[Canonical HTML article](https://algoway.trade/blog/mt5-optimization-criterion-ranking.html)
Published: 2026-11-07
Modified: 2026-11-07
# **MT5 Optimization Criteria: Why the Best Pass Changes**
The top optimization result depends on what you asked MT5 to reward. The pass with the largest ending balance can lose to another pass on profit factor, and neither ranking tells you how much confidence to place in the sample.
For automated trading, this matters before a parameter set becomes a running strategy. Keep strategy selection separate from execution setup: the [AlgoWay MT5 installation guide](https://algoway.trade/blog/algoway-mt5-expert-advisor-installation-guide.html) addresses the execution side, while a tester ranking describes a historical experiment. A high score does not validate the orders you will send tomorrow.
## Read the objective before reading the winner
MT5 offers objectives including Balance max, Profit Factor max, Expected Payoff max, Drawdown min, Recovery Factor max and Sharpe Ratio max. Balance favors the highest ending balance; profit factor compares gross profit with gross loss; expected payoff measures average profit per trade. These are different questions, even when they appear beside the same parameter values.
The criterion guides candidate selection in genetic optimization. A complete search evaluates the selected grid without using that criterion to decide which combinations survive. Changing how you compare completed results is therefore different from changing the objective that guides a genetic search. See the [MetaQuotes optimization reference](https://www.metatrader5.com/en/terminal/help/algotrading/optimization_types) and our [genetic versus complete search guide](https://algoway.trade/blog/mt5-genetic-vs-complete-optimization.html).
## Three passes, three plausible winners
Consider this invented arithmetic example, not an MT5 test or a recommendation. Each pass starts with 10,000 account-currency units, has no deposits or withdrawals, and reports the following completed-trade results. Gross losses are shown as positive magnitudes.
| Measure | Pass A | Pass B | Pass C |
| --- | --- | --- | --- |
| Gross profit | 6,000 | 3,000 | 1,200 |
| Gross loss | 3,000 | 1,000 | 600 |
| Net profit | 3,000 | 2,000 | 600 |
| Ending balance | 13,000 | 12,000 | 10,600 |
| Trades | 150 | 100 | 10 |
| Profit factor | 2.0 | 3.0 | 2.0 |
| Expected payoff per trade | 20 | 20 | 60 |
Balance max favors A: 10,000 + 3,000 = 13,000. Profit Factor max favors B: 3,000 / 1,000 = 3. Expected Payoff max favors C: 600 / 10 = 60. Nothing about the underlying trades changed. Only the question changed.
C's average comes from ten trades, whereas A's comes from 150. That does not automatically make A reliable or C useless. It means the winning score omits a fact you need to inspect. Likewise, B's higher profit factor does not establish its deepest floating loss. The table contains no equity path, so it cannot answer an equity-drawdown question.
Before comparing real passes, check that dates, starting funds, sizing policy, costs and data settings are comparable. A change in order size can change money-based scores without improving the entry rule. Record those controls alongside the ranking, rather than letting one large number do all the talking.
## A green Result cell is not necessarily a high displayed score
MetaQuotes documents a separate Complex Criterion that combines deal count, drawdown, recovery factor, expected payoff and Sharpe ratio. Its formula is not disclosed; the documented scale runs from 0 to 100. Do not invent weights to reproduce it.
The [MQL5 optimization-criteria chapter](https://www.mql5.com/en/book/automation/tester/tester_criterion) explains that this complex score also controls Result-cell coloring even when another criterion supplies the displayed number. Values below 20 receive red shading and values above 80 dark green shading. A cell's color and its number can therefore describe different assessments of the same pass.
When comparing exported figures with the screen, write down the selected criterion. Treat color as an additional cue, not a substitute for the column definition. Neither a green cell nor a 0-to-100 score is a probability that the strategy will succeed.
## Choose a shortlist with explicit rejection rules
A useful review starts by specifying the conditions that would disqualify a candidate. For example, a research plan might reject passes exceeding a stated equity-drawdown limit, flag very small trade samples for further evidence, and then rank the remaining passes by its chosen objective. Those are review rules you define; selecting a built-in criterion does not silently enforce them.
Keep the winning inputs, the objective and the unselected measures together. If a balance winner has an unacceptable equity path, inspect the [difference between balance and equity drawdown](https://algoway.trade/blog/mt5-balance-equity-drawdown.html) before accepting it. If a result depends on one isolated input value, compare nearby values instead of assuming the peak is stable.
Reserve later data for a separate check using the [MT5 historical forward-testing workflow](https://algoway.trade/blog/mt5-forward-testing-historical-vs-live.html). Choose the selection rule before reviewing that holdout. Repeatedly changing the objective to crown whichever pass looks best on the holdout spends the independence you meant to preserve.
The record worth keeping is specific: which pass won, under which criterion, with which constraints, over which dates. “Best” is a label. Those details make it an experiment someone else can inspect.