NERDQUANT / OUR METHODOLOGY
Our methodology
Define the rules. Preserve the results. Explain the limits.
Each study sets out a specific historical question. Its report is the source for the precise parameters, date range and assumptions used.
1. Specify the experiment
We identify the market, event calendar, entry clock, position-management rules and final cutoff. For timing comparisons, the same selected dates are used wherever the question requires a matched comparison.
2. Keep the complete sample
Profitable and losing periods are included. A study may select events using a stated calendar condition, such as an ECB week followed by FOMC the next week. Results from different date ranges are not treated as proof that the filter itself caused an improvement.
3. Measure more than profit
- Average profit per event: total period profit divided by the number of tested event periods.
- Average holding period: time from the first entry to the final close, in calendar days including weekends. It is not the average life of every individual position.
- Average maximum equity drawdown: the mean of each event period’s deepest floating-equity decline. Dollar and percentage maxima are reported separately where applicable.
- Largest drawdown: the largest observed within-period decline in the tested sample. It is not a maximum possible future loss.
- Win rate: the proportion of tested periods that finished profitable, not a guaranteed probability of future success.
4. Disclose execution and costs
The initial published EURUSD study uses MT5 one-minute OHLC modelling, a 100 ms simulated delay and a $10,000 reset for each event. Its results are price-profit equivalents before swaps and commissions. The simulation does not reconstruct actual news-time fills.
These settings are specific to that study. Always read the specification in each report before comparing results.
5. Separate research from validation
Ideas explored after earlier tests are not untouched out-of-sample evidence. Broader historical testing, realistic costs and forward testing are further steps; none is implied merely because a report is published.
Cumulative totals across independently reset tests are sums of separate results. They are not continuous portfolio equity curves, compounded returns or annualised performance.