Intensive Review Genetic Builder Intensive Review

[Pharaoh]

Genetic Builder Intensive Review
By Pharaoh​

Normally I'm way too busy (doing things like supervising the work on my new pyramid and telling that Moses fellow that I will not let his people go ) to do an intensive review, but this is one I just couldn't pass up.

Genetic Builder is software that creates EAs. Its fairly new and there aren't any reviews for it yet (FPA Review Page is at GeneticBuilder | GeneticBuilder.com reviews and ratings by Forex Peace Army).

I've been looking for something like this for awhile. I sometimes have ideas for an EA, but I don't have the time to sit down and learn MQ4. I haven't read the manual yet, so will just tell you what I learned from the Genetic Builder sales info on the website for starters.

As I understand it, you start by selecting some general parameters. Then you turn Genetic Builder loose and it goes through hundreds (thousands?) of combinations and lets you sort out the ones that show the most promise. It seems to be more loosely organized than some other EA builders where you specify a set of If This, Then That logic. The reason is that Genetic Builder is designed to try things you may not ever have thought of.

One very interesting feature is the way to split the test data. You pick a date range of data for initial EA creation, holding back some data for followup testing. Then take your best generated EAs and run them again on the older data used to create them as well as on newer data to see how well they hold up against "unknown" data. This will quickly kill any EAs that are locked into success only from data from a specific date range.

Looks like this thing needs to run for a few days to chew through all the possibilities. Guess I'll dust off a spare computer so that it can run without disturbing me.

So, my plan is to sit down this weekend, read the friendly manual, and then try to see what I can create.


A few of the questions floating around in the back of my mind:

1. Just how much control do I get for telling these EAs what to do? For example: Can I specify things like money management, adding trailing stops, etc.

2. If I get a very promising EA, is there a way inside Genetic Builder to try to further improve it by adding/removing components and/or altering settings, or will I have to figure that out inside of MT4's MQL editor?

3. Since the forex market conditions change over time, would it be a savings of time and effort to develop EAs using the last 2-4 years of data instead of the last 10 years?

 

[RahmanSL]

"Normally I'm way too busy (doing things like supervising the work on my new pyramid and telling that Moses fellow that I will not let his people go )"

Don't forget to check on those embalming people;yah! Evidence have shown they have been rushing some jobs.
=======================

At last, a super-duper EA!

"...develop EAs using the last 2-4 years of data instead of the last 10 years".
If you check the second pyramid (I think) of Giza, fifth inner chamber, third level down, you should be able to develop your super-duper EA based on the hieroglyphics curved into the side walls and ceilings.

Once you have that EA ready, we can incorporate that into my "top-top secret" trading strategy and we WILL be holding the forex world "reproductive organs".

 

[Mark Fric]

Hello Pharaoh,

I'm the author of Genetic Builder, so I can answer all your questions.

First of all, Genetic Builder is not a strategy building wizard, in a sense that you have to specify the strategy rules which is then transformed to MQL.
It is something totally else, and I might say superior to that
It is strategies generator - which means don't need to know and define how exactly your new robot should work. You only have to specify building blocks you want to use (indicators, price, time etc.), then specify some general conditions like max size of SL/PT, trading time range etc., and GB will find the best combination of price patterns, technical indicators, and entry and exit rules to generate new unique strategies and writes the corresponding strategy code.

It can really generate thousands of new random strategies per hour and backtest them right away. The number of combinations of various indicators and parameters is almost infinite, so you can let it run for a few years! and it will still generate new unique strategies.

Because they are generated randomly, most of the strategies will simply make no sense and will be dismissed right away, but those that survive the filtering process will be new unique strategies based on the preselected building blocks that will be profitable.
GB will generate hundreds of such strategies, and you only have to evaluate them and choose the best ones.

By the way, there are many exciting features planned in the future, also a strategy wizard that will allow you to manually build or edit a strategy and possibly optimize it in GB.

1. Just how much control do I get for telling these EAs what to do? For example: Can I specify things like money management, adding trailing stops , etc.

I think you should try it to see how it works. You cannot (yet) tell it the rules for entry or exit, these are generated randomly.

Percentage based Money Management is included in the generated MQL, trailing stop is not yet supported, but it will be added in the next version.

2. If I get a very promising EA , is there a way inside Genetic Builder to try to further improve it by adding/removing components and/or altering settings, or will I have to figure that out inside of MT4 's MQL editor?

yes. there is a feature called Strategy parts optimizer which allows you to take a generated strategy and try to generate better strategies based on the existing one, only with different exit rule (for example).
Or different rule for long entry.

3. Since the forex market conditions change over time, would it be a savings of time and effort to develop EAs using the last 2-4 years of data instead of the last 10 years?

depends on which timeframe you use, testing on 4 years of data could be sufficient for shorter timeframes

 

[Pharaoh]

Hi Mark!

Thanks for letting me get to play with your software and answering questions.

I read through the manual and have a better idea of what's going on. I got a little tied up over the last 2 days, but should be able to turn it loose on a spare computer today or tomorrow.

 

[Pharaoh]

That was fun. I hadn't used that computer for a month. While it was collecting dust, I renewed my anti-virus contract. Dusted off the computer, cranked it up and it said the AV software was expired. Sent it to the website, which said the AV software was fine for another year, but the copy on the computer kept having issues. Ended up having to do multiple updates to Windows Vista just to be able to install the latest version of my antivirus, but finally got that going.

Installing Genetic Builder was very simple and straightforward.

One issue I encountered. Not all the options are displaying on all the screens (see attached). I should be able to tell it which strategies to dismiss, but that part isn't showing. The monitor is maxed out at 1600x900 so I don't think that's the issue.

i decided to go ahead and see what I could do without full control. I ran it in Evolution mode. It was much quicker than I expected and went through 100 generations within a few hours while only using a small amount of the available memory. It maxed out the fitness after about 20 generations.

Overall, I'm very pleased with the potential, but need to figure out how to get access to all the settings.

Mark, any thoughts on how fix the access on the other computer or should I reinstall it on my main system that's running Windows 7?

 

[Pharaoh]

I found another area that could use a slight upgrade. My favorite uncooperative computer was churning through the evolution process and decided to reboot due to an windows update - the last I saw, the fitness of my best strategy was at about 80% and the overall population fitness had just broken 50%. 20 minutes later, I came back into the room and the computer was awaiting login. The issue is that Genetic Builder didn't save anything. It would be nice if it could retain its work under these circumstances.

One other small request for a feature upgrade. Its possible to tell it to close at the end of the day. For the most part, end of the day doesn't worry me. I'd like to see a "close 2 hours before weekly market close" feature. That would remove the risk of weekend gaps.

 

[Mark Fric]

thanks for the bug report. That issue with settings seem to be related to font size - it seems you have increased your default font size in Windows so the texts are bigger than normal and that's causing checkboxes to move away from the panel.
Anyway, I'll look at it and fix it.

The idea of weekly closing is interesting, I'll add it to the future version.

If you want to try Genetic Builder on another computer just send me PM or contact me by email and I'll generate new license for you.

 

[Mark Fric]

an update has just been released that should fix the problem with invisible settings. As I expected, it was caused by increased Windows fonts, it should work now.
You can get it from the GB download page.

 

[Pharaoh]

Just changed the font settings on the weekend and I can see everything (spent awhile digging those out of my control panel). Stiil I think I'll download the new version.

One flaw I noticed while playing with it this weekend. Dismiss strategies with negative OOS P/L didn't seem to work. I've seen them pop up in Random Generation as well as in Gen 1 of Genetic Evolution. Guess I'll see if that's still the case with the new release.

BTW - Decimation rocks! The best results I've seen so far were from generation 1 an evolution build with decimation set to 10. Took a LONG time to get that first generation, but the results were worth it.

Quick question - Assuming I'm telling it decimation of 10 and population size of 100, it will work through and create 1000 strategies. What happens if during that process I go into the results and manually kill off some of the less promising strategies? I.E. If I go in when there are 600 strategies that it made and I kill 100, will it notice that 100 have been "pre-decimated" and work it's way up to 900 strats before decimation occurs, or will it keep working until there's a total of 1000 strats in the databank before decimation happens?

If the latter, this is a cheap and dirty way to push decimation past the max setting of 10.

Got a few dozen more questions for you, but I want to get a look at the new version to see what else may have changed. Plus, I'll be able to reset my font size back to "you really need new glasses" with it.

 

[Pharaoh]

The new version is running.

The good news: The font size issue is fixed, so I can pretend I don't need glasses a little longer.

The bad news: Dismiss Strategies with negative OOS P/L still isn't working in either Random Generation or Genetic Evolution (even during the initial generation).

Hopefully that can be fixed with the next update.

I've been tweaking my Evolution settings. I've noticed that if I run 100+ generations, evolution has a bad habit of getting stuck in a rut and I still haven't quite been able to keep it from doing that. By this, I mean that certain strategies with tiny variations take over and there will be a dozen or more strategies with only tiny difference, then another set of near identical results. Maybe I should increase the mutation factor? That would result in more bad results, but should also encourage more variety.

Another plan I have is that I've started saving my best results from each build with labels of A, B, and C. When I get enough A's and B's, I'm going to load them all into the databank and use them as the initial sample for a new round of Genetic Evolution.

New feature suggestions:

In addition to "Move SL to BE", a trailing stop feature would be a nice addition.

Allow different Min/Max for SL vs TP. This could be used to create a fixed or limited range risk/reward ratio.

There's a max trades setting. I think there should be a minimum trades setting. I keep getting some strategies that only make a dozen or less trades across several years worth of data. Even if every trade in and out of sample is perfect and it backtests perfectly in MT4, a tiny handful of trades isn't enough to create any confidence that the EA will have future success.

On the other hand, for people who really want to trade longterm and have large enough data sets to get enough trades, adding Weekly to the available periods would be a nice addition.

Tiny suggestion - include a list of what changes are made with each update. That will enable users to check out new features they may not be aware of and to make sure that fixes and additions really work properly.

Being a glutton for punishment (and running the program on a spare computer), I'd be happy to see decimation values above 10 available.


Overall, I'm still having way too much fun playing with this program.