@Spearhead when I agreed on trading being learnt, I meant that you can pick a few books and learn about support/resistance, setups, pattern formation etc and give it a shot thinking that the good old days are just right round the corner. However, the reality of being consistently successful is a very different one. If you look at my previous posts, you will see that I am with you on. I am on the same boat as you on your second paragraph. Your 3rd paragraph is what I'm having an issue with. I have 2 sets of strategies; the first is a group of strategies that I think I can make functional by learning basic - intermediate coding. The second set actually consists of on one single strategy which has nothing to do with setups etc. That is the one that I would count on to make very good returns. However, it is not something that I would want to give to anyone to code for me, yet I do not have that trusted someone who could get the job done.
the situation you are proposing is not similar to a programmer joining a mayor firm for a fixed salary, it is more like joining a startup, where there is no warrantied income, but there is unlimited potential. What kind of split is your friend doing (the one that has a similar arrangement)? An idea for an alternative. Hire a coder to write a simple strategy and get him to write it in a way that the signal generation and the risk management are isolated enough so that you can change them later and plug in other strategies?
maybe you should checkout Ernie Chang's book. it gives a good overview of what it takes to setup an algo trading business... including details like what kind of strategies can be effectively automated. (the book includes a few strategies with excel and matlab examples) http://www.amazon.com/Quantitative-Trading-Build-Algorithmic-Business/dp/0470284889
Firstly, I call them "simple" as there are only a few aspects that need to be covered to make them work and the coding for it would be easy enough (and your suggestion for TradeStation/NinjaTrader seems sound). In reality, categorizing them after extrapolating data and qualifying them takes a lot of work. So they are only simple, only after all the work is done. I know that they are high probability as I have traded them time and time again over a long period and they have fared well.
I sure will. Someone also PM'd me the following one and highly recommends it: http://www.amazon.com/dp/1905641796...UTF8&colid=XGJ4POPBKZA5&coliid=I1J45N7AEL17UA
why not start with the goal of coding some of the qualifying, categorizing and extrapolating of the data... that way you'll free some of your time as you start to code...
Could you guys express your opinion on automated strategy builders that use genetic programming? My unqualified opinion is that these can work if: A) the strategy displays robustness by working on different products B) the strategy shows robustness by working on different time frames ( I think P. Kauffman endorsed this.) Seems you would save a lot of time in research and development.
Nice list of well thought out questions. Rarely do you see that here on ET anymore. I mostly agree with whatever else has been said. Realize that the following opinions are from someone who doesn't believe in looking at charts or bars as a good tool for developing trading strategies. So, if that's your primary method then we might be coming from fundamentally different viewpoints. Also, I believe that new or retail traders GREATLY overestimate the value of their trading idea and underestimate the value of execution, programming, fees, relationships, capital, experience, etc. Everyone goes through this process and then eventually comes out the other side as their career progresses. I can't really recommend you do this, but maybe as a thought experiment if you were to consider hiring an experienced trader from somewhere like Getco, Jump, Infinium and conduct a phone interview they would reveal much more about their running strategies than your average retail trader. Why? Because they know most of the value is in the execution framework and that you need to know lots of details about their strategy to evaluate how it fits within you firm's risk tolerance. The value the trader brings is insight into configuration and parameterization of the strategy and the ability to adjust over time as conditions change. 1) you probably shouldn't learn a language. Too broad. Best bet is something integrated into a platform where many of the execution issues are handled for you. Like TT's ADL (for futures), whatever ORCs language is (for options) or whatever stock automation platform suits you. You likely won't pick the best one at first so you'll have to learn two. But that's OK. 2) See above. When to use what requires a decent amount of knowledge. High productivity languages (C#, Python) would be less discouraging. 3) Depends on how smart you are. Not trying to be cheeky. I've seen someone create a viable strategy knowing absolutely nothing about programming in about 10 days. Some people can never do it. 4) There are a lot of trading decisions that would be painfully complicated to automate. What you describe is fairly trivial. Someone with the right tools who knows what they are doing and has done it before a few dozen times could bang out your complicated scenario in 1 day. 2 max. Examples of painfully complicated are: a) intricate math models that require lots of calibration, b) identifying areas of market congestion, c) identifying pairs that generate (at times) free convexity, d) guessing the position of other market participants and deducing their puke points 5) This is hard. It depends on your network of programmers. The problem is that if you cannot program it is impossible to determine empirically if someone is a great programmer. So then you are forced to go with circumstantial evidence that fits well on a resume: where did they go to school? have they worked at a legit shop before (like Getco, Jump, etc). That way you'd end up hiring someone very expensive with at least a low double digit percent change they can get done what you want. Cheaper option is to look at who contributes to or has created open source projects that you use or admire and contact them. Cheapest option is to find an excellent high school kid. There's about 1 kid per 10 high schools in tech areas of the country who would suffice. Finding them requires laying a lot of groundwork and casting a wide net. 6) Most people, I have found, have a decent sense of integrity if they like you and are treated fairly. Most strategies tend to migrate from shop to shop (0+, fx basis, rolls, etc). It's generally not a big deal. 7) Not sure what your trade frequency is. maybe have an override button. 8) Except for some very narrow cases, automation doesn't free you from the monitor. Most exchanges require automated strategies to be monitored at all times. Not a big deal at first. More of a headache down the road. I view automation as an easy way to make money on the obvious stuff across lots of products. 9) Should be able to do latter, depending on your data source and strategy logic. 10) I think we are quite a ways away from a complete death of discretionary trading. Consider that even today, in lots of markets most trading is done over the phone. Even if you only consider electronic markets, humans are generally able to consider many more nuances than a computer. Algos can always take the easy, fast money over and over and over but it is very hard to program non-deterministic algos that are able to learn and make sense of different market conditions in a rational way. I'm not saying it can never happen that discretionary trading (by humans) dies completely but I'd be very surprised if it's less than 10 years out. 11) No idea. I don't trade stocks. 12) I have a very few that have run 7 years now unaltered. Most are somewhat seasonal. 13) Generally, for me, new technology and the rise of more predatory algos.
GP has a very strong tendency to overfit the data. So they can require significant fwd testing after the paramethers have been adjusted to validate the model.