Project: Samurai - Blog

*DISCLAIMER* - This blog will be written very informally, and there probably won't be much analysis in the very beginning.

The purpose of this blog is to catalog a new project that I'm working. The project concerns a new automated trading methodology inspired by a few ramblings by members on this board.

The project requires a few phases. The first phase is the application construction. The second phase will be the analysis period. The third phase will be the actual trading phase. The final phase will be the retirement phase.

The trading methodology will vary because the basis of the system relies on probabilities, not really any particular system. In fact, the actual trading system may vary, and several trading systems may be explored. What will be discussed at some point is how we get from the third phase to the final phase.

The program itself will be monitored and most of the project will be focused on the daily results. Trades will not be posted in real-time.

Just to be sure, the blog, at first will be mainly for myself. I do plan to eventually make the blog more helpful, but bear with me folks, only the outline of the system has been developed.

Today is the second day of the project, and we're moving along.

The Quote service that will feed the multicast queue is finally up and running after some difficult debugging and web searching. Once the problem has been resolved, I suspect we will have a ready service that will broadcast the quotes where the other parts of the application will pick them up.

Once the Quote service is complete, we'll move onto the databasing service that will store the information in a database. A secondary service will maintain the database, cleaning up and compressing records on a daily basis.

Once this is complete, the first full phase will be underway as we collect the necessary data to do analysis. Obviously, analysis will be continuous, but we should be able to collect enough information in the first few weeks to determine whether or not this application will be viable.

While the first phase is underway, we will start on the building of the management components and the robot itself.

The quote service should be completed by tomorrow and the database service will be operational the next day. We can start databasing quotes as early as Wednesday if all goes well.

As promised, the Quote Multicast service works like a charm. So I'm sending quotes across a multicast queue and another program is picking up the quotes and now databasing. So also as promised, the quote databasing service is now operational.

Apparently, I logged in and started testing out the database right in time. Within just a few minutes of testing, my database managed to capture over 2000 ticks. When I double checked the items, I noticed the momentum was just insane. The quotes managed to move up and down about 50 pips on the EUR/USD. I have no idea what the impetus was.

I did notice that MBT actually offers negative spreads during these times of high volatility. There was a period of at least 4 seconds where the spread managed to be < 0 with a maximum negative spread of 6 pips! What I also noticed is that only one side of the bid/ask was moving (the ask) while the bid stood in place. I also happened to notice that once the bid caught up with the ask, it jumped a good deal lower, and then continued to follow the ask downwards...makes you think...

The volatility is excellent, and I have a good feeling that my trading strategy will work with the robot. My main concern are the fail-safes. The system relies heavily on the ability of order server to collect the orders and make sure that the spread doesn't go out of line. With that said, I do believe I can make some mathematical adjustments so that we don't necessarily have to be strict with the boundaries (more on this later...just making an observation for myself).

I managed to make a little more progress finishing another important service in the set of applications that will be run.

I have completed a logging service that will track error messages from the various applications.

There are 2 more pieces to complete before I can let phase 1 go into play: The maintenance service that will compile the ticks and zip them onto the server for me and the management services link into all the various apps. I am trying to get all of this done by tomorrow so I can let it run on Friday and figure out if there is anything else that might need to be done over the weekend. The idea is to have full quote databasing and archiving starting on Sunday.

During the 2 weeks after that, I'll be working on completing the robot and the various administration applications.

Again, I'd like to start trading this project live by Q4, but I may wait until mid-October to release so I can track a full month of quotes to find out how wild everything gets.

No progress today. Life happens.

Minimal progress was made over the weekend. I flew out to Florida for a much needed vacation. I started on the Maintenance service on the flights, but I was having a hard time figuring out how I wanted to make it happen and once I finally did figure out how I wanted the service to be architected, it was time to get off a plane. So I did make some progress, but not enough to consider phase 1 done.

I have commitments today that may prevent me from completing the Maintenance service, so we'll see. Definitely by the end of the week, I'd like to get it up and running. I will definitely push back a go-live date until October 15. That should give me a few volatile days to analyze.