Topic: General tips

Hey all you fine people.

I just wondered if any of your long-time users of the software had any general tips.

Do you find the strategies to be more reliable in live trading on the longer or shorter timeframes?
Do you find that over-optimising a strategy based on the backtest can actually improve the backtest results but hurt live performance?
Are any particular pairs more reliable when designing strategies using the preset indicators, and are some more prone to wild movements that cannot be predicted?
Do you know this week's lottery numbers?

Thanks all, and happy trading!

Re: General tips

Hi Magic147,

1. This one can be a "doozy". In general there is less noise in the higher timeframes (> H1) but it depends on what you mean by "reliable". I would say as a general rule that the lower the timeframe the more often you will need to swap out strategies that become unprofitable. Personally, I don't trade anything less than M15, mainly to decrease the risk of adverse moves due to news events and to decrease trading costs like spread and slippage.

2. You've hit the nail on the head with that one. That's exactly what over-optimising or "curve fitting" is. The less restrictive you are with your strategy, and the less degrees of freedom you give it, can minimize the chance of curve fitting.

3. IMO, no. I generate strategies on all available pairs. You are better off diversifying and trading as many pairs as are sufficient to diversify your risk. For example it wouldn't be wise to trade only the the following pairs:

USDCAD
USDJPY
AUDUSD
EURUSD

Why? Because you are heavily exposed to the USD. 4 pairs with USD, and 1 each with AUD/JPY/CAD/EUR.

When I generate strategies I also don't care about PAIR correlation, like EURUSD and GPBUSD. I do not care that the pairs in general are correlated, what is important is that the individual strategies that are generated do not have balance/equity curves that are correlated to each other. This can be performed using the EA Studio setting "Detect balance lines correlation" for strategies generated on the same pair. For different pairs (like EURUSD and GBPUSD) you are going to have to run them on a demo/live account then perform a correlation analysis.

Getting back to the main part of the question, you might consider generating strategies on higher time frames to negate "wild swings".

Re: General tips

burrup.lambert wrote:

Hi Magic147,

1. This one can be a "doozy". In general there is less noise in the higher timeframes (> H1) but it depends on what you mean by "reliable". I would say as a general rule that the lower the timeframe the more often you will need to swap out strategies that become unprofitable. Personally, I don't trade anything less than M15, mainly to decrease the risk of adverse moves due to news events and to decrease trading costs like spread and slippage.

2. You've hit the nail on the head with that one. That's exactly what over-optimising or "curve fitting" is. The less restrictive you are with your strategy, and the less degrees of freedom you give it, can minimize the chance of curve fitting.

3. IMO, no. I generate strategies on all available pairs. You are better off diversifying and trading as many pairs as are sufficient to diversify your risk. For example it wouldn't be wise to trade only the the following pairs:

USDCAD
USDJPY
AUDUSD
EURUSD

Why? Because you are heavily exposed to the USD. 4 pairs with USD, and 1 each with AUD/JPY/CAD/EUR.

When I generate strategies I also don't care about PAIR correlation, like EURUSD and GPBUSD. I do not care that the pairs in general are correlated, what is important is that the individual strategies that are generated do not have balance/equity curves that are correlated to each other. This can be performed using the EA Studio setting "Detect balance lines correlation" for strategies generated on the same pair. For different pairs (like EURUSD and GBPUSD) you are going to have to run them on a demo/live account then perform a correlation analysis.

Getting back to the main part of the question, you might consider generating strategies on higher time frames to negate "wild swings".

Great advice B :-)

Diversification and risk-weighted returns is what this game is about

4 (edited by hannahis 2020-07-26 15:41:18)

Re: General tips

burrup.lambert wrote:

1. This one can be a "doozy". In general there is less noise in the higher timeframes (> H1) but it depends on what you mean by "reliable". I would say as a general rule that the lower the timeframe the more often you will need to swap out strategies that become unprofitable. Personally, I don't trade anything less than M15, mainly to decrease the risk of adverse moves due to news events and to decrease trading costs like spread and slippage.

Yes lower time frames may seem to have more noises (if you don't know how to filter them out).  However, lower time frames EA is more versatile and reacts faster than higher time frames. EAS open and close based on Bar Open/Close data and imagine using a H4 EA that only open or close on 4 hours intervals! In a volatile market like forex, can a strategy be robust with such lagging reaction? Can a lagging EA consider as "stable" just because we use a longer time frame data does it mean you are using a more stable execution?  Data vs Execution is a different ball game.  Like driving, you use a Big Map (data) to plan which is the shortest route to a particular destination but when we are driving, we use a GPS (real time execution) to navigate us with real time response to get from point A to B.  Imagine using a GPS that will only update your driving position every 4 hours, you would have missed your turning points and go off the course.  Unless the trading style is position holding whereby the EA is supposed to trade and hold positions for weeks or mths.  I personally only use M1 time frame EA and don't venture anything more than M15 (maybe in future if I want to develop position trading EA but not now).


burrup.lambert wrote:

Because you are heavily exposed to the USD. 4 pairs with USD, and 1 each with AUD/JPY/CAD/EUR.

Provided you are using the same EA for all these different pairs and hence you are overly exposed, i.e the EA has the same risk management in terms of TP, SL and trading rules, hence the EA can repeat it's mistakes in all these different pairs.  Whereas, If I were to use different EA/strategies/trading theory, I may not be overexposed because I'm not using the same decision-making strategy (EA) to repeat my "mistakes" elsewhere in other pairs.  With different EA/trading styles installed in different pairs, I may not be necessarily overexposed.  It all depends on what do we mean by exposure (risk or market conditions).  If it's about risk, different EA or trading style has different risk exposure (SL, TP, trading rules).  If it's about market conditions such as volatility, then different EA has different trading styles/rules such as swing, scalper, intraday etc to handle these different market conditions.  So besides just plainly looking at the currency pairs, we also need to examine our portfolio composition of EA/trading rules etc.




burrup.lambert wrote:

When I generate strategies I also don't care about PAIR correlation, like EURUSD and GPBUSD. I do not care that the pairs in general are correlated, what is important is that the individual strategies that are generated do not have balance/equity curves that are correlated to each other. This can be performed using the EA Studio setting "Detect balance lines correlation" for strategies generated on the same pair. For different pairs (like EURUSD and GBPUSD) you are going to have to run them on a demo/live account then perform a correlation analysis.

Getting back to the main part of the question, you might consider generating strategies on higher time frames to negate "wild swings".


Balance line correlation does not necessary mean that the EA has the same trading styles.  It's like assuming if you are in the same income group, you must be in the same profession.  The same group of people who earns $10k a mth, does not mean they must be in the same profession.  Likewise, you can't really distinguish the EA simply by their Balance line correlationship.  For example, 2 very different strategies that somehow has very good entry and exit rules, they both would have the same balance line profile.  The balance line is a reflection of the accumulative net profit the EA make, not a reflection of the types of strategies it contains.  We can't read too much into the balance line to determine what kind of strategies is contained in the EA, unless you are going to zoom into the balance line to examine it closely.  But if you are backtesting it over large data period, I personally don't think the balance line is a good indication of whether the strategies have closely correlation entry/exit rules because I have a number of EA that have the similar balance lines but these EA are made up of different trading rules and have different trading behavioral and trading metric performances.

5 (edited by burrup.lambert 2020-07-29 18:27:09)

Re: General tips

I replied to this but it didn't post so I will make this one short and sweet!

Yes lower time frames may seem to have more noises (if you don't know how to filter them out).  However, lower time frames EA is more versatile and reacts faster than higher time frames. EAS open and close based on Bar Open/Close data and imagine using a H4 EA that only open or close on 4 hours intervals! In a volatile market like forex, can a strategy be robust with such lagging reaction? Can a lagging EA consider as "stable" just because we use a longer time frame data does it mean you are using a more stable execution?  Data vs Execution is a different ball game.  Like driving, you use a Big Map (data) to plan which is the shortest route to a particular destination but when we are driving, we use a GPS (real time execution) to navigate us with real time response to get from point A to B.  Imagine using a GPS that will only update your driving position every 4 hours, you would have missed your turning points and go off the course.  Unless the trading style is position holding whereby the EA is supposed to trade and hold positions for weeks or mths.  I personally only use M1 time frame EA and don't venture anything more than M15 (maybe in future if I want to develop position trading EA but not now).

I agree about getting in and out of trades quicker. I always generate EA's with "Always Use" for the TP and SL. I don't want any strategies hanging in the air waiting for an indicator based exit.

How are your M1 strategies going hannahis? Do you find you need to rotate new strategies in more often than higher time frames? What is the average shelf life for a M1 strategy?

Provided you are using the same EA for all these different pairs and hence you are overly exposed, i.e the EA has the same risk management in terms of TP, SL and trading rules, hence the EA can repeat it's mistakes in all these different pairs.  Whereas, If I were to use different EA/strategies/trading theory, I may not be overexposed because I'm not using the same decision-making strategy (EA) to repeat my "mistakes" elsewhere in other pairs.  With different EA/trading styles installed in different pairs, I may not be necessarily overexposed.  It all depends on what do we mean by exposure (risk or market conditions).  If it's about risk, different EA or trading style has different risk exposure (SL, TP, trading rules).  If it's about market conditions such as volatility, then different EA has different trading styles/rules such as swing, scalper, intraday etc to handle these different market conditions.  So besides just plainly looking at the currency pairs, we also need to examine our portfolio composition of EA/trading rules etc.

The methodology I spoke of is mainly to negate black swan type events like the swiss floor break.

Balance line correlation does not necessary mean that the EA has the same trading styles.  It's like assuming if you are in the same income group, you must be in the same profession.  The same group of people who earns $10k a mth, does not mean they must be in the same profession.  Likewise, you can't really distinguish the EA simply by their Balance line correlationship.  For example, 2 very different strategies that somehow has very good entry and exit rules, they both would have the same balance line profile.  The balance line is a reflection of the accumulative net profit the EA make, not a reflection of the types of strategies it contains.  We can't read too much into the balance line to determine what kind of strategies is contained in the EA, unless you are going to zoom into the balance line to examine it closely.  But if you are backtesting it over large data period, I personally don't think the balance line is a good indication of whether the strategies have closely correlation entry/exit rules because I have a number of EA that have the similar balance lines but these EA are made up of different trading rules and have different trading behavioral and trading metric performances.

Indeed. I intentionally omitted strategy correlation from my reply to focus solely on PAIR correlation. I personally generate strategies with the "detect similar trading rules" setting OFF.

I don't have a problem if EA Studio generates 10 strategies that are all just variations of two moving averages. I only care that the balance lines are not correlated. I figure that I can diversify that risk by constantly generating new strategies and adding them to the portfolio, the under performing ones (with or without similar trading rules) will get replaced with new profitable strategies that again, may or may not have similar trading rules.

I'm sure Rich can pipe in if he wants regarding trading strategies with similar rules like just varying indicator parameters wink.

6 (edited by hannahis 2020-07-29 19:09:59)

Re: General tips

Hi Burrup,

How are your M1 strategies going hannahis? Do you find you need to rotate new strategies in more often than higher time frames? What is the average shelf life for a M1 strat.egy?

I can't really comment about how long the shelf life because I kept improving my EA with newer and newer version of the old profitable one and hence I so call delete those old profitable ones to give space for the new and better ones.  Nevertheless I ran a number of M1 for more than 6 mths and they were still profitable and I didn't continue on because the news ones out perform the old ones in terms of pf and net profit.  But I think I'm coming to near the end of EA generation/improvement and would likely stop making new version and start using the current ones.  The early stage of EA development is where I will test out the EA for 2 to 3 mths and select the good ones to use as preset and test these new versions for 3 mths and again select the best ones as preset to generated better ones, this process is repeated several times till I'm satisfied with the overall outcome.  That's the reason why I can't really comment how long the EA last.  But they definitely last more than 3mths.  And I have no intention to keep refreshing them because I aim to develop EA that is robust which hopefully means long lasting cos I don't want to spend the rest of my life keep making new EA every few mths and hopefully have an early retirement smile

7 (edited by ats118765 2020-07-30 02:09:09)

Re: General tips

burrup.lambert wrote:

I'm sure Rich can pipe in if he wants regarding trading strategies with similar rules like just varying indicator parameters wink.

https://atstradingsolutions.com/wp-content/uploads/2020/07/Pipe.png

Nothing to add mate. :-)

Hannah is dabbling in a space (M1) that I fear to tread...so can't comment.

In relation to why I insist on a SL at all times....here is a thought to consider.

The reason for me at least is not as straight forward as simply 'protecting the downside' as we know that major 'crisis' events can shoot straight through any stops....so for adverse exposure protection....diversification with a small bet size is the only way out of this dilema.  Being diversified and having small exposure to CHF during the Swissy depeg was your only saving grace. Stops were not observed.

.....but the reason that I actually use stops on all my individual solutions embedded within a portfolio is to ensure the portfolio is not carrying warehoused risk. At all times my portfolio is 'deleveraged' as much as possible and no individual return stream is carrying excessive unrealised risk. That means that the portfolio is always optimised in terms of it's capability to absorb FUTURE RISK events.

The times you see a portfolio fail is when 'they are already highly exposed' to future tail events.

One of the central reasons that portfolio managers can go down the gurgler is that the portfolio is not prepared for really unfavourable conditions AT ALL TIMES.   Using stop losses on all solutions ensures that the portfolio AT ALL TIMES is not overexposed in risk NOW for a future event to turn that risk exposure into a disaster. The use of stops is like having a shock absorber in your portfolio. It keeps your portfolio lean w.r.t intrinsic risk.

Portfolios are like risk sponges and allow you to carry far more risk that what would otherwise be possible with a single return stream. So given that understanding you need to manage the intrinsic risk that lies in the portfolio at all times.

Just some food for thought for those that like to swim naked :-)

Diversification and risk-weighted returns is what this game is about

Re: General tips

Hi everyone,

what do you think of EA like that: https://i.ibb.co/qFcH8bz/small-gain-big-lose.png

These kind of EA always scare me because they are not very stable, they can have 15/20 good trade and boom just 1 trade to wipe all the profit, how can I avoid that ?

Tweaking stoploss and takeprofit ?

But whatever I do with the TP and SL the ea can still exit at +5pips even if my MIN TP is at +25pips (exit condition has been met)

9 (edited by zenoni 2020-07-30 12:12:32)

Re: General tips

I reccommend to use stop-loss for the disaster, but usually exit using on some dynamic criteria. I prefer multiple exits, both hard and dynamic ones. Just try to follow the principle - cut you losses short and let your profits run.

Re: General tips

jordytr971 wrote:

Hi everyone,

what do you think of EA like that: https://i.ibb.co/qFcH8bz/small-gain-big-lose.png

Run away fast....and for god sake.....don't look back :-)

Diversification and risk-weighted returns is what this game is about

11 (edited by burrup.lambert 2020-07-30 20:23:53)

Re: General tips

jordytr971 wrote:

Hi everyone,

what do you think of EA like that: https://i.ibb.co/qFcH8bz/small-gain-big-lose.png

These kind of EA always scare me because they are not very stable, they can have 15/20 good trade and boom just 1 trade to wipe all the profit, how can I avoid that ?

Tweaking stoploss and takeprofit ?

But whatever I do with the TP and SL the ea can still exit at +5pips even if my MIN TP is at +25pips (exit condition has been met)

It's hard to be precise because the statistics are not mentioned but it seems that the Average Loss is much greater than the Average Win. This usually occurs because the SL is several times the TP, for e.g, SL = 100 pips but the TP is only 10 pips.

I can think of several solutions:

1) Prevent this during strategy generation by changing the minimum/maximum pips for SL and TP.
2) Do not trade this strategy by itself. Incorporate it into a portfolio so that other strategies can offset the drawdown.
3) Use some type of trade/magic filtering tool to not take (or copy) trades that don't meet you minimum acceptance criteria such as: AverageWin / AverageLoss * -1 must be > 1 (the multiplication by -1 is if your AverageLoss is denominated as a negative number), or some type or return/ drawdown metric.

https://i.ibb.co/xhph46z/Capture.png

The green strategies are probably what you want to trade. The red strategies you probably want to avoid, and maybe you want to trade the orange one (purely based off their Average Win/Loss).