Nasdaq Short-Only EA Tests: How to Read the Evidence

Read a report as a chain of evidence

Begin with identity, then behavior, then performance. Identity means the code version, preset and contract match the advertised product. Behavior means the report’s trades reflect the stated direction, entry and exit rules. Performance is the result produced by that specific combination over those specific dates. Skipping the first two steps makes an attractive profit figure difficult to interpret.

A report should let a reader trace at least several entries from the stated signal to the actual transaction. Include an example winner, an example loser and a trade from a difficult rebound. Showing only the best trade explains very little about the process. If the signal uses completed candles, the example should show the confirmation candle and the subsequent executable entry separately.

For a Nasdaq-focused product, identify the underlying contract rather than assuming a broker symbol is self-explanatory. Contract value, trading hours and financing can differ between similarly named instruments. Store the specification used for the test so later changes do not silently alter the meaning of a preset.

Worked example: two reports with the same net result

Imagine two test reports each ending 20% above their starting deposit. Report A reached a 7% maximum equity drawdown and completed 180 trades. Report B reached 35% drawdown and completed 12 trades. Those invented numbers do not tell us which system is better for every purpose, but they show why final profit alone cannot describe the experience of holding the strategy.

Next suppose most of B’s profit came from one trade. A missed fill, earlier stop or different evaluation period could change its result substantially. The relevant sensitivity test removes or perturbs that event and explains what happens. For A, investigate whether many similar trades are really independent or merely a series of entries during one favorable market regime.

A high trade count does not cure overfitting. A larger sample can still be dominated by one instrument, one session or one directional period. Publish the distribution of results by regime and time, and explain whether the chosen parameters were selected after inspecting those same trades.

Publication standards for crash performance claims

Use an exact claim such as “this version, on this broker symbol, over these dates, produced this simulated result.” The more general statement “it makes money in crashes” requires evidence across multiple relevant episodes and adverse control periods. Even then, future execution remains uncertain.

Show losses beside profits. If bullish periods are unfavorable, explain how the mandate handles them and whether the filter was tested without hindsight. If the product adds to losing positions, disclose that behavior before a reader sees the balance curve. A simple label such as trend-following is insufficient when exposure scaling drives the account risk.

Finally, preserve the original evidence rather than recreating a balance line from an image. A plotted performance chart should be traceable to actual report data. Until that packet exists for a new SELL-only robot, the collection can explain the methodology without implying a result has already been achieved.