Doing opposite of what the model says

[fordewind]

Say you've written a strategy. During back-testing, you find out that the strategy is losing money on most of the positions it enters into. Is it viable to use the strategy but instead do the exact opposite of what the strategy tells you in terms of positioning?

I understand that such a behavior requires further investigation, but if we ignore that, can you rely on such a strategy?

Depends on how much data you`ve tested it with.Test it a little more and it`ll revert to positive i`m 100% confident.

 

[qxr1011]

Many roads lead to Rome...

From where each of us is there is only one road that leads to Rome (best case scenario)

 

[Simples]

Technically a losing position can always win if reversed (even discounting spread, commisions and costs). It requires never realizing the loss (ie. no stopp-loss) and is based on hindsight of the current trading plan already being a loser.

In practice, people strive for systems with some form of risk management (aka stop-losses and targets). These will realize paper-losses into real losses (SL="create-loss"), or place limits on profitability per trade. These rules are non-reversible from a practical trading perspective, as they will operate on quite different data (derived from flipping chart upside-down).

Remember, a "losing system", might be tomorrow's winner. Declaring a system "loser" is also a prediction, when not just hindsight analysis.

 

[Moataz Elmasry]

@Moataz Elmasry ... you trade mainly bitcoin correct? Is that what your strategy is trading in?

I admire your investigative skills. But no, this isn't for bitcoin, it's for futures.

 

[JackRab]

I admire your investigative skills. But no, this isn't for bitcoin, it's for futures.

I thought so, because it doesn't seem like a mean-reverting strategy would work on that

 

[Moataz Elmasry]

Technically a losing position can always win if reversed (even discounting spread, commisions and costs). It requires never realizing the loss (ie. no stopp-loss) and is based on hindsight of the current trading plan already being a loser.

In practice, people strive for systems with some form of risk management (aka stop-losses and targets). These will realize paper-losses into real losses (SL="create-loss"), or place limits on profitability per trade. These rules are non-reversible from a practical trading perspective, as they will operate on quite different data (derived from flipping chart upside-down).

Remember, a "losing system", might be tomorrow's winner. Declaring a system "loser" is also a prediction, when not just hindsight analysis.

I agree with you, a strategy with concrete risk management rules is what everyone should aim for.

The system was backtested on 3800 days worth of data, and on 90% of those days, the market behaved the opposite of what the strategy expected. Wouldn't you call this a strong enough signal.

 

[alex314159]

Yeah it works, and as mentioned before the key is consistency. Say your PnL = F(X1...Xn, C) with X your parameters and C your cost. I'd first set C=0 and test different X1...Xn. Take the top performers (making a ton of money) and the bottom performers (losing a ton of money). Retest the former with a punitive C and the latter with a punitive -C, and see where you get the max absolute P&L. Of course overfitting warnings apply!

 

[bpr]

https://www.elitetrader.com/et/thre...-copying-losing-strategy-with-reverse.310160/

 

[Moataz Elmasry]

The system was backtested on 3800 days worth of data, and on 90%

https://www.elitetrader.com/et/thre...-copying-losing-strategy-with-reverse.310160/

Thanks for the link, I can conclude from that post that position parity can be considered another parameter in the strategy and hence reversing positions isn't exactly fading the strategy.