By Cree Lawson
You run an EA over the dates shown in a report and get a different result. Before changing the settings to make the curves match, I would save your report and compare the two setups.
There are several ordinary reasons for a difference. The useful task is to identify which ones apply and whether they matter to the strategy you are evaluating.
Match the version and configuration
Start with the EA version and the exact preset. A current version can contain changes that weren't present when an older report was produced. Testing the current binary over the older dates doesn't recreate the earlier software.
Then compare symbol names, timeframes, date boundaries, deposit, account currency, leverage and sizing mode. Record the terminal build and the broker or history source. A familiar-looking input screen isn't enough to establish that the configurations match.
Some EAs depend on several symbols or timeframes. Their data requirements need to be included in the comparison too.
Read the execution assumptions
Look at spread, commission, financing, execution delay and contract specifications. A strategy seeking small gains per trade can be particularly sensitive to costs relative to those gains. The question is how sensitive this EA is under the conditions you intend to study.
The tester's tick mode also matters. Broker-provided real ticks and generated ticks are different inputs. Availability and gaps should be checked rather than inferred from a setting's name. The MetaTrader documentation on real and generated ticks explains the distinction.
Find the first meaningful divergence
Instead of comparing only the final profit, inspect where the trade sequences first separate. Did one test enter and the other stay flat? Did both enter but at different prices? Was a position rejected because of margin or volume restrictions?
The testing journal can help identify operational differences. Save relevant messages alongside the report.
Once you have a candidate explanation, change one documented condition at a time and rerun where feasible. Several simultaneous changes can make the result harder to interpret.
Separate reproduction from robustness
Reproducing a report asks whether you can recreate a stated setup and result. Testing another feed or execution assumption asks how the EA behaves when the environment changes. Both are useful, but they deserve separate labels.
A mismatch doesn't by itself establish that either report is dishonest. It may expose an incomplete setup description, an unavailable version or a sensitivity that needs further study.
My goal would be an explanation another person can follow: what matched, what differed, and what happened when the difference was tested. Keep unresolved questions in the record instead of adjusting settings until the final profit looks familiar.
Educational discussion. No named EA or broker is evaluated here. Reproducing a historical result does not establish future profitability.
