The perceived issue in your PS is valid, but his point is the opposite of what you think and actually reinforcing your point. One may implement pattern recognition of varying quality or use ML, but the fact is many actors in the market have already taken that to extremes. So where's the edge there? The bar is high. When I started out, I was super excited about mapping all these patterns I was seeing in charts to computer software. Sadly I was several decades too late with my IQ/maths competency* **. I certainly wasted more than a few months. *As a counter point, it's of course super easy to write a generic trend following or mean reversion algo, whether it will be profitable or not in a particular time slice depends on exogenous variables though. **I don't rule out that one day I will actually be trading an automated system that lasts longer than a few months, but if so it will come about as a result of what I've learned from my discretionary trading, and price certainly will not be the only input.
I don't agree. My experience is that chart patterns can become extremely difficult/complicated. To judge if something it difficult or not you should know everything about that subject. So in this case the question is: do you know all potential happening patterns? And that question can never be answered because nobody knows if the knows all potential chart patterns. I encountered very difficult patterns in my journey to become a daytrader. I am figthing with them for decades, and bit by bit I get closer to managing (some of) them correctly. After more than 20 years of studying I even found new patterns that I did not know before. Often I ask myself: Maybe these pattern are not difficult and the problem is that I am too stupid. But again I cannot answer that question as I don't know if these patterns are maybe not difficult for other people... Discussions between people are strongly influenced by the knowledge, or lack of knowledge each participant has. From their personal point of view they can be both right, while at the same time in reality one is wrong. A says you can open this box, and B says you cannot. Who is right? It all about knowledge. A lacks knowledge that B has. So A is wrong while he thinks he is right. Many discussions on ET are in a similar situation.
I'm surprised by some of the comments here. Computerized pattern recognition have been around for a long time. In Norway, we have an advisory firm called InvestTech which was founded in 1997 originating from a reseach project at the University of Oslo using AI a few years prior. They utilize a fully algorithmic/systematic approach where classical TA patterns comprise the bulk of their models in addition to seasonal studies and insider trading (if someone in upper management or members of the board buys or sells stock). Everything is fully automated. Disclaimer: I never used any of their services, nor will I make any comments on the efficacy of their approach. Triangles, wedges, support and resistance, head and shoulders, etc., it's all there fully automated. And since it's fully automated they can and run their models on 30.000 (!) stocks, currencies and indices. Here's their recent charts for S&P 500. Short term (buy): Medium term (buy): Long term (buy):
I measure the validity of any system on the returns and drawdowns they have. There are thousands of systems that generate profits. Some people feel they are very good because they make 10% a year. While for others making 50% per year is not good enough. What are the numbers for Investtech the last 10 years? That's the essential thing. All the rest doesn't matter. I see they sell subscriptions, but nowhere any information about performance. Performance ise the only reason why people would subscribe.
The difficulty grows exponential in relation to the shrinking of the timeframe. virtusa..... yep.... i tried to automate my daytrading for a little over 10 years. i ended up breaking the code into 2 pieces... setup identification and entry/trade management. i manually identify setups and still use the entry/trade management part of the automation. cheers toucan
In my prior post I specifically said that I would not comment on the efficacy of their approach. The point was to show just one firm which implements computerized pattern recognition and have been doing it for over 20 years - something several here seemed to believe was not possible in 2023. I think they have or had a fund, though. I can see if I find anything on performance.
Here's an article from Finansavisen in 2021. Investtech Invest crushes the stock market with technical analysis | Finansavisen (www-finansavisen-no.translate.goog)
I have very long term commodities system automation, it takes many trades. Month after month it eks out small percentages, but when all markets are trending hard and thereafter, system does well. Have 2 scalping methods automated but doesn't take many trades in 24 hours, risk are ticks and profit same on 80% of position, other 20% shoots for larger profits. I could increase trades on one of them, profits be larger but drawdown triples up. At my age, don't want hassle. Linda Raschke is old school, hands on trader, doubt she done much programming.
Sure. I have it. Here is the public demo on ET I have been running for years while trading investors' accounts: https://www.elitetrader.com/et/thre...st-time-decay-and-market-drift.364633/page-51 (Free unlimited trial/test. Only for capitalized investors). [ Consistently profitable here means that is practically guaranteed you will make a profit every year with very bearable DD (<5%). ]