Home

Principals

Institutional Trading
Strategy Design

Strategy #1 – "JET"
Bond Intraday
Institutional Only

Strategy #2 "JET" Euro Currency Intraday
Institutional Only

Strategy #3 "SIB
Bond" Daily
Institutional Only

Strategy #4 Trend & Cycle
S&P 500 Daily

Institutional and Corporate Clients FAQ

Free Articles

Contact Us

EIGHT SIMPLE RULES to avoid
over-optimizing a trading strategy

KeyPoint Market Analytics’ trading strategy seeks to uncover general market character.  Experience, along with a few simple guidelines helps us to accomplish this. 

Optimization is the process of determining “optimal values” for the variable inputs that go into a trading strategy. These inputs include indicator parameters, time values, or price functions. Optimization of a trading strategy is necessary to determine precisely what is valid and what is not. Every trading strategy must therefore be optimized to some extent.

The most significant risk of any trading strategy development is “over-optimization”, also referred to as “data mining” or “curve fitting”. The process essentially mines historical data and attempts to identify rules and parameters that fit past data. Over-optimized trading strategies are typically characterized  by at least one of (i) too many rules or conditions (ii) too few trade occurrences per rule (iii) erratic results with small adjustments in input parameters. Over-optimization will result in trading strategies where the historical performance exceeds future performance.  

To avoid the inherent risk of over-optimization, KeyPoint Market Analytics strictly adheres to these 8 simple rules:

1. We cut the number of “and” statements. Experience has taught us what to avoid. We generally use only one “and” statement per entry or exit rule. For example, if “pre condition 1” exists and Momentum “crosses the signal line” then buy.

There are of course reasonable exceptions to using a single “and” statement rule. It may be acceptable to use a second “and” statement depending on the situation. For example, if “pre condition 1” exists and momentum “crosses the signal line” and “close greater then the moving average” then buy.  In this case we have added a long-term trend filter in the form of a moving average with the additional “and” statement.

Over-optimization will occur if additional “and” statements are added. For example, by adding “and close greater than 6 bars ago, and fourteen period RSI less than forty five. These conditions will result in fewer trade occurrences which increasingly move away from a general discovery of market character. 

2. Most of our strategies use only one basic entry principle. We might have several entry rules but they all use the same entry principle. For example rule #1 might be “if pre-condition 1” exists and momentum “crosses the signal line” and “close greater then the moving average” then buy. Rule #2 might be if Bollinger Band Difference was “less than 5 in the previous 3 periods” and momentum “crosses the signal line” then buy. Both rules use precisely the same “momentum crosses signal line” entry, but each supplies different circumstances by which it may be followed. 

This ensures the strategy generates a significant number of total trades and a significant number of trades from each individual signal. In other words, it’s desirable to have a trading strategy that is generating several hundred total trades, but we also want to make sure that no individual trade signal has just a few trades in order to avoid the risk of a curve fit signal.

3. We apply robust input parameters. Any input parameter used in a trading strategy is an optimizable parameter. We’re looking for a wide range of input parameters that are satisfactory. We also look for a gradual falling away on each side of the optimum value. Strategies that display this characteristic are referred  to as “robust”.  If our strategy does not display this characteristic it is discarded.

4. Many of our rules are portable with little variation between  markets. The portability of entry/exit rules between markets reinforces that the rules at work are not curve fit to a particular data set. Therefore, portability between markets is definitely a strong positive. However, experience has taught us that portability is not a necessity as it is also a characteristic of markets that they have their own personalities.

5. Exit rules follow the same principles as entry rules. All evaluations that we apply to entries apply equally to exits. We always test any new trading approach with very simple exits such as dollar stop loss, swing point stops, profit objectives, or a profitable daily open. If any particular approach does not show promise, subjecting it to different exits may lead to over-optimizing.

All exits used in a trading strategy are the same no matter what the entry signal. If entry rules all follow the same general principle, then the same exits should work with all of them.  Mixing and matching exits and entries is very often just another form of over-optimizing.

6. We review  buy and sell trade results separately. This way we can see how each performs within different types of market trends. During strong trends in either direction, we want to see our trading strategy produce positive results or minimize loss when taking trades against the prevailing trend. We’re looking for a trading strategy that  rarely falls out of sync regardless of the trend or lack of - trend.

7. Entries and exits are not optimized based on whether it is a long trade or a short trade. While we concede that markets often behave differently on the way up than they do on the way down, and we definitely could optimize to get better results on this basis, we also believe this approach may be a small step closer to over-optimization. Since our goal is near zero trading strategy degradation after release, we use mirror image signals for either side of the market.

8. We review  the length of test track record several ways. It is clear that the longer the historical test period the more reliable the results are likely to be. Equally important though, is the total number of trades. Generally, more then 500 trades provides a comfort level with the degree if reliability. 
 
We also consider how often long or short a trading strategy has historically held a position in the market. For example, a strategy that holds a position 40% of the time, it is less likely to fall victim to over-optimization than one that holds a position 10% of the time during the same test period. Any strategy that adheres to our guidelines governing entry and exits, and is “in the market” a significant percentage of the time is extremely difficult to over-optimize.