Strategy performances decline over time?

[drmark27]

I think you should expect degradation over time, but that doesn't mean the system can't make money. If you find that backtested, profitable strategies regularly fail to make money in live trading, then I would suspect your process. One thing I have seen clearly over the years is that backtesting is not backtesting. Much of what I have seen called backtesting is nothing less than an insult to the whole endeavor.

I have traded historical-data-backtested strategies since 1999, in stocks, futures, and funds. Once I have discovered and implemented a profitable trading strategy, its performances always got worse and worse over time. None of my strategies remained profitable after 10 years.

I am curious whether other strategy traders have seen the same pattern? And how your guys manage this issue?

For me, I generally start to reduce my position in a strategy if it is unprofitable for a year. If the strategy remains unprofitable for another year or two, I stop trading it altogether. [My strategies usually generate dozens of trades a year; so a couple of years give me enough sample to suspect that the world has changed.]

 

[qlai]

One thing I have seen clearly over the years is that backtesting is not backtesting. Much of what I have seen called backtesting is nothing less than an insult to the whole endeavor.

Would you mind expanding on this please?

 

[comagnum]

There are definitely differences in the markets today compered to the 90's. In the post 2005 algo/HFT era, price action has changed.

Stanley Druckenmiller is considered to have the best multi-decade hedge fund performance, he shares his insights/observations on how price action.T.A. have changed in this interview.

https://asymmetryobservations.com/2...n-his-use-of-technical-analysis-and-instinct/

 

[drmark27]

Would you mind expanding on this please?

I was referring to things like failing to account for transaction fees, data mining, future leaks, overfitting, and small sample sizes.

 

[tommcginnis]

I was referring to things like failing to account for transaction fees, data mining, future leaks, overfitting, and small sample sizes.

Errrrr, the only thing in that list that is [mildly, possibly, not-definitively] pertinent to backtesting would be transaction fees. Issues with data mining, future leaks, over-fitting, and small samples -- these are modeling issues: the horse to the backtesting cart.
They're important, for sure -- but are settled or not before you get to the backtesting.

 

[drmark27]

In my world that's just semantics, but I won't argue with you.

Errrrr, the only thing in that list that is [mildly, possibly, not-definitively] pertinent to backtesting would be transaction fees. Issues with data mining, future leaks, over-fitting, and small samples -- these are modeling issues: the horse to the backtesting cart.
They're important, for sure -- but are settled or not before you get to the backtesting.

 

[tommcginnis]

In my world that's just semantics, but I won't argue with you.

You criticized others for sloppy language, but that doesn't apply to you?

If you were the manager of such a person, who would you hire?

 

[IamaMars]

This is a very important observation that every trader should take into consideration. Change is inevitable, and even strategies do change. Using old-age trading strategies in the current forex market is a sure way of failing, if you don't make the necessary adjustments to it. The same applies to the future, since strategies that work now may not work then. A trader should always test and prove a strategy's winning potential on demo before going live with it

 

[maler]

It's normal. It would be arrogant to believe no one else can find
the juicy bits you are feasting on. In time, someone that tolerates
more pain for less reward will come along to eat your lunch.
Time to search for another morsel?

 

[gentlerich]

One can refer to strategic performance reports which allows a trader to review a trading system or live trading results so that they can evaluate these trading systems. Anyways you can maintain it ,if the key metrics are considered which are :

• Total net profit
• Profit factor
• Percent profitable
• Average trade net profit
• Maximum draw down

However, Back testing method can be used to measure the performance of trading system.