Hello friends.
I hope you are all well.
I developed an interesting strategy based on neural networks and trades a sole index ETF. The NN is trained using volume, direction, volatility indicators, market sentiment and price action, target is the classic 5 days ROC. It buys when NN Indicator is above a cutoff value and rising, it has a dynamic hedge (buying the dip on support zones), and sells when an SL is hit or when NN indicator turns down.
It works very well but it is susceptible of event-driven DDs, such as the covid event during 2020.
¿How could I filter such event-driven DDs? Ideas needed.
Thank you friends, and best wishes!
I hope you are all well.
I developed an interesting strategy based on neural networks and trades a sole index ETF. The NN is trained using volume, direction, volatility indicators, market sentiment and price action, target is the classic 5 days ROC. It buys when NN Indicator is above a cutoff value and rising, it has a dynamic hedge (buying the dip on support zones), and sells when an SL is hit or when NN indicator turns down.
It works very well but it is susceptible of event-driven DDs, such as the covid event during 2020.
¿How could I filter such event-driven DDs? Ideas needed.
Thank you friends, and best wishes!
Rename
QUOTE:
How could I filter such event-driven DDs?
This is indeed an important and deep question.
There is a nice wikipedia article about the term "Rare Events" (https://en.wikipedia.org/wiki/Rare_events) and a field of mathematics/statistics called "Extreme Value Theory" (https://en.wikipedia.org/wiki/Extreme_value_theory) which tries to asses these events.
In the context of the stock market (or more general financial time series) my experience is as follows:
1. The worst events (extreme drawdowns) do not happen often enough to allow any useful statistical analysis. I ususally say you'll need 100 or more data points before you can draw any sound conclusions.
2. Every single of these rare events is different. This makes a meaningful analysis very difficult and any prediction impossible.
Sorry.
... but once you accept the idea that there is no "filter", no way to predict these catastrophic events, there is hope:
There are several techniques to "damp" the effect of these negative rare events.
1. Diversification. Trade more than a single or a few securities. Trade a larger number of non-correlated securities probably from different markets (and several strategies at once). A single rare event which affects just one company will have a much smaller effect on your portfolio of securities (and strategies)
2. Hedge. A huge catastrophic event (like 9/11) will affect most securities and most strategies. Also, back in 2001 exchanges were closed for several days, so no stop-loss order prevented huge losses. A counter-measure for such cases is a "market neutral strategy" i.e. for any long position of $1000 you short $1000 of a broad market ETF. This short position will (roughly) neutralize any concurrent losses in your "long" portfolio.
There are several techniques to "damp" the effect of these negative rare events.
1. Diversification. Trade more than a single or a few securities. Trade a larger number of non-correlated securities probably from different markets (and several strategies at once). A single rare event which affects just one company will have a much smaller effect on your portfolio of securities (and strategies)
2. Hedge. A huge catastrophic event (like 9/11) will affect most securities and most strategies. Also, back in 2001 exchanges were closed for several days, so no stop-loss order prevented huge losses. A counter-measure for such cases is a "market neutral strategy" i.e. for any long position of $1000 you short $1000 of a broad market ETF. This short position will (roughly) neutralize any concurrent losses in your "long" portfolio.
Your Response
Post
Edit Post
Login is required