Quote from syswizard:
Taking advantage of today's electronic markets is the key. But as we have observed, getting timely, consistent, accurate data into a robust trading platform that doesn't get hung-up, well that's a challenge. day now.
This is an opportunity to have the key but not the key itself. A robust and stable trading platform with this huge amount of data is very hard to have even if you develop it itself and control the things. Also what trades you decide to perform depends from the historical research you do.
I personally do not know of indicator for historical intraday data. I suppose it is not because it is algorithmically and mathematically hard to create, but because of the data problem. Firstly you have to get these intraday data from somewhere, they are very expensive and for limited amount of time. Secondly you have to have a platform to store and process them. Thirdly you have to have indicators.
The first problem is insolvable for me as I see it. Even If I can get the terabytes of data the COTS (Conventional of The Shelf) databases cannot handle this amount of data.
I am getting my EOD data from CSI.
Accurate EOD data are supposed to be easy to get but my research shows this is not the case. The same problems exist with the EOD data too. CSI up to now is the best but they also have problems- for example I mentioned sometimes the data are updated not only for the last day but up to 6 months back. If this is what you have for EOD data which are thousand times simpler I don't want to think about what it is with intraday data.
I agree with intraday there are much more opportunities but without reliable data - forget it. Not only data but platform is important too. I remember using Schwab to trade options. Options is a high frequency trading. I had an opportunity for a good trade, prepared everything for the trade. In the moment after I hit the Buy button, the speed of trading was so high that StreetSmart blocked. StreetSmart is a very fast desktop application and the problem was not in it but at the servers. Now, in this moment I do not know if the trade was executed and if executed, how. I cannot see how the underlying stock moves to anticipate how the option will behave. My preliminary calculations were for a very quick movement with a profit potential at least 50%. The problem was I don't know what happens in the moment, if the stock is against the anticipated movement or not. So I called my broker to sell all positions. After several hours I was able to see what happened. My initial analysis was right and I would be with a good profit, but as you see, the communication congestion prevented a very good trade.
Man learns every day and this accident showed me that there is something wrong. Options have a very big potential for profit. If you are close to the strike the options are moving very very fast sometimes with a light fast speed. Big potential - yes, but hardly practical for high frequency trading.
Now, I don't say I am not using options, but I am using them in a different time frame to prevent this kind of accidents. From then I began to pay attention to the data problems. A lot of research showed me that the data problem will not be solved and will get worse and worse.
I do not want to get in much details but my research shows that very soon the vendors of real time data will hit the wall. Man have to decide what to do then. Trading options definitely is most profitable. But very few have the liquidity and the regular spectrum around the strike. Which leaves a few ETF's and several stocks. Banking stocks which used to be probably the most traded with options now are in such condition I wonder they will exist anymore.
For high frequency trading of stocks - you need a lot of capital in order to get a reasonable gain in very short time frame. Also you need a liquid stocks and predicable stocks not penny stocks.
Which leads to the bottom line for the current situation - just a few ETF's and several stock options for a high frequency trading.
I am using EOD for historical analysis and intraday data from the broker to enter the trade. But how I do enter the day does not matter too much because my time frame is 5-7 days. You can get bigger movements in this time frame, not intraday. Also this approach allows me to use proprietary indicators on much bigger time frame back in the history - I have reliable EOD data and do not depend from the inherent inaccuracies of intraday data.
