ATR 25 to 32: what does a 69% backtest profit increase mean?
A difference between two historical results starts an investigation; it does not finish one.
The founder’s research record reports net profits of 97,450 for ATR 25 and 164,950 for ATR 32, a relative difference of approximately 69.3%. The larger number alone does not establish a more robust strategy.
Where did these numbers come from?
The two endpoints are from the founder’s iron ore backtest research notes. The chart path is illustrative, not an original monthly equity series. Currency, source data, fees, execution assumptions and complete run configuration are not yet disclosed publicly, so this extract does not support full reproduction or comparable performance claims.
Why are two points insufficient?
Compare neighbouring parameters, not just selected winners. Investigate isolated peaks, while recognising that a flat neighbourhood is not independent validation. Check changes in trade count, holding time and risk exposure before forming a testable explanation.
What should be tested next?
Freeze rules and versions and examine a pre-specified time period. Retain every attempt and apply data and cost assumptions consistently. Publish limitations, unsuccessful cases and unfinished checks alongside results.
| ATR | Recorded net profit |
|---|---|
| 25 | 97,450 |
| 32 | 164,950 |
Questions
Is 69.3% an account return?
No. It is the relative difference between two historical net-profit endpoints.
Can the illustrative curve be reproduced?
No. Reproduction requires the original series, data version, rules, costs and execution configuration.
Sources and scope
- Founder’s owner-supplied research extract. Complete raw data is not published.
- scikit-learn · Avoiding data leakage
External sources support the definitions and method discussion; they do not validate WeiMind’s research results.
For education and research only. No investment advice or promised returns. Backtests do not predict future performance. Any live demonstrations are historical records delayed by at least one trading day.