- Over-optimization means finetuning the parameters to get the best results over the backtesting period and giving us the wrong idea that the system works very well in real-time execution.
- Long testing periods are good, and it is important to identify market regime changes to backtest over relevant periods.
- Asymmetric trading ideas include using different criteria for entry and exit signals.
1. What is so dangerous about over-optimization?
Basically the result of curve fitting, your strategy works flawless with past set of time ranges, but it will not be able to work in different probabilistic scenarios that have not happen yet.
2. How long should a testing period be if you are serious about building a profitable trading strategy?
9 to 11 years of data, so we can run backtest on different market events.
3. Why should you avoid asymmetric trading signals?
To avoid curve-fitting, using the wrong time range of data will end with bad results, Example: using an indicator cross at 20 for long entries but 15 for shorts).
1. What is so dangerous about over-optimization?
That you reoptimize the same optimization. You should optimize a different parameter first before doing so and always change one parameter at a time.
2. How long should a testing period be if you are serious about building a profitable trading strategy?
minimum of 10 years when trading Forex, but that will be difficult if you look at bitcoin.
3. Why should you avoid asymmetric trading signals?
when your code becomes too complicated it will be hard to derive from the code why a strategy isnât working. The key factor described in the article is to keep it simple and asymmetric.
1 Like
-
What is so dangerous about over-optimization?
â Over-optimization can lead to curve-fitting which can be misleading because that can generate
amazing results that will not be achievable in the future. -
How long should a testing period be if you are serious about building a profitable trading strategy?
â The data set used for testing should cover a period of about 5 years and ideally 9 to 11 years. -
Why should you avoid asymmetric trading signals?
â Using asymmetric criteria can add complexity and lead to curve-fiting setups.
1 Like
- Curve fitting. Using past data and creating a program that is fitted to the events of the previous years. Future data is unknown and can be totally different to previous years.
- Ideally, 9-11 years
- Data from previous years will not necessarily be the same as future data. As we donât know what external factors will affect the trends in the market.
1 Like
-
What is so dangerous about over-optimization?
You create an algorithm that is not profitable when a small change occurs to the pattern in the asset you are investing in
. -
How long should a testing period be if you are serious about building a profitable trading strategy?
As long as possible. 10 years would be great, but with crypto we have one coin with 10 years of data. So gathering the max data and testing against a variety of coins will help reduce the chance of curve fitting. -
Why should you avoid asymmetric trading signals?
If your strategy is going to go long and then short the entry exit signals for long/short entries should be similar. If the entries are not symmetrical you can find your self with a curve fitted algo.
1 Like
-
What is so dangerous about over-optimization?
It might lead to curve-fitting; the function might fit perfectly about the past but not at all about the future. -
How long should a testing period be if you are serious about building a profitable trading strategy?
From 9 to 11 years is a period of time enough long to get many market conditions -
Why should you avoid asymmetric trading signals?
Asymmetric adds complexity; any mathematical model might become impossible.
1 Like
Response:
-
By over-optimizing, you allow too broad of a spectrum to be considered for the data and run the risk of curving the results where it will not be as accurate.
-
Ideally, using 9-11 years of data for your testing period is best to see how it performs against the conditions of the past decade.
-
Keep your system symmetric. Not doing so will give your system too much freedom and will provide the user with mixed signals when it comes to buying/selling. Use the same values for your indicators
1 Like
- Over-optimization can be dangerous because it can lead to curve-fitting, which can make a future trading strategy perform poorly because it relies too much on following curves from the past that are very different from the new market conditions.
- Simple strategies should be used to test on data from about 10 years (which we donât have nearly that for most cryptos in 2020, but thatâs what the passage recommends for stocks)
- Using asymmetrical data where your long position strategy is greatly different than your short strategy is another cause of curve-fitting, itâs best to use more symmetrical strategies.
1 Like
- What is so dangerous about over-optimization?
Over-optimization can lead to curve-fitting of the past data that can be misleading because it is no longer a good strategy going forward to future trading strategy. - How long should a testing period be if you are serious about building a profitable trading strategy?
A testing period for profitable trading strategy is between 9 and 11 years. You have a large amount of market conditions that have happened with this time frame. - Why should you avoid asymmetric trading signals?
One should avoid asymmetric trading signals because this can lead to curve fitting. The more variables you use, the more you are prone to curve-fitting.
1 Like
-
The danger of over-optimization lies in the tendency to curve fitting, which can lead to losses as the market dynamics continuously changes thru differing periods. The present market conditions may not mirror (curve fit) your projected outcome.
-
An average testing period of 10 years is an optimal timeframe for building profitable trading strategies. If the strategy during this period is profitable it lowers the threshold of curve fitting, limiting its freedom.
-
Asymmetric trading signals will increase the degree of freedom which will easily allow for curve fitting.
1 Like
1.What is so dangerous about over-optimization?
It can lead to curve-fitting - a given system being unable to work on a different data set. It may have exploited certain market inefficiencies from the past, but canât guarantee the same results in the future. Neural networks are a good example. They are great at optimization but tend to fail in successful trading systems because of excessive data curve-fitting.
2.How long should a testing period be if you are serious about building a profitable trading strategy?
As long as possible, ideally 9-11 years. The longer the trading period, the more statistically significant the data set is and the less likely curve fitting will be, because it takes a wide variety of market conditions into consideration.
3.Why should you avoid asymmetric trading signals?
They automatically increase a strategyâs degree of freedom and makes it more prone to curve-fitted solutions.
1 Like
-
What is so dangerous about over-optimization?
over-optimization is celled also curve fitting; it means that the strategy is optimized only for a certain dataset over a certain period of time - here it works perfectly but it will have problems with new data -
How long should a testing period be if you are serious about building a profitable trading strategy?
it depends on the selected time frame but for a long term trading ideally 9 - 11 years of data for testing to cover as many market situations as possible -
Why should you avoid asymmetric trading signals?
it is better to use symmetric trading signal; asymmetric signals - separate criteria for long and short positions can lead to curve fitting and are not very reliable over long periods of time
1 Like
1. What is so dangerous about over-optimization?
It can lead to âcurve fittingâ, where you try to fit the data to fit your strategy.
2. How long should a testing period be if you are serious about building a profitable trading strategy?
Ideally 10 hours or more for more accuracy.
3. Why should you avoid asymmetric trading signals?
They provide more complexity which will likely lead to curve fitting.
1 Like
I guess you mean years 
Carlos Z
1 Like
- makes it less likely to accurately predict future price.
- 9 - 11 years.
- Adding separate criteria for longs and shorts automatically increases the strategyâs degrees of freedom and makes it excessively prone to curve-fitted solutions
1 Like
- What is so dangerous about over-optimization?
Curve fitting is the risk to avoid when optimizing a system as it will provide a strategy that is more focus on the historic data points that the underlying price movements. Curve fitting will work very well on old data but not on new data.
- How long should a testing period be if you are serious about building a profitable trading strategy?
Several years
- Why should you avoid asymmetric trading signals?
Although trading will always be asymmetric, biases will change through time with macro economics trends : as it is hard to predict what the future will look like, asymmetric trading signals will work better on any circumstances.
1 Like
- What is so dangerous about over-optimization?
Overoptimization can lead to creating a tool that shows excellent results with the preselected data, but show very poor results with other data, including future. - How long should a testing period be if you are serious about building a profitable trading strategy?
Testing data should not be shorter than 30 minutes, but the more recent data the better. Some new instruments that are new may only have a few days, weeks, or months of data. - Why should you avoid asymmetric trading signals?
Asymmetric trading signals tend to lead to over-optimization.
1 Like
-
Over optimization can lead to curve fitting (ex: no predictive capability even though it was able to predict all previous items successfully).
-
Time frames greater than 30 minutes , and optimizations should be carried out for long periods of time, ideally 9-11 years of data should be used for the process.
-
Adding separate criteria for longs and shorts automatically increases the strategyâs degrees of freedom and makes it excessively prone to curve-fitted solutions.
1 Like
-
What is so dangerous about over-optimization?
Curve fitting, twiking so much that it will cor perfect on past data but not realistic for future data. -
How long should a testing period be if you are serious about building a profitable trading strategy?
Ideally 9-11 years. -
Why should you avoid asymmetric trading signals?
It tends to curve fitting.
1 Like