- This topic is empty.
-
Don’t necessarily go by the last 5 years Tamas. Have a look at mine for the last five years
For me this is very much a diversification system to add to my other trend following systems. It works the small end of the market and has some nice years and some rubbish ones. As a stand alone system this would be hard to trade!
The last couple of weeks I have been testing on markets outside the US and Australia. The primary reason is to find an MOC system I can run outside of US hours in order to maximize my funds when they are not being used. Originally the ASX was the best place for this but I have had issues with fills and the system was not matching the backtest in reality because of this.
So I looked at other markets. My primary search was to find markets that have a decent size, so that I wouldn’t face the same issues I had on the ASX. There are really only four other markets to look at that have a decent size, Hong Kong, Tokyo, London and Euronext. Other markets like India and South Korea are both roughly the same size as ASX in market capitalisation so I ruled them out. Plus if the algos are running on the ASX they will be rampant in India.
Scott has been working on this for a while so the two of us combined forces and did some research. MetaStock seemed the best place for data, so I got a subscription and set up Amibroker on a different computer, so that I wouldn’t have any issues with my existing database.
It hasn’t been a particularly smooth journey, and the MetaStock website doesn’t have half the information you need to accomplish this, but we got there in the end, although Scott is still having issues.
I decided that to try and avoid survivorship bias I would not try and use a historical constituent stock index, but would run price and volume filters on the whole stock market, therefore the only survivorship bias would be for stocks staying solvent.
Each stock market has its own peculiarities that affect how you backtest. After a lot of research we decided to drop the Hong Kong market, as stocks there are traded in blocks of shares of differing sizes. There are 100 lots, 200, lots, 300 lots, 500, lots etc etc… and to complicate matters we couldn’t find a list that gave any consistency. Shares seem to move in and out of the lot sizes and we couldn’t find any way of knowing what they were in the past and what they would become in the future. Too hard basket! Maybe if I could find a HK stock broker I could find out how they go about it.
Tokyo is simpler. They have 1000 lots and 100 lots and by 1st October they will all be 100 lots. Simple to code. However I could not find a combination of parameters that gave me a system I would want to trade. I thought for a while I had, but I had set the minimum share price to 1 Yen which gave me 50% CAGR but in some cases I was over 20% of the market for a particular stock, and when I changed the min trade size to 50 or 100 Yen to eradicate this, the CAGR went down to 20% with -20% MDD. Not worth trading especially considering the unreliability of the data. The data has to be paramount for me, as it is intrinsically unreliable, so if I can’t find a system that will give me a really good CAGR and MDD then I won’t trade it. I have to go into this with the assumption the CAGR will be less than the backtest and the MDD will be worse.
I then started to look at LSE and there I found a good system, but it’s making all it’s money in the last two years and starts to tail off fast the further back you go. I don’t trust that.
I will look at Euronext next but I’m having a break for a few days to think things through. I thought I would document this for two reasons; one to help me consolidate my thoughts, and also to let everyone else know about it, as it is often something that we have mentioned on the forum, or on group chats.
- You must be logged in to reply to this topic.