Understand MT5 equity and balance drawdown, profit factor, expectancy and losing streaks, with worked examples and a transparent EA #001 report case study.
The largest number in a backtest report is rarely the most useful one. Net profit tells you where an experiment ended. It does not tell you how much risk accumulated on the way, whether a few trades carried the result or whether the settings were selected after a large search.
Read a report in layers: test conditions, exposure and drawdown, trade distribution, then profitability. This guide uses worked examples and a supplied EA #001 historical report to show how those layers fit together. Arithmetic examples are educational; they are not projected trading outcomes.
Read the test identity first
Before interpreting a ratio, record the EA version, symbol, broker, dates, deposit, account currency, leverage, modeling and input preset. Determine whether the report is a baseline, an optimization winner or a test on untouched data. The same profit figure means different things under those conditions.
Check the test duration and trade count together. A one-year test is not a two-year test, even if the manual describes a two-year comparison standard. Likewise, Total Deals is not interchangeable with Total Trades. A position can involve multiple executions, including partial entries or exits.
Identify costs in the ledger. Missing commission or swap assumptions can materially affect a small average trade. An incomplete record should lead to an explicit uncertainty, rather than a favorable default assumption.
Balance and equity answer different questions
Balance primarily tracks realized account changes. Equity includes the current effect of open positions. A strategy may keep its balance curve smooth by postponing realization of losing positions while its equity experiences a substantial drawdown.
Compare the two curves and inspect open exposure at the worst equity point. Ask whether positions recovered because the original hypothesis played out, because exposure increased, or because the test happened to end after recovery. Ending a test while a basket is still open can also affect what a casual screenshot appears to show.
Maximum monetary and relative drawdown can differ
For a peak of 10,000 and a subsequent trough of 8,000, the decline is 2,000, or 20% of that peak. Later, a decline from 25,000 to 21,000 is 4,000, or 16%. The second episode has the larger money loss; the first has the larger percentage decline.
MT5 reports maximal drawdown in account currency and relative drawdown as the largest peak-to-trough percentage. Their paired values can therefore describe different episodes. Absolute drawdown is a different measurement: the decline below the initial deposit. MetaQuotes defines these fields separately. [1]
Do not use “absolute” as a casual synonym for the largest monetary peak-to-trough drop. On a product page, label the exact report field and currency. If a summary gives a smaller percentage beside the largest cash drawdown, also inspect the relative-drawdown field before stating the worst percentage risk.
A worked EA #001 example
The supplied US100.pro report covers 1 October 2025 to 1 October 2026, with USD 5,000 initial capital, 1:100 leverage and a 3% risk input. It reports 275 trades and USD 11,243.74 net profit. These are historical results for one supplied configuration, not a verified live record or an independently reproduced result.
| Report field | Supplied value | What to notice |
|---|---|---|
| Profit factor | 1.41 | Read alongside drawdown and costs |
| Winning trades | 37.82% | Most trades were not winners |
| Maximum relative equity drawdown | 41.23% (USD 2,651.04) | Largest percentage equity decline |
| Maximal equity drawdown | USD 3,276.55 (28.49%) | Largest monetary decline, another episode |
| Maximum relative balance drawdown | 38.17% | Does not replace equity drawdown |
| Maximum consecutive losses | 19 | A long losing sequence occurred |
The lesson is about reading, not endorsing the configuration. Reporting only 28.49% would omit the larger relative equity decline. The supplied trade ledger records zero commission; that cost observation must travel with the result. The original report is linked in the source materials below. [2]
Profit factor is useful, but not a certificate
Profit factor divides gross profit by the magnitude of gross loss. If profitable trades total 14,100 and losing trades total 10,000, the factor is 1.41. It describes that sample. It does not say that future profits will exceed losses by the same ratio.
Inspect sensitivity to the largest winners and the selection process. A high factor from a handful of trades is fragile evidence. A high factor selected from thousands of configurations can reflect search luck. Compare factors only after aligning costs, test windows and meaningful risk assumptions.
Avoid treating any fixed threshold as universally acceptable. A ratio above one can coexist with an intolerable drawdown, heavy concentration or inadequate evidence. Use several measurements to decide which question to investigate next.
Win rate needs average win and average loss
Consider an illustrative system that wins 40% of trades, earns 2 units on a winner and loses 1 unit on a loser. Its average outcome is 0.40 × 2 − 0.60 × 1 = 0.20 units before costs. A second system that wins 80%, earns 0.25 and loses 2 has 0.80 × 0.25 − 0.20 × 2 = −0.20 units.
The arithmetic shows why win rate cannot stand alone. In your actual report, calculate from net trade outcomes consistently, including costs without subtracting them twice. If trades have different sizes, also study exposure-normalized outcomes; money totals can overweight later, larger positions.
Expected payoff summarizes average historical outcome per trade in the report. It is not a reliable payout promise for the next trade. Ask how spread, commission and a few unusual trades contribute to that average.
Read streaks, concentration and recovery time
Maximum consecutive losses describes what happened in the sample, not a future upper bound. A 19-loss streak makes it reasonable to investigate clustered losses, sizing behavior and recovery duration. Future conditions may produce a worse sequence.
Break results into months or quarters, long and short trades, and market regimes relevant to the strategy. Ask whether one exceptional interval explains most profit. Measure time under water, not only the size of the deepest drop: a shallower drawdown that persists for months can create different practical pressure from a brief decline.
Loss recovery is asymmetric
After a 20% loss, recovering to the original capital requires a 25% gain on what remains. The general arithmetic is loss fraction ÷ (1 − loss fraction). A 40% decline needs approximately 66.7%; a 50% decline needs 100%.
This is capital arithmetic, not a prediction of recovery. It explains why increasing risk to repair a drawdown changes the experiment and can compound exposure. Document any change in sizing instead of presenting the combined track as an unchanged strategy.
Your report-reading checklist
- Confirm exact test identity and how settings were selected.
- Compare balance with equity and inspect floating exposure.
- Read maximal, relative and absolute drawdown separately.
- Pair win rate with average win, average loss and costs.
- Inspect trade count, concentration and losing sequences.
- Record what remains unverified and what would invalidate the conclusion.
Frequently asked questions
Which drawdown should I report for an EA?
Name the field precisely. For trading exposure, relative equity drawdown is especially important because it includes open-position effects. Provide the monetary measurement as well; do not merge the two episodes into one invented statistic.
Is a profit factor of 1.41 good?
It means gross profits exceeded gross losses in that sample. Its usefulness depends on sample size, costs, drawdown, selection and performance on unseen data. It is not enough by itself to approve live use.
Does 100% history quality prove a backtest is realistic?
No. Data quality, modeling choice and live execution realism are separate issues. Inspect the data and journal, then investigate costs and execution assumptions independently.
Sources and further reading
Original educational guide by MT5 Robot Lab. Platform details are checked against the linked documentation; worked examples and diagrams are identified as illustrative. Last updated: .