RBOT's review rules, not a calibrated probability or an investment rating.
The interface uses three bands: 0–39, 40–69 and 70–100. These are presentation thresholds, not statistical confidence intervals. Read the conclusion, rationale, evidence-confidence notes, unknowns and next actions together. The evidence-confidence score is also a review assessment, not broker authentication.
Identical input can produce different AI wording or scores when model behavior, prompts or available evidence change. Comparisons require matching report type, period and review conditions; RBOT does not claim deterministic scoring across models or releases.
What happens when a report has many trades?
The parser computes supported summary statistics before limiting the rows passed to analysis. The AI request contains at most 5,000 trade rows and may contain fewer to stay within the request-size limit. Anonymous browser submissions have concentration recalculated on the server from the rows actually submitted; those results describe that sample. The authenticated generation route parses the original report on the server and can use full parsed concentration statistics. Neither route establishes that the original file is genuine.
How do I upload, analyze and publish a report?
Export an MT4/MT5 HTML or HTM report, import it into RBOT, inspect the parsing results, and run the AI review only when you want the extracted evidence sent for analysis.
Step 1: Inspect the browser-parsed record
Use the report analyzer. The upload limit is 8,000,000 bytes; supported text encodings include UTF-8, UTF-16LE and UTF-16BE. Check warnings and missing fields before relying on a number. The sampled growth curve is normalized to 100 and is not a broker's live equity feed. MT4 users can consult the official tester-results guide when exporting and reading source results.
Step 2: Understand the AI data transfer
Clicking AI risk review sends extracted metrics, trade rows and necessary report text to RBOT's server and DeepSeek. A successful review creates a server-held result with a sharing credential valid for one hour. Expiration of the credential does not promise immediate file deletion. Default free access is three successful reviews per IP in total, not three per day; configuration and on-screen availability may change.
Step 3: Choose the sharing scope
A private link can be read by anyone who holds it and is not added to the public catalog. Only a public report offers the public widget. Publication currently requires one record, at least 30 trades, and a conclusion of at least 40 characters; otherwise a public request stays private. These checks control listing eligibility, not statistical reliability. Browse public reviews to see the format. Published share records do not inherit the one-hour credential lifetime.
What evidence should you check before using a review?
Check the source report, measurement definitions and unknowns before treating an AI review as supporting evidence for any decision.
- Reconcile available trades and headline metrics with the original export; save the source file and review date.
- Confirm the reported drawdown's balance/equity basis and whether costs and cash flows are represented.
- Look for sufficient period coverage and separate out-of-sample or forward observations rather than assuming a trade-count threshold validates the strategy.
- Treat missing settings, actual slippage, ownership and future performance as unknown unless separately supported.
RBOT's parser calculations and AI interpretation are different evidence layers. A report page can make them easier to inspect; it cannot repair a false, incomplete or selectively chosen source record.
Where do these definitions and limits come from?
Metric definitions come from official MetaTrader documentation, while product limits and workflow descriptions come from RBOT's current application implementation reviewed on 8 October 2026.
The technical sources are linked beside their claims. This guide documents the parser, browser worker, analysis and sharing workflow; it does not present an independent performance study or claim a human review of every submitted report. Limits may change with configuration and releases. Confirm current limits in the analyzer before uploading.
To start MT4/MT5 report analysis, upload your exported report, check the parsed metrics, and use the AI review to organize questions for further verification.