When evaluating any trading system, hold it to these standards:
- It is profitable in a wide variety of market groups
- It is not curve-fit or over-optimized.
- It is profitable across a range of parameters.
- Its logic and rules must be fully disclosed with no black box aspects.
These four tests are not difficult to state. They are very difficult for most systems to pass. A system that works only in the market it was designed for fails test one. A system whose rules were fitted to historical data to produce the best backtest results fails test two. A system that works only when parameters are set to a precise value fails test three. A system whose entry and exit logic cannot be fully explained fails test four. Trend following passes all four, and the reasoning behind each test is worth understanding in detail.
More on Curve Fitting
- Trend following is not a curve-fit. Curve-fit systems customize the trading rules differently for each market you trade, producing unrealistic results. Trend following rules are the same for each market.
- Computer technology can be easily used to over-optimize a trading system and produce something that looks good. By testing thousands of possibilities, you could create a system that works. However, trying to produce a magical or perfect system falls apart in the real world.
- Trend following parameters or rules work across a range of values. System parameters that work over a range of values are robust. If the parameters of a system are slightly changed and the performance adjusts drastically, beware. For example, if a system works great at 20, but does not work at 19 or 21 you have a system with poor robustness. On the other hand, if your system parameter is 50 and it also works at 40 or 60, your system is much more robust (and reliable).
- Traders often only focus on future profits when looking at a system. The key, however, is risk control (or money management). If you control your risk and let your profits run, you position yourself to make bigger money throughout the long term.
- A good system with robust and adaptive parameters must not require re-optimization. Trend following uses indicators and parameters that adapt to changing market conditions.
Why These Five Points Matter in Practice
The first point, that trend following rules are the same for each market, is the most direct answer to the curve-fitting critique. A curve-fit system finds the parameters that produced the best results in each specific market’s historical data. It does not discover a general principle. It discovers the particular setting that happened to work in the past for that market, which is essentially memorizing the data rather than understanding it. When the future arrives and the market behaves differently from its historical pattern, the curve-fit system fails. Trend following avoids this by applying the same rules everywhere. The same breakout logic, the same volatility-based position sizing, the same exit criteria. If the rules work in one market, they work because they reflect genuine market behavior, not because they were tuned to one data set. For the specific rules and their cross-market application, see the TurtleTrader rules.
The second point about computer over-optimization is the practical danger. Given enough computing power and enough historical data, you can find parameter settings that produce an extraordinary backtest. The process is called data mining. The result looks like a brilliant system. The real-world performance is typically poor because the “discovery” is an artifact of the search process rather than a genuine edge. A trend following system tested across many markets with consistent parameters does not have this problem because the parameters are not tuned to any specific dataset. They reflect the underlying logic of the approach.
The third point about robustness at a range of values is the practical test. Take a system’s key parameter and change it by 10 to 20% in both directions. If performance collapses, the parameter is curve-fit. If performance remains reasonably consistent, the parameter is robust. The TurtleTraders used 20-day and 55-day breakout periods not because those exact numbers were optimal for some historical dataset, but because breakout-based entry works across a range of lookback periods. The edge is in the concept, not the precise number. A system that only works at exactly 20 days is not a system. It is a coincidence captured in code.
The fourth point brings risk management back to the center. The most beautifully constructed entry system is worth little without correct position sizing and risk control. A system that finds brilliant entries but allows uncapped losses will eventually be destroyed by a single position that moves violently against it. Trend following’s approach to risk is built into the system as deeply as the entry rules. Position sizing based on current volatility, maximum risk per trade as a defined percentage of equity, and stop exits at predefined levels ensure that no single trade can inflict catastrophic damage regardless of how the market moves.
The fifth point completes the picture. A system that requires constant re-optimization is a system that was curve-fit from the start. Every time the market changes and performance deteriorates, the optimizer runs again and finds new parameters that fit the new data. The cycle repeats. Each optimization looks backwards at what would have worked, producing another curve-fit that will fail forward. Trend following parameters adapt to current market conditions through the volatility measurement built into the system, not through re-optimization. The rules are recalibrated continuously by current price action, not by periodic searches for better parameters. For the full story of how these principles were tested in one of the most rigorous trading experiments ever conducted, see the TurtleTrader story and the broader trend following framework.
Frequently Asked Questions
What is curve fitting in trading systems?
Curve fitting is the process of customizing a trading system’s rules or parameters to produce the best possible results on historical data. The result looks like a powerful system in backtesting but fails in live trading because the rules are memorizing the past rather than reflecting a genuine market principle. A curve-fit system works on the data it was tested on and fails on new data.
How do you test whether a trading system is curve-fit?
Apply the four standards: test it across multiple market groups, verify the parameters work across a range of values rather than only at a precise setting, check whether slightly changing a parameter causes performance to collapse, and ensure the logic is fully transparent. A robust system remains profitable when its parameters are varied by 10 to 20% in either direction. A curve-fit system does not.
Why does trend following use the same rules for every market?
Because the rules reflect a general principle of market behavior, that prices trend, rather than rules optimized for a specific market’s historical data. Applying the same entry and exit logic across crude oil, currencies, bonds, and equity indices simultaneously tests whether the approach captures something real about markets rather than something coincidental about one data set.
Why is re-optimization a red flag for a trading system?
Because it means the system was curve-fit to begin with and needs periodic re-fitting to keep producing acceptable results. Each re-optimization produces new curve-fitting that will eventually fail. A genuinely robust system does not require re-optimization because its parameters reflect real market behavior that does not change from year to year.
Why is risk control more important than optimizing entry signals?
Because a system with mediocre entries but excellent risk control will survive and compound over time, while a system with brilliant entries but no risk control will eventually be destroyed by a single large loss. The entry gets you in. Risk control determines whether you are still trading five years later. Letting profits run while controlling losses is the mathematical foundation of trend following’s long-term edge.
Trend Following Systems
Want to learn more and start trading trend following systems? Start here.
