Removing the golden handcuffs

fan27 said:
This coming Friday will be my last day at my company of thirteen plus years. The company builds and sells SDKs and SAAS products centered around digital imaging. I have learned a great deal about what it takes to build, sell and support software. Time to take what I have learned and apply it to my own ideas. Having a family and children (one under 5 years old), my scarcest commodity is time. Not for long!

For the past year and a half I have been working on an SDK/App (C# .NET Core) that uses machine learning techniques to dramatically automate finding trading strategies. It currently supports EOD as has exceeded my expectations in terms of usefulness. I have two more items to complete to fully prove out my idea before I can polish it up and make it a commercially viable solution.

1. Add support for intraday futures data and be able to find profitable strategies.
2. Add ability to automatically generate code from those strategies to be used in an external platform for live trading (considering AlgoTerminal as the first platform to target).

Assuming I get those two items figured out, I should a fully functioning product by October 1. I have a considerable financial runway but am only giving myself until the end of 2018 to get some initial traction with the product.

Good times!!
fan27
More...

Good luck with the project! Any plans to open source it? Sounds very useful
 
Polygon.io said:
Good luck with the project! Any plans to open source it? Sounds very useful
More...
Thanks! No plans to open source it. I am however open to a mutually beneficial exchange of services/licensing in certain situations.
 
I wish you well with this.

How will yours compare to the products already commercially available that do the same or similar?

How do you plan on competing, or coexisting with them?

Thanks

fan27 said:
This coming Friday will be my last day at my company of thirteen plus years. The company builds and sells SDKs and SAAS products centered around digital imaging. I have learned a great deal about what it takes to build, sell and support software. Time to take what I have learned and apply it to my own ideas. Having a family and children (one under 5 years old), my scarcest commodity is time. Not for long!

For the past year and a half I have been working on an SDK/App (C# .NET Core) that uses machine learning techniques to dramatically automate finding trading strategies. It currently supports EOD as has exceeded my expectations in terms of usefulness. I have two more items to complete to fully prove out my idea before I can polish it up and make it a commercially viable solution.

1. Add support for intraday futures data and be able to find profitable strategies.
2. Add ability to automatically generate code from those strategies to be used in an external platform for live trading (considering AlgoTerminal as the first platform to target).

Assuming I get those two items figured out, I should a fully functioning product by October 1. I have a considerable financial runway but am only giving myself until the end of 2018 to get some initial traction with the product.

Good times!!
fan27
More...
 
userque said:
I wish you well with this.

How will yours compare to the products already commercially available that do the same or similar?

How do you plan on competing, or coexisting with them?
Thanks
More...

The similar products I have seen are mainly geared towards users who do not write code and are a desktop applications. The platform I am building will be an SDK component (compiled - no source code) which is used by a console application with full source code provided. The platform will be fully extendable with the ability to build custom code generators targeting any ATS platform.
 
Performed a live trading test against an IB paper trading account with AlgoTerminal this week and the executed trades matched the back tested results (except for a few ticks here and there due to real time fills). I ran 12 strategies concurrently generating about 30 trades per day. So far AlgoTerminal looks like a solid choice as an ATS. My next task is to generate code from my platform that can be plugged directly into AlgoTerminal. I am designing this in such a way so that my platform will translate strategies/algos into a generic data structure which can be used by any number of code generators. The idea is that a user of the platform could look at the source code for the AlgoTerminal code generator I am writing, and write their own generator for whatever platform they are using for live trading.
 
I will be wrapping up the AlgoTerminal code generator in the next couple of days. This was a bit more effort than estimated but the result will be a common way of building other code generators. My top priority is to be live trading ASAP. Once that happens I will polish up the product a bit and get it to some traders who are currently running algos in some C# based platform. Preferably some people who have decent programming chops and could write their own code generator based on what I did for AlgoTerminal while using the source code as a template. PM me if the above describes you and you are interested in trying out the tech.
 
I am now generating strategy code from my system which can be "dropped" directly into AlgoTerminal and executed. That being said, I have discovered a show stopper data issue in AlgoTerminal which if it cannot be resolved, I will have to abandon this platform and look for an alternative for live trading. Hope to hear from their support on Monday and will report back to this thread. The irony of the situation is I have automated everything from finding strategies, optimizing for different orders, and combining the strategies based on performance metrics with the output being code for an ATS (currently AlgoTerminal). StockSharp looks like a possible ATS alternative. The key feature I need is to be able to execute multiple strategies together and to have a component that controls risk of the entire group of strategies.
 
fan27 said:
The key feature I need is to be able to execute multiple strategies together and to have a component that controls risk of the entire group of strategies.
More...
So does this mean AlgoTerminal doesn't allow different strategies to communicate with each other?
 
You must log in or register to reply here.
Back
Top