Curve Fitting

Curve fitting is the process of designing a financial trading system that is rigorously optimized for an exclusive set of historical data of a given asset.

This is done so with the primary purpose of displaying a profit for that specific period.

Traders, when developing a trading strategy, rely on two testing grounds, back-testing, and forward-testing.

Back-testing is taking a set of trading rules, and applying them to the past, to judge what would have been the results had these rules been implemented then.

Forward-testing is taking these same rules, and applying them in real time.

Both are equally as important, since if a system doesn’t show any promise on historical data, it’s fruitless to continue in real time.

If real-time testing shows constant losses despite successful back-testing, again, something’s not right.

Understanding Curve Fitting

Technically speaking, curve fitting is a subset of back testing, but with a twisted methodology.

Traders seek to employ classical trading techniques with their own twist to see how their system performs when back testing.

On the contrary, curve fitters aren’t concerned about real trading with technical analysis when back testing.

Rather their interest lies in adjusting their trading rules to match the exact conditions of the market for exclusively the back-testing period, timeframe, and asset.

For example, a real forex trader who trades the GBP/JPY may genuinely aim to optimize their system.

This is commonly done by making some adjustments in terms of placing wider stops for this volatile currency pair based on a tried and tested tool such as the Average True Range (ATR).

However, a curve fitter will just look at the GBP/JPY over a specific time period, note down its high and low for those days/hours/minutes, and then place exact profit targets within the trading range and stop losses outside the trading range.

Consequently, curve fitted “systems” never perform when traded live. This is because the exact set of data, in terms of volume, spread, range, bulls and bears, support and resistance, oversold and overbought, all in relation to the exact time, etc.… are never repeated.

In recent years there has been a rise in the number of vendors offering their automated trading robots, purely showing curve-fitted back-tested results.

This is performed with the hope of enticing inexperienced traders.

The only way to ensure the robustness of a system is to back-test and forward-test, not curve fit, as bending the rules to fit an exact set of data simply doesn’t work in real time.

Any past results showing accelerating northern equity curves are merely an illusion.