What does it take to become an algorithmic trader 'hotshot'? My guess is creativity. All these people who walk on trampled paths make me wonder who set them on this track the first place. For instance why does nobody use fundamental & technical data? Why does nobody try to model behaviour? Why does nobody use AI for riskmanagement? Why does nobody use agent design for microstructures? Why aren't there loads of papers on news sentiment interpretation? etc.etc. Perhaps there are a few who work on these things, but I haven't seen many of them. Instead people look at fibonacci ratios, Elliot-Wave-Theory, Candlesticks, Indicators and whatnot. There are the chartists and value fundamentalist, but to me it doesn't make sense why people don't work together and produce testable, quantifyable ideas.
Many of the things you mention here ARE well trampled paths I'd suggest. I'd be hesitant to ever use the phrase: "why does nobody..." unless you know they don't. That's 6.5 billion+ people you know aren't doing something I'm not sure where you have been looking or what has coloured your perception - you don't say. For example, Reuters producing news feed products specifically for automated/algorithmic trading system consumption highlights the fact that semantic news analysis is well in the mainstream and not the just the preserve of multi-billion dollar investment banks: http://www.finextra.com/fullstory.asp?id=16254 For $50 a month you can subscribe to specialist services that use semantic news analysis technology to guage market sentiment in real-time and deliver you the results. In other words, that innovation ship has sailed. Furthermore, if any of the techniques you mention yields a trading edge then anyone with the requisite knowledge will most likely prefer to exploit the edge to make money rather than share and collaborate and/or write papers on the matter. It's the nature of the beast. Do you spend countless hours and $ researching how to make money in the markets in order to share it with other people or to....as was driving force in the first place...make money?
Once you share a new trading idea it takes about 3 years for it to find its way into the main stream. The market becomes more efficient and the edge disappears. The ratio between winners and losers remains constant but the barriers to entry and the 'quality needed in the next edge' increases. If you have an edge and you hope to benefit from it it makes very good sense to keep it to yourself. The only people who advocate 'sharing' if ideas and skills are those who have none of their own. Those offering to 'share their edge' are just repackaging rehashed ideas that have been published again and again. (Any exceptions here??? Don't think so.) Just a thought on a slow, cold, Saturday.
I just reviewed my post and it seems I was a little ambiguous. Reuters' new products and the sentiment-based subscription service from a company not related to Reuters are two distinct things. I mentioned both to illustrate the point that this particular technology is now pretty much commoditized. The subscription service I was referring to has a published track record with high winning % but fairly low frequency of trades. However, the efficacy of semantic news analysis has been debated before on ET with the general consensus of ET members being that it is does not work.
TradeStation will soon provide a fundamental database( don't know when... ). I was eager to develop fundamental systems as it was new to me.Why do you think it's not a good approach? What does the world "semantic" means ?
I am really ignorant about algorithmic trading but am curious: who uses Elliot Wave or candlesticks, etc. to automate their trading? I always thought that was somewhat like the astrology part of the trading world. On the other hand I thought I had heard of attempts to monitor news feeds to generate trades. I've always been under the impression that algorithmic approaches were much more probabilistic in nature. I've met only one in-bank algorithmic modeler and he was not surprisingly less than forthcoming. Any ideas?
When people say it doesn't work it doesn't mean it can not work. Yes, Reuters has these tags, but that doesn't mean much. You always have to integrate it, into a system. Tradestation already has components for fundamentals AFAIK. I think that Tradestation is not really a plattform to develop advanced models on.
Absolutely! I think EW and candles are as good as anything (tossing a coin, throwing a dart). Trading as such is probablistic. I find programm trading much more powerful, because one can backtest, scale and automate.
Yes, I would say that reading a news article is pretty tough for any computer. Even worse it's understanding it. What I'm wondering is what happens if the article is structured in a way that can be more easily understood by a computer, for example with an xml file. Not that I'm saying it would work, and I'm definitely not saying that it would still be like a person reading, but it could help the computer to get more information out of it. If you have more articles made this way, feeding for example a neural network, this might give some results.