The Trade-Robots Situation Update Christmas 2014
A regular update is something that any blog should have – especially one that has been fallen to inactivity in the recent path. There are many reasons for this,. But they all run around a limited time budget as we head into end of year frenzy with some consulting work and focus on our trading agent. Things are moving. And here are the details.
Time is limited – and we are super busy
There is little time until end of the year, and a lot of work to be done. We keep developing strategies – but our core programming capacity (that is me) is currently out working for a customer in a nice consulting contract. While this will run until end of March 2015, or even a little longer, pre-Christmas that was just a little – chaotic, also together with some private elements and, as I have to confess, the release of the awesome “Dragon Age: Inquisition” game… that sadly also had an impact on time. Being the author of most blog posts I also was waiting to see what happens. Having recently (like in the last days) taken delivery of a nice Surface 3 Pro tablet (well, sort of tablet) I am now, in the nice situation to also be able to just write something on the road or in a plane. Blog posts will keep coming now – and whoever looks for a good mobile device, the Surface Pro 3 from Microsoft is a really interesting Laptop in a Tablet form factor.
Our new Trade Agent is coming along
Slower than hoped for, but we are making progress on the trading agent front. I decided to skip the interim step of a desktop based application and have moved straight to our web based server system – web based in so far as we will use a website for managing strategies…. Obviously not for trading.
We are handling some last critical gaps and design issues in our core framework that are now popping up – once we actually look at the whole “complex server to run hundred strategies at a time” setup. Among them the lack of a standardized symbology. Work on that will likely be finished in the next days and then “the data can flow”. Our current status is that the new server can connect to providers and can load strategies, but we have still issues in our data broker (distributing the data to strategies) and order handling is not tested.
And while we are at it, the new symbology will close some functional gaps in instrument handling that just happen to get in the way of trading FOREX accounts. For example, having a pure FOREX background, we currently go through a front month determination for every symbol – totally unneeded for forex. A blog post about the Reflexo Instrument Symbology will soon be released – sadly there is no decent industry standard to just use.
Our Forex account is open and ready for testing
And it is with Oanda. There are two main reasons here. First, FOREX brokers with easy to access API are still not that common – and Oanda offers a HTTP based free API. This allows us to actually develop against them without a cost. Second, and a lot more important – Oanda allows arbitrary position sizes down to 1 base unit. So, I can trade 1 USD. Why is this critical? Because there is a large step missing in our Trading Agent development setup – the possibility to test the core logic of our risk management and order handling without risking a lot of money. Trading Futures happens in significant amounts as smallest possible denominator – and it is not fun to realize the new software got into a loop and closed a position 122 times (as seen in NinjaTrader). As such, Oanda will be actually the first broker we use for real trading in beginning of 2015 with the new framework – because it allows us to start with a small 50 EUR account and do infrastructure testing on this small amounts.
On the strategy front – we broke through 300 setups
Amazing how much more speed we have gained since we fixed the bad data bug that came and hit us hard in the august timeframe. We just broke the 300 optimized and validated setup count, although obviously most of those setups have a very low statistical trust value assigned. They simple have not been observed long enough. With all the new stuff we have in the pipeline, though, there is a high probability that we will keep the count in the 300 setup range, even if a lot of them get invalidated in the next months.
As explanation: we do not consider a setup to be something we trade for real unless we have some months of good real time simulated experience (a practice we may change with Forex because we can just put a setup to start with a super small value). And given that most of those setups that made it into the validation stage started to come around in the last 3 months… most of them just lack the age to be taken “too serious”. We expect to invalidate a good part of them, but to also replace them with new strategies at a same or growing speed.
Over the holidays, we are on holidays.
And that is really how it is. We are in the process of winding down our back testing operation – no sense in paying all the electricity when there are no people to keep the new strategies flowing. We are shutting down all activities, just leaving critical infrastructure services running (email, this website, for example) and collecting market data. People are on holidays – your truly blog post author is in on the way to Belek in Turkey now where I will enjoy 2 weeks of intensive golfing, food and beverages and some time for working on our core code. Because, seriously, I cannot really turn into non working mode for 2 weeks.
After the holidays, the goal is Trading
After the holidays, our main and only goal is to give our new Trading Agent a chance to take over. Depending on the state of affairs we will start deploying it in early to mid-January towards a test with OANDA. It will take us a month or two to get the issues out of the system… but then we will start replacing our current infrastructure with the new trading agent. 2014 was an interesting year, and I would really love to start 2015 based on our new platform capabilities, even if that means delaying the trading start for the next year by a month.
From there we will move as fast as possible to integrate publishing of trading results for a portfolio of strategies into Collective 2. This likely will happen in March / April timeframe. Rather safe than sorry with timeframes after the experience of 2014, where one programming delay piled up on top of the next delay.
But more details on that will be published on the go, while we are working through the quirks of our new trading infrastructure.