Structural edge. I have never seen or heard of a algo telling the firm to buy an exchange membership or implementing a new technology (eg. AI-based News Sematics Analysis) to expoit a specific edge. ............................................................. - In terms of statistical edge, computers are faster finding statistical "edges" than humans... (I don't consider conventional back-tested models as having an edge) AI, NN, GP, AIS... If there is data and if it involves numbers... computers are faster and sophisticated... True edges involves the financial industry themselves where numbers and data can't get involved or quantified... ( In terms of fundamental analysis, computers are better... Humans tend to hit more home runs... algo-based fundamental analysis tends to be more consistant. If you want to make money, algo-based fundamental analysis does more... Fortunately, the industry is turning risk emphasized more than ever... Analysts are going to get killed. Give it a few more years... Analysts = Marketers... )
It does not matter how the computers will trade. It still will be a 50-50 game. It is a sum zero game right. Some computers win some lose. Same thing will go for humans we beat a computer we lose to a computer. and on and on and on. And you can never forget those stupid peple who just get lucky. Like that #7 place lady in the cnbc competition. She is a god damn waitress for crist sake, I would be surprised if she has any knowledge what so ever about markets, balance sheets, TA, etc etc etc. ( I am not trying to say that she is stupid just stating that she got lucky)
A looong time ago (in computer development time), a computer beat the best human player in chess. Yes, chess is different in that it has a fixed amount of variables, but aside from qualitative data, the amount of quantitative data that a human being can process for decision taking in short term trading, is not that big either. In other words, short term human traders, don´t use that much more data than would be available on a chess board anyway.. in my opinion.
As a programming problem, chess is completely different from trading. Chess has a fixed playing field, fixed inputs and a fixed set of rules. With enough processing power a computer can just apply brute force to look-ahead at every possible move it could make and counter-move its opponent could make in response to determine which is its best next move. Add in some pruning logic and standard openings and you have a world-class computer player. Trading is nothing like that. There are no hard and fast rules. Many inputs cannot be easily described/input to a program (breaking news). What works one day can fail the next. Simply applying brute force computing power does not insure success for a trading program.
what i said was: a short term trader, who doesn´t use qualitative input (fundamental analysis for instance), but just uses the order book, some leading indicators, historic data for his instrument and his leading indicators (charts) is using fixed inputs as well. therefore, this game this trader is playing, is very much like a little bit more complex chess game.