My stab at an automated trading system

I am also running automated trading systems with EOD data using Java as software language. However, instead of changing position sizes at market open I do it shortly before market close. I have tried understanding Python but for some reason I find Java language easier to read and write.
Good luck with your endeavors.
 
I recommend reading Foundations of Trading by Howard Bandy. I just finished it and it was a real eye opener for me. It is a cheap book, to boot.
 
HobbyTrading said:
I am also running automated trading systems with EOD data using Java as software language. However, instead of changing position sizes at market open I do it shortly before market close. I have tried understanding Python but for some reason I find Java language easier to read and write.
Good luck with your endeavors.
More...
Thanks HobbyTrading. Just so I understand it right, do you mean you generate signals to trade based on EOD from the previous trade day, and you buy the shares the next day just before close? Or are you just talking about rebalancing position sizes?
 
All algos
Bigchazza said:
Anyone heard of Talend? It's an open-source java-based data manipulation tool and I'm using it to build a trading system. I have messed around with python for a couple of months but just don't click with it, especially when I know how easy it is to code EXACTLY what I want from scratch. Also I'm on a steep learning curve with trading and investment in general, so forcing myself to code technical indicators from scratch helps me to undertand them.

BACKGROUND: UK-based, I have worked in IT for many years, mostly on the data side. I got interested in the stock markets a few months ago, realising that property wasn't a great investment option any more (here in the UK, due to tax changes). Whether or not stocks are a better choice I can't say for sure, but the data-driven nature of it makes it endlessly appealing to me. I have read a few books on trading, listened to loads of podcasts and just soaked up as much as I can, and I'm pretty sure that I am not by nature a discretionary trader. I can look at charts but I'd rather look at the data behind them. And if I'm doing that, why not automate it? I love robots!

GOAL: a system running my chosen strategies in the cloud, popping off trades while I'm walking the dog or doing my day job. I'm looking at EOD data initially so it will probably only be popping off those trades when the market opens (LSE initially although I intend to trade US markets once I fully understand the tax side). Some features I am aiming for:
  • Decent backtesting capability (EOD data)
  • Fully configurable (meaning if I want to try a new strategy I just plug the parameters into an interface. Job done. A strategy is just parameters)
  • Modular (i.e. separation of concerns, service-oriented, loosely-coupled... whatever you want to call it. E.g. the regime filter is a callable service that just takes a date, a market and a filter type and returns BULL or BEAR - it doesn't care about the thing that calls it)
  • As cheap as possible (cloud, data, brokerage costs)
  • Frequency: daily (I'd like to look at intraday eventually but let's walk first)
WHY I AM DOING THIS: because I need a focus. I get obsessed with projects. And I go deep, so I might as well be obsessed with something from which I can expect $$$. Also I'm not getting any younger, and the more I look at my private pension funds and the people responsible for managing them, the more I feel I should take that responsibility myself.

WHY I AM POSTING ABOUT IT HERE:
  • As a way of talking through ideas with myself
  • In the hope that someone will point out any crazy misconceptions I have
  • Who knows, someone out there might be vaguely interested
I'll be sharing more details about the system soon. It's pretty much built as a proof of concept.
More...

All algos run on 2 forms of information: volume (of buyers and sellers), which is an independent variable, and resulting price, which is a dependent variable that follows the law of supply and demand. This information is generally considered accurate and is available to the world. To go from a algo to a strategy you need a 3rd information vector based on time, which is most often derived from the past data (which is perfect knowledge) such as channels (Keltner, Donchian), horizontals (daily pivots, highs, lows, etc.) curve-linear (Fibo lines) or dynamics (MACD, RSI). It is use of this plurality of data that leads to algo strategies.

One other thought: back testing with Ninja 8 is still corrupt; best use Trade Station Easy Language programs – excellent back testing. The many thousands of programs I mentioned are written for TS; Ninja sometimes.
 
Last edited:
I've been working on my own tool for a while now, whenever I have time. Building a good looking frontend is time consuming as is the backend. Multiple data providers, fast charting, tabular analysis, custom scripting syntax, not to mention the backtesting engine with a million traps to avoid. I'm going to pull the order management from my autotrader, which took a long time to create by itself.

For me, I dramatically underestimated the time required. If you want something to go into production soon, I'd suggest 3rd party. If you don't mind building it for 3-4 years or more to get anything barely usable, build your own.
 
Bigchazza said:
Thanks HobbyTrading. Just so I understand it right, do you mean you generate signals to trade based on EOD from the previous trade day, and you buy the shares the next day just before close?
More...
That would be "the conventional way". I do it slightly different. Suppose I have a strategy which takes the latest three EOD prices into account. And today is Wednesday. I would run the review at about 10 minutes before the Wednesday EOD and get the current price at that time. Combine it with the EOD prices of Monday and Tuesday. And place my orders still before Wednesday ends. The assumption in this approach is that not much will change in those last 10 minutes, and thus that the current price can be taken to be a good approximation of the imminent EOD. The benefit of this approach is that you avoid any overnight price gap, which you would be exposed to if you'd place your order on Thursday.
 
cafeole said:
I recommend reading Foundations of Trading by Howard Bandy. I just finished it and it was a real eye opener for me. It is a cheap book, to boot.
More...
Hi cafeole on your rec I just listened to a podcast interview with him - seems worth paying attention to so I'll get that book. Thanks.
 
Bigchazza said:
Hi cafeole on your rec I just listened to a podcast interview with him - seems worth paying attention to so I'll get that book. Thanks.
More...

ValeryN has an automated trading journal and recommended Bandy. This book is an abbreviated version of his Quantitative Technical Analysis book.
 
You must log in or register to reply here.
Back
Top