Reading Assignment: Common Backtesting Mistakes

[Jekaterina]

1. Over optimization will lead to copying how market acted in the past, thus system will be not efficient in the future market conditions.
   2.According to the article 9-11 years time period is enough to run backtesting.
   3.Doing asymmetric trading signals means that system has been over-perfected according to the chart. AKA curve fitting. Better to keep it niiceee and siimmpleee .

[Jason_Mack]

1. What is so dangerous about over-optimization?

Accidentally curve fitting.

2. How long should a testing period be if you are serious about building a profitable trading strategy?

Optimizations should be carried out for long periods of time, ideally 9-11 years of data should be used for the process in order to ensure that a large amount of market conditions become available. If a simple strategy yields profitable results across a ten year period then the probability of curve fitting is greatly reduced as the system has limited degrees of freedom to artificially “fit” all those different market conditions.

3. Why should you avoid asymmetric trading signal?

Adding separate criteria for longs and shorts automatically increases the strategy’s degrees of freedom and makes it excessively prone to curve-fitted solutions.

[AmonRa]

1. Too much complexity optimizes curve-fitting results, which will only help for past datas. Less parameters and the more simple strategies to avoid curve-fitting results.

2. A testing period of about 10 years of data should be used for the process.

3. Because it will fit past cycles and won’t be necessarily accurate to reflect on the upcoming trades.

[dsutter]

1. Over-optimization is dangerous because the more optimization, the more likely you are curve-fitting to past data and creating a trading strategy that may not perform well in the future.

2. Ideally, the testing period should be in the range of 9-11 years.

3. You should avoid using asymmetric trading signals, because they are likely to become less reliable over time due to changes in macro-economics, plus it is a sign that you may be curve-fitting.

[Leo777]

1. What is so dangerous about over-optimization.?

The dreaded word Curve-fitting,keep the code simple," Elegant Strategies," to avoid complexity.

2. How long should a testing period be if you are serious about building a profitable trading strategy.?

Over a 10 year period of testing, with a simple strategy, if that yields profitable results then the probability of curve-fitting is greatly reduced.

3. Why should you avoid asymmetric trading signals.?

By developing Simple Symmetric strategies, you greatly reduce the chance of curve-fitting.

[Corneile]

1. Curve Fitting, creating strategies based on favourable past performance that won’t be able to produce favourable results in the future.

2. /usually a year or two. but 9 to 11 years for accuracy.

3. Because you add additional degrees of freedom to your strategy which ends up being prone to curve fitting.

[Uyan]

1. Over-optimization leads to curve-fitting, which reduces the predictive power of your financial models and strategies greatly.
   If your model is precisely fitted to a particular data set, how can it adapt to future market conditions?

2. If you are serious about building a profitable trading strategy, a testing period should be at least 30 minutes, and ideally 10 years. A longer testing period can provide more varied market conditions, such as boom-bust cycles and black swan events.

3. You should avoid asymmetric trading signals because they take into account “macro economic variables” that inevitably change throughout history. In other words, they can lead to curve-fitting.

[Alex_Chaz]

1. over optimization means “fine grid optimization” which is more likely to result in curve fitting
2. years, preferably 10 years according to blog post
3. It makes it excessively prone to curve fitting

[GiuZuk]

1 Can produce errors to calculate future data, due to a “copy” of the old past dates, a curve-fitting.
2 for traditional markets, 9, 11 years, but for crypto I would say from birth.
3 it could add data to all the past dates, and make the curve-fitting more evident.

[andreypoe]

1. Over-optimization can lead to curve fitting in your trading strategy.
2. A period of 9-11 years will be ideal.
3. Adding separate criteria increases the strategy’s degrees of freedom and makes it prone to curve-fitting.

[Nadiat_Wasit]

1. Over optimisation can lead to unwanted curve fitting.

2. In the traditional market 9-11 years.

3. Because can lead to a prone curve fitting, because of separate criteria long and short term.

[Keith2]

1. It can lead to curve-fitting. This means that you have based your strategy purely on past data which may not replicate in the future.

2. Between 9 & 11 years so that the strategy takes account for all events which have occurred during this longer period.

3. It can lead to curve fitting.

[tojimmys]

1. What is so dangerous about over-optimization? It can result in “curve-fitting” which means that you have essentially coded past performance rather then creating a predictive formula.
2. How long should a testing period be if you are serious about building a profitable trading strategy? the article suggest 9-11 years.
3. Why should you avoid asymmetric trading signals? This is one of the optimization strategies that can increase the likelihood of curve-fitting.

[barney]

Reading Assignment: Common Backtesting Mistakes.

1. What is so dangerous about over-optimization?
   It may transform a strategy into a “curve-fitting” strategy, that may be perfect for past data, but not future data.

2. How long should a testing period be if you are serious about building a profitable trading strategy?
   Ideally 9-11 years.

3. Why should you avoid asymmetric trading signals?
   Adding separate criteria for longs and shorts automatically increases the strategy’s degrees of freedom and makes it excessively prone to curve-fitted solutions.

[Roman_Mazuryk]

1. we don’t want to generate trading strategies with absolutely astonishing results that will not be achievable going forward

2. ideally 9-11 years, but it is not possible yet for most of the crypto assets. So I can assume that we should take the cryptos with 4+ years history.

3. Adding separate criteria for longs and shorts automatically increases the strategy’s degrees of freedom and makes it excessively prone to curve-fitted solutions.

[Geri18]

1. The purpose is not to generate trading strategies with tremendous results of the past that are useless going forward, but a strategy with the highest possibility to work in the future aswell.
2. Thinking long term, ideally it should be at least 9-11 years.
3. Adding separate criteria for longs and shorts makes the strategy excessively prone to become curve-fitted.

[Dimski45]

1. What is so dangerous about over-optimization?
   The future does not always adhere to the past
2. How long should a testing period be if you are serious about building a profitable trading strategy?
   9 to 11 yrs
3. Why should you avoid asymmetric trading signals?
   It increases potential for curve-fitting

[dudge17]

1. Curve Fitting , it might be over adjusted or tailored to past events that might not be usable in the future due to market conditions or future data sets
2. Test periods should be 9-11 years if serious about building a profitable trading strategy
3. Avoid asymmetric system because the information can lead to adverse selection, incomplete markets and is a type of market failure

[andreaspensold]

1. What is so dangerous about over-optimization?
   With more complexity in the algorithm, the risk of curve fitting increases and results may not fit future events.

2. How long should a testing period be if you are serious about building a profitable trading strategy? Very long time frames with about 9-11 years.

3. Why should you avoid asymmetric trading signals?
   This is a kind of curve fitting and my not fit future trends.

[AndyEken]

1. With over optimization there is a tendency to curve fitting, meaning you potimize based on known data and match your algos to work with that specific data.
2. According to the articel 9-11 years (is that really right). If you take 6 months at the time, back-test that period, then take another period and do the same, that hopefully will do the trick.
3. If the market trends change, the algo might not behave well, and it is easier to curve fit your algos using assymetric criterias.