Topic: Generator - Working Time Setting

mel8331 wrote:

Can the Generator working time parameter be set. that when it is at 10000 minutes, there is no time limit on the Generator working time ?



When you set 10000, the Generator will work 10000 min, which almost 7 days.

Re: Generator - Working Time Setting

Popov wrote:

When you set 10000, the Generator will work 10000 min, which almost 7 days.

I wanted to explore if even better strategy could be found it it run for longer period.

What I do is, I am using FSB Pro for trading / demo and at the same time using it for generating strategy. So it make more sense to me if I can let the generator run "forever" if I chose, because it running all the time.

Re: Generator - Working Time Setting

mel8331 wrote:
Popov wrote:

When you set 10000, the Generator will work 10000 min, which almost 7 days.

I wanted to explore if even better strategy could be found it it run for longer period.

What I do is, I am using FSB Pro for trading / demo and at the same time using it for generating strategy. So it make more sense to me if I can let the generator run "forever" if I chose, because it running all the time.

Restarting or letting it run will have similar results over a period of time. The generator gives random results, some of them are good and some are not so good. The length of the run does not necessarily improve results.

I did both with FSB, ran it for two weeks sometimes... the results were no better than running it for a few hours often.

That is my experience, maybe someone else has had different results.

My 'secret' goal is to push EA Studio until I can net 3000 pips per day....

4 (edited by mel8331 2014-11-14 05:58:48)

Re: Generator - Working Time Setting

Can the Generator running time limit be raised to 99999, that  will just mean just changing some figure in the software code, so no major changes in the code needed ?

Re: Generator - Working Time Setting

mel8331, I can think of no reason you would like to do that... But it should be possible to run the Generator endlessly, if it does not increase load to your computer with time. If it takes the same amount of RAM and CPU(in general) to process each strategy, which I believe is the case, you could run it "forever".

You probably already know, that the Generator is working by creating random strategies, and backtesting them. If the strategy is good enough, it puts it on your screen, and adds it, as well as the old ones to the Collection in your local FSB Repository.

This process has consequences. If you run the Generator long enough you might get the best 100 strategies that are possible in the current conditions. However, best on backtest in profit or another criteria does not mean best in the real world.

Best in a backtest means totally over-optimized/curve-fitting strategies. You probably already know what curve-fitting is, but for the sake of all other users that may not know...

Curve-fitting (over-optimization) is what it says it is. You end up with a strategy so tightly tailored for the historical market data you have, that it can not preform nearly as well in the real market, as it did when backtested.

Re: Generator - Working Time Setting

One idea that I use is to start to build the strategy I want by adding a couple indicators and setting whether to use stops and take profit and break even..

I check the chart to see if there will be entries in suitable places.

Let the thing run for an hour and examine the number of consecutive losses and win loss ratio, and then set the the minimums

Then add, perhaps volume. and see what occurs after the generator runs for a while....

and then I test whether 'long or short only' will help....

next is 'entry time'

etcetera.

The reason I do these is to evaluate whether I can make a strategy with those indicators.

If there seems to be a possiblity that I have a strategy that will be workable..... I start to examine all of the statistics and  then 'tighten up the 'acceptance criteria' in the control panel.

So........ much of my time in building a strategy is spent examining the chart and manually developing a strategy, the generator is a supplement most of the time.

For example..... today I am building a strategy to work with breakouts, I put a couple momentum indicators on the chart, let the generator run for a while, had a look at the chart and decided to try with volume..... a few minutes on the generator and back to the chart, to see if the results were going to be appropriate.

Then a moving average to see what that would do.....

etcetera.

I still have a lot of work to do on this strategy.......... the one thing I know already is...... hours on the genrator is not going to give me what I want, I have to structure my approach in a better way as I know where I want the entries to be.

As Yavor says..... using the generator for a gazillion hours will lead to curve fitting and be ineffective.

Time is best spent on the chart to build the strategy and the generator and optimizer used for some fine tuning.

I have used the generator to find strategies...... ran it for days and days.... not a lot of luck; we have to control the thing and develop strategies with the chart and then perhaps allow the generator to add options that may improve performance.

This is just an opinion........ take from it what you might want.

Nuno , if he will, has a lot to add about developing a strategy and using the generator.

My 'secret' goal is to push EA Studio until I can net 3000 pips per day....

7 (edited by mel8331 2014-11-14 22:53:02)

Re: Generator - Working Time Setting

One technique I use in the hope of preventing curve fitting is at the start of a strategy generation, I check the Martingale option.

Since is not viable in the long term, the random strategy selection will eventually select strategy that do not need Martingle, ie if you unchecked this option, it does not affect the result.

I think that adds fitness to the selection process instead of just pure curve fitting.At least I am hoping it will not allow strategy that has consecutive losses that will not be viable if double the bet.