@denilson1
I've been "monitoring" LEAN development since 2014. I remember that time it was not really usable to any serious algo development.
But a lot effort must have been done on the way since then, because recently I realized how mature and rich platform it has became (e.g. they have Docker image so one can debug algorithms against Docker!).
IMO it's the best retail algo platform these days. I might be wrong though, nobody's perfect
I've been "monitoring" LEAN development since 2014. I remember that time it was not really usable to any serious algo development.
But a lot effort must have been done on the way since then, because recently I realized how mature and rich platform it has became (e.g. they have Docker image so one can debug algorithms against Docker!).
IMO it's the best retail algo platform these days. I might be wrong though, nobody's perfect
Last edited:
I dont' understand...
You code a ruleset, you test it but you should place the trades manually.
Placing trades manually you only acheive a delay of seconds or minutes, I dont' think this kind of delay can turn a losing strategy into a winning one.
About institutions, I think we had more "fat finger" mistakes from humans than from computers.
Agree, performing trading on faster TFs like ticks/minutes the delay is the case with no doubts. But even models running on tick or minutes data requires "recalibration" from time to time, which is also kind of manual intervention. There is no model functioning indefinitely forever.
Hence I mentioned the dream of common retailer thinking that one ex4 file combining several combination of indicators makes the forever profitable trading robot.
In ETF or Stock universe algorithm usually search sophisticated models consists of several data sources and containing thousands of assets and based on algorithm also its combinations (e.g. looking for cointegrated combinations or portfolios). If the strategy is running on some higher timeframes the humans can have last look option before placing the trade.
See this article Using QuantConnect to Automate Investing Strategies: Chacon Diaz and Di Virgilio Wealth Management . I do not know how much of marketing is included inside, but it's somehow how it can work.
Hope that makes sense.
Last edited:
Hi @Amazon
apologies to be a bit negative here, but IMO hiring programmer to create some proprietary (meaning your very own) back testing and live trading platform is not the right way. I'm professional software developer for whole of my life (~25 years), I do actively know several languages including C# and Python, and I do not have pretensions to write my own back-tester and live trader platform. Another problem is that working with external programmer you are locked-in to work with the programmer who is the only one who properly knows the code.
I guess for the same reasons guys above were kindly suggesting to use some more or less standardized and mature software packages (QSTrader, Zorro) for custom strategy development. Unfortunately both suggested platform are not capable of C# development, as you mentioned you require.
Hence I'd see like the most viable option for you is using LEAN by QuantConnect . It's open source system, unlike Zorro, strategies/algorithms can be written in C#, unlike QSTrader and Zorro, it's free if you run it on your own machine or VM, unlike Zorro, and last but not least it's capable of real trading with IB, unlike Zorro and QSTrader, as well as and many other brokers, unlike QSTrader. The LEAN allows trading Forex, Futures, Options anything you can imagine, and have got back tester as well equipped with data (depends what you need, for some data you might need at least demo account with broker) , so you should not need external database to store your data. And even though you need custom data source, you can plug custom data through csv files (e.g. I've been using minute data for my back test). If you have ever been using Quantopian the LEAN is the same concept except LEAN is capable of live trading.
The QuantConnect have got also rich documentation and Bootcamp (accessible after free registration) which guides you through algorithm creation process in form of hands on coding exercises (you can chose C#, Python, or do both of course) through the platform so you should be able to code your strategy after few weeks spending of learning the platform.
Once you break through to any package you chose you can focus on writing only your algorithm and stop wasting time with unnecessary trading support code (like e.g. resilient broker connection, which is huge work per se, capital management, risk management support, handling data quality issues, you name it) and finally the code will be only yours. No further stealing of your intellectual property.
I know it's not you were asking for but hopefully it helps a bit.
Regards
Hello Mcquant,
Thank you for your message. Not negative at all and very helpful. I am not a prgrammer so advice from expereinced people is always welcome. I undertand what you mean about custome built systems. I have looked ath Lean just now and it looks good, but makes very little sense to me. Iwill have to get someone to install it and work on it a little for me.
I agrre that once you breakthrough a certain point it gets easy. I have previously worked with Javascript and also R which I gor familiar with. It tooks many of the boxes but I will have to see how it is when loaded up.
If you have used it you may be able to tell me if the stratgeies re stored on their server or remain on on my server. I know some of these programs require you to upload it because the platform is on their server, which means they can view your code.
Thank you
Some interesting comments about auto trading. I completly understand the views expressed.
But i am stil not sure why one would not trade it on auto completly. I understand the slippage etc and market price changing quickly, but I am sure these can be coded.
Wth regards to entries and exits, i think these would be the most important requirements tostart even thinking about autotrading so maybe the concern regarding exits could be eliminated.
The biggest thing I can think of is regarding how long the system will continue working for before it changes its outcomes. I understand this aspect and more importantly what causes the systems to fail when they do. This is one of the major factors of my work and the system itself. Although I ahve not proved this in the way I prefer, via an auto tested system and data. I ahve been using the approach manually for some time.
But I do not use indicators and agree with the points on using these.
I also agree with the fact that are way too many combinations to make it possible, once again I believe there is a way. I have devised my method and in theory it looks and makes sense, but any issues with it will be discovered when I can actually backtest it and maybe even run it live.
But i am stil not sure why one would not trade it on auto completly. I understand the slippage etc and market price changing quickly, but I am sure these can be coded.
Wth regards to entries and exits, i think these would be the most important requirements tostart even thinking about autotrading so maybe the concern regarding exits could be eliminated.
The biggest thing I can think of is regarding how long the system will continue working for before it changes its outcomes. I understand this aspect and more importantly what causes the systems to fail when they do. This is one of the major factors of my work and the system itself. Although I ahve not proved this in the way I prefer, via an auto tested system and data. I ahve been using the approach manually for some time.
But I do not use indicators and agree with the points on using these.
I also agree with the fact that are way too many combinations to make it possible, once again I believe there is a way. I have devised my method and in theory it looks and makes sense, but any issues with it will be discovered when I can actually backtest it and maybe even run it live.
you can use this tutorial "Custom Trading Server A How To Guide" to make it running on your local machine or server.
Once you set up and run the LEAN on your own server nobody can see your code. It requires some effort but it worth it.
The QuantConnect allows running it life on their infrastructure, but as you said, your intellectual property can be compromised then.
Just a few cents from a stranger…
It's really kind that you offer to use your strategy. But consider that programmers with experience in automated trading might (hopefully) have their own strategies within their very own comfort zone and risk parameters, and therefore no use case for other strategies other than basic curiosity. If you find one you might need to simply pay her/him.
I read your posts twice and I really don't understand what you have and what you don't have. Did your programmer develop a standalone C# app? Taking data from where, do you do order management by yourself, how do you want to visualize and backtest and operate and monitor?
Have you considered (besides any error that could arise like code malfunction, data feed server down, ...) your trading costs, slippage, ...?
Do you have a basic understanding of statistics = can you put a number on "how sure are you your results are not random"?
If you continue down this road, it will be a long journey, but can be rewarding, especially if you bring an open mindset to the table and enjoy learning about many side topics …
All the best!
It's really kind that you offer to use your strategy. But consider that programmers with experience in automated trading might (hopefully) have their own strategies within their very own comfort zone and risk parameters, and therefore no use case for other strategies other than basic curiosity. If you find one you might need to simply pay her/him.
I read your posts twice and I really don't understand what you have and what you don't have. Did your programmer develop a standalone C# app? Taking data from where, do you do order management by yourself, how do you want to visualize and backtest and operate and monitor?
Have you considered (besides any error that could arise like code malfunction, data feed server down, ...) your trading costs, slippage, ...?
Do you have a basic understanding of statistics = can you put a number on "how sure are you your results are not random"?
If you continue down this road, it will be a long journey, but can be rewarding, especially if you bring an open mindset to the table and enjoy learning about many side topics …
All the best!
Thank you for your reply Android24, I understand what you are saying. From my own experience over the last 14 years or so with programmers, I find that most ofthe mare very good programmers but very very few also good traders. There is an edge they can develop as programmers which makes some of them proftable some of the time. But most i have come across are actually limited by their programming skills, the ones who are also good traders and understand the markets fully are already doing well and you will rarely find them on the forums or public eye. If theyare curious it is only because they may have a strategy that can be improved.
Fortunately my programmer is backnow so I dont have the requirement any more, but yes it is a standalone C#App taking its data from IB and have taken all those things into consideration.
Iknow it is along road, have been at this lark for years now, so seen most of the difficulties( I think). In terms of statistics my trading is proof enough. I have a high percentage of winners, even while using part of the system manually, automation will only increase the opportunities and improve results. I can see, and also define when the system will work and how it will perform qith a fair degree of acuraccy.
Whether it can be transferred to code 100% , I will soon find out, I personally believe it wil be down to the coding ability that will eventually define this, but am confident that it can be done.
Thank you
Interesting what you say about your experiences of only few people having both skillsets, decent programming and market understanding.
I haven't had the pleasure (?) to have a programmer working for me yet.
But if I will employ one in the future, only for non-critical parts of the machine… I wouldn't want to miss out making all the beautiful mistakes by my own
Maybe you enjoy this podcast episode:
chatwithtraders.com
I haven't had the pleasure (?) to have a programmer working for me yet.
But if I will employ one in the future, only for non-critical parts of the machine… I wouldn't want to miss out making all the beautiful mistakes by my own
Maybe you enjoy this podcast episode:
213 · An Obsession with Edge, Powered by Automation w/ John Roberson
John Roberson’s a proprietary trader at KTG (Texas), who begun exploiting one “tiny edge” in 2007, and today, runs a portfolio of 80+ automated strategies. Listen, as John describes the techniques he uses to maximize market anomalies...
chatwithtraders.com
Interesting comment on Quantopian - if this 200M$ VC money no issue had to close as users could not program anything useful, there is little chance you will be able. Strategy development require to pass >5 years learning curve. The comment that fully automated is a retail dream is correct, in the past - I am currently testing a new non-programming platform which I think will completely change this space.
Similar threads
