- ago
The new Genetic Evolver is a great tool.

To improve transparency and possibilities even more I'd suggest to add more visualizations of the progress and results of the evolver.

After it has stopped/finished the evolver could call plugins/extensions which implement the existing Optimizer Visualizer API.

Instead of Optimizer Parameters there would be columns like "Generation" "Gene-Combination-Id" and so forth along with the usual performance metric columns exactly like in an optimizer result set.

The API could have an additional "bool ForEvolverResults" to allow an extension to find out why it is running.

I'd expect WL to add the first visualizer "Tabular Results" and leave others for creative API users and extension developers.

The table (and result set) should probably contain strategies that survived a certain number of generations only.

Such visualizers could have the following benefits:
* more insight in the Evolver Process which in itself will spark new ideas
* more insight in promising Genes (aka working strategies)
* Detection and measurement of over-optimization


5
666
14 Replies

Reply

Bookmark

Sort
- ago
#1
QUOTE:
Such visualizers could have the following benefits:

...or introduce unnecessary complexity that would overload the GUI to the extent of scaring away its potential users 😂
0
mjj37
- ago
#2
First of all, it's an amazing first cut at this.

Along the line of DrKoch's suggestions, it would be nice to make this a more generic tool that can be extended. It would be helpful to be able to:
a) tweak the genetic algorithm settings
b) target the evolution more, such as:
1. Having an existing strategy and solving for just iterations of that strategies entry conditions and exit conditions.
2. Having an existing portfolio of strategies, and solving for a strategy(s) that mitigate risk (they could even be strategies that lose money but mitigate enough risk to run them).
3. Having an existing strategy and solving for the top conditions that improve the strategy's performance (eliminate the co-variants)
c) Allow the user to save these search settings so they can be reproduced/updated
d) For the hedge strategies, it makes sense to have those be based on stock index futures as they can be overlaid on top of existing stock based strategies with less margin impact.
e) Defining a list of viable strategies and solving for the best combination of strategies, their parameter settings, and their ideal weighting.
0
- ago
#3
You forgot
4. Increasing the subscription price to match competitors' offers 😂

Just visited the website of a backtesting software with genetic evolver alternative which costs an order of magnitude greater p.a. than WL7 !
0
Glitch7
 ( 83.25% )
- ago
#4
The best outcome would for the zealous WL7 users to spread the word to others, and help get our subscription counts rising!
0
mjj37
- ago
#5
Have you guys considered doing a podcast/presentation on:
a) Better Systems Trader http://bettersystemtrader.com/
b) https://futures.io/

For me personally, I would be happy to pay a significant amount upfront to own a version of the software in addition to paying the annual subscription for future updates. I've spent thousands of man hours at this point converting my trading over to WL7. My biggest fear is that you guys go in a different direction with the platform and I'm left high and dry. That's why I keep my own code stack for execution and do much of my research on WL7.
0
Glitch7
 ( 83.25% )
- ago
#6
That looks interesting, thanks for the lead. Let me see if Volker wants to try and get something brewing over there!
0
- ago
#7
May I ask who you envisioned as the target audience for the Genetic Evolver?
0
Glitch7
 ( 83.25% )
- ago
#8
Myself :) Hopefully some others find it useful too!
0
- ago
#9
Okay, just trying to understand more about the direction you're taking the software.
As mjj3 pointed out, when you've invested thousands of man hours in a particular software technology, it pays to understand as much as possible about where it's headed.
0
Glitch7
 ( 83.25% )
- ago
#10
Makes sense. FYI we drive much of the new development based off of the most highly upvoted feature requests in the wish list. And this item happened to be at the top, so other users were supporting the development besides myself.
1
- ago
#11
Thanks Glitch.
You mentioned helping to spread the word about Genetic Evolver. And Eugene mentioned raising the price.
So that I understand its positioning and purpose better, is it fair to say that Genetic Evolver is similar in concept to something like Adaptrade Builder, albeit simpler in scope?
If so, since it may not be of value to all users, would it make sense to develop it as a cost option?
0
Glitch7
 ( 83.25% )
- ago
#12
I think Eugene was making a joke when he mentioned raising the price, we don't have any plans to do that.

My comment was just a general one, I'm hoping to see some strident WL7 users eventually start showcasing the product's features in the trading vlogs, articles to TASC magazine, etc. I think that would reallty accelerate our growth.

So no, I don't wish to make this an added-cost feature, I like having it there as a core tool.
0
- ago
#13
Okay, I didn't realize Eugene did jokes ;)
I'll wait for some product reviews and then forward the links to anyone I think might be interested.
1
- ago
#14
The discussion wandered a bit away form the original intend...

To bring it back here an example graph showing the Evolver's progress:



It shows some metrics of the then-top strategy vs. generation number.
Lacking the API mentioned in the original post, this graph was generated using an alarm clock, reading values off the screen and typing this values into excel...

If you want to see such graphs for your own Evolver adventures please vote for this #FeatureRequest....

... its worth the click... ;)
0

Reply

Bookmark

Sort