MoGo - new genetic optimiser for NT 6.5

[-Swig-]

MoGo v1.1 released

Yes, I do have boolean parameters, but those are not included in the optimization universe. I also run the latest version of NT.

Marc

This was the problem. MoGo relied on being able to read all strategy parameters from NT, but NT was blowing up when it tried to fetch the value of a boolean property.

Version 1.1 is out which fixes this. The (minor) downside is that the CSV exported when you select 'Save all trials' in MoGo will not include any non double/int parameters that your strategy may have.

Link:

http://mcraig.net/site/Development/NinjaTrader/MoGo/tabid/59/Default.aspx

-Swig-

[bibulous]

This was the problem. MoGo relied on being able to read all strategy parameters from NT, but NT was blowing up when it tried to fetch the value of a boolean property.

Version 1.1 is out which fixes this. The (minor) downside is that the CSV exported when you select 'Save all trials' in MoGo will not include any non double/int parameters that your strategy may have.

Link:

http://mcraig.net/site/Development/NinjaTrader/MoGo/tabid/59/Default.aspx

-Swig-

Thanks! Seems to work now.

Marc

[-Swig-]

Source code posted

Glad to hear!

The source code is now available from the website for anyone interested.


-Swig-

[laparker]

Thanks Swig
It's up and running and performing well.

-thanks

[CRCatNinja]

Swig:

Love working with this. Hey...I tried it with Walk Forward Testing and it forces you to stay in front of your computer to restart at every new Look back Period.

Just a thought...somehow allowing it to run the entire WF without manual continuation.

Thanks again for your work

[-Swig-]

CRCatNinja,

There is specific functionality in MoGo for this. If you hold down <Ctrl> the first time you click OK to start optimisation, MoGo won't re-display the parameter window until you hold <Ctrl> again.

Please refer to the usage instructions on the website for more details.

-Swig-

[Buttermilk]

variation in MoGo Output

Swig

With constant input parameter, I am getting significant variation in output. Is there MoGo parameters that need to be set to minimize this problem?

[-Swig-]

Hi Buttermilk,

This is largely to be expected, since MoGo randomly creates, mates & mutates its candidates. In that respect it is non-deterministic.

Genetic optimisers in general do not guarantee to find the global maximum (i.e. optimal solution); rather they find a decent solution in vastly fewer trials. They can get 'stuck' in local maxima, particularly if the surface of your fitness function in parameter-space is very bumpy, which it's likely to be for pretty much any trading strategy with more than a couple of parameters.

You can improve performance by increasing the number of generations and the population size (note I don't think it's a good idea to increase the mutation rate), but naturally the optimisation will take longer as you are running more iterations. The only sure-fire way to find the optimal solution is to exhaustively search every combination, which is what the default NT optimiser does. And it takes forever.

FWIW I often use MoGo with around 20 generations and a population size of around 1000. That's given me reasonably consistent results from run to run (though never quite identical).

Swig

With constant input parameter, I am getting significant variation in output. Is there MoGo parameters that need to be set to minimize this problem?

[rGQRGs7]

Hi,

If the space state is huge a genetic optimizer won't find a minimum quickly.
So, is it possible to give mogo a trial or set of trial values to start with.

Thanks

[StarBright2009]

De-Installation of MoGo

-Swig-

How does on De-Install/Un-Install MoGo from Ninja Trader?

Thanks

[StarBright2009]

Can anyone answer the question. If I add an optimizer and later wish to remove it from Ninja Trader, what steps should I take?

Thanks

[-Swig-]

Hi,

If the space state is huge a genetic optimizer won't find a minimum quickly.
So, is it possible to give mogo a trial or set of trial values to start with.

Thanks

Hi rGQEGs7,

MoGo doesn't support this at present, however the complete source can be downloaded from the website if you wish to add it yourself.

[-Swig-]

-Swig-

How does on De-Install/Un-Install MoGo from Ninja Trader?

Thanks

Hi StarBright2009,

After installation MoGo will appear in Windows' 'Add/Remove programs' list under Control Panel. Clicking Remove on the MoGo entry will uninstall it.

[pespinar]

Dead-end detection

It is posible to disable it? It always pop up with the same question, and it is a nuisance when you try to optimize several instruments at once...

Un saludo,

Pablo

[-Swig-]

It is posible to disable it? It always pop up with the same question, and it is a nuisance when you try to optimize several instruments at once...

Un saludo,

Pablo

This dialog shows up if MoGo hasn't found a new untested combination of strategy settings after one million attempts. As such the only reason I can think of as to why you get this is because your population size and/or number of generations are set too large for the range of possible values for all your strategy parameters, so MoGo is running out of candidates to evaluate.

My suggestion would be to reduce one or both of these values until you don't get the popup anymore. If you're regularly getting seeing it, you may as well use the default NT 6.5 optimiser because MoGo is doing an exhaustive search more or less.