I am looking at my poll again. Now we've got sixty votes. People investing below 50 thousand dollars could also be lucky. I am not saying it HAS to be luck, but for example I've been there twice (around 25 thousand) and went straight back to zero where I came from, due to great performance by my system, but no money management at all (I was reinvesting everything, buying as many contracts as possible and without even selecting my best systems). I am impressed with the others. There's about 40 people, two thirds of all voters, who really seem to know what they are doing. I don't think I'd get these percentages anywhere else on the web, or in any other section of this forum.
Travis - I am interested in getting into automated trading. Can you talk more about what software/broker you use, whether you developed your own systems or had help with them, what your time commitment to maintaining/monitoring the system is, etc. I thought I had found my niche selling credit spreads but after a rough patch recently I am looking for a methodology that is shorter term. And given that I have a job and other responsibilities I cant be in front of the computer all day. So I am exploring whether automated trading would fit my needs. Thanks
Oh, wow, that's great. I was just waiting for someone to ask me this question. Do get into automated trading, by all means - just look at the poll if you have any doubts. Over 33% are investing over 1 million dollars. That tells you everything. Could you do that with discretionary trading? For how long would you keep the money and for how long would you sleep at night? I use excel and Interactive Brokers. Here is the link on their API software page, for alternatives to using excel. Everything begins from here: http://www.interactivebrokers.com/en/software/highlights/apiHighlights.php?ib_entity=uk http://www.interactivebrokers.com/php/apiUsersGuide/apiguide.htm Without knowing any programming languages and having a liberal arts background, I opened an account with IB, studied some vba (a lot, actually, but I am not gifted in the field of programming), and modified their sample excel file and built, in three years, all the 30 systems that I have now. They all run smoothly - my pc doesn't crash, but it does breathe heavy - on just one excel file, at the cost of zero dollars, unlike tradestation, I think. And I can customize it all I want and very quickly, with all the advantages of excel, like live charts and so on. Not to speak of IB Demo and paper trading accounts, that allow you to test your strategies live without using real money. I must say that, even though I didn't try any other brokers, IB works really well and efficiently. So does Excel. Time commitment to monitoring/developing. That's the tough part. It took me about 10 hours a day for three years (because I also worked on it while at my regular job). Before that, I had spent three more years (same time commitment) developing trading systems on tradestation but I wasn't good enough to make them automated (nor were they good strategies - curve fitting and over-optimizing). Even before that, I had spent another 4 years losing money with discretionary trading (options, stocks). But now I don't have much left to do. Actually, I need to stay away from it or I'll interfere with my system and blow out my account as usual (it happened over 30 times in the last ten years - always starting with a very small account though). Now I would say I need 10 minutes a day. But of course you could go on developing it forever, changing the colors of cells, and customizing it infinitely (and probably I will). How much time would I need? 10 minutes. How much time do I spend on it? 3 hours day on average, right now. Help from others? Yes. Absolutely. It was impossible to find real people I knew in real life who knew, cared about trading systems or even knew about trading. So I looked on the internet. While playing the Risk strategy game online, I met one guy who's helped me a lot for years with programming and configuring various settings on tradestation and excel and other programs. That's the first person I have to thank. The second group of people is all the people who helped me at these two forums: one is Elite Trader and the other one is a similar Italian forum. Other than that, it's all my hard - very driven - work. I was really bad at math, but the desire to make money has made me study a lot of things I would have never imagined.
Also consider metatrader and tradestation. Both have very extensive support forums, deep and rich with information and examples. Both of them also have extensive back-testing and optimization features. We did lots and lots of backtesting with tradestation and used add-on genetic optimizers. Additionally the transition to live trading is easy for both along with backfilling your strategies so they can transition to live trading. If you are technically adept, then the API route is good too. But from a starting standpoint, metatrader and tradestation shorten the learning curve significantly. Plus a really cheap and worth every penny option is to take the tradestation courses. They are amazingly cheap for the extensive training they give you. I've taken those courses and even brought my 10 year old son along to learn programming. Worth every minute and every penny. Even if you don't end up trading through tradestation, the platform is one of the easiest and fastest to develop your ideas on. We used the tradestation platform, then executed elsewhere. What I did was deposit $5000 at tradestation to open an account, and never traded it. But by opening the account, it made the platform only $99 per month. Of course Mt4 is free, but even if you use Mt4, do the tradestation training course. They teach you the basics of backtesting, execution, and strategy development. Trader 5of7 @ TheCollectiveFX.com
There are nuances to automated order execution. Tradestation markets their product as an all-in-one software tool. Backtest, optimize, and automate. Using only tradestation to develop and deploy your ATS is woefully inadequate IMO. TS is absolutely terrific for constructing a system model quickly, that mostly reflects what can be achieved by live trading. However, the backtest results don't reflect daily mark-to-market of the account, which is a fundamental flaw. The performance statistics are based on a trade-by-trade evaluation instead. Some metrics, such as net profit, are not sensitive, obviously. other metrics, such as profit factor, Sharpe Ratio, etc. are very much sensitive as the mark-to-market frequency will directly influence the result. TS should base its backtest calculations on an absolute standard, which by definition is daily mark-to-market. The other inadequacy of the backtest is the lack of walk-forward optimization. I had to build a tool in C++ to do this, and the code is rather lengthy. Without WFO, the trader is just gambling IMO. Order execution is a BIG problem for an all-in-one approach. There are various order types that can leave live market positions discrepant from simulated market positions. This is the result of the faulty fill assumptions of the simulator, primarily the "fill if touched" assumption. Because of this, there are only crude ways of re-syncing the sim and account MP's, such as using MIT in lieu of limits, which result in nasty slippage a lot of the time. My take is that robust execution must be implemented by order execution macro commands. Consequently, a separate execution script must be carefully developed that examines actual account market position and fill histories at the close of each bar. This allows the script to issue macro commands to the broker platform independently of what the simulation is doing. As such, more creative ways of overcoming fill limitations can be implemented, such as order stacking, which acts as a great statistical filter that smoothes out fill distributions. I think TS gets it, because they have a range of order entry macros that you can use to create your own execution script. Ninjatrader also provides a dynamic link library so that NT macros can communicate with various broker platforms through the API. I just wish NT would provide these same macros for use in ninjascript, so that one could avoid using the TS data feed altogether for live trading. rt
I don't get it. Do you believe this holds true for all automated systems, no matter the timeframe and no matter how simple they are? Do you find that many people agree with you? Premise: I don't know exactly what walk forward optimisation is, but for this very reason your statement made me worry, since obviously I am not using it. Say for example I have a system that goes short every friday night and sells every monday morning? Should I also worry about walk-forward optimisation? I would say no, but what do you think? I would say no, because I don't see in what way I could ever change and optimise such a system (and similar systems).
There are basically two walk-forwards. One is random walk-forward testing. The other is walk-forward re-optimization. Both are based on strategies that are optimized. ie. You set stops and targets based on the range of the market which is based on optimizing for a period in the past. Other things that can be optimized are things like the moving average lengths, etc. When you optimize you run the risk of being "curve-fit", in other words those settings will only work on exact same identical conditions in the future. Such strategies are considered "less robust". To avoid this, you would take multiple periods in the past (which your optimization didn't see), and test your settings on them. If your strategy still performs well, your strategy would be considered "robust" as it passed multiple walk-forward testing. A robust strategy doesn't need as frequent re-optimization. Walk-Forward Re-Optimization is another approach. This operates under the assumption that the near future conditions will be similar to the recent past. So you would optimize your settings for the most recent week (for example) then use those settings for the next week. Then next week, you do it again, constantly adjusting. A third approach not mentioned is Strategy Filtering. This is essentially watching closely the performance of your strategy, and plotting the equity curve. If the strategy is making money, you continue to let it trade with live money. If it begins to perform badly, you "switch it off" and put it into simulation mode only until it begins performing well again. We are strictly autotraders (system traders) who have been doing this for a very, very long time and have tried all these methods. We have found the best approach is to NOT put your eggs in one basket and find yourself as many strategies (EA's, systems, etc.) as you possibly can. The reason is that no system on the planet does well in all conditions. So find yourself 20! Then find the settings that are the most robust. Next treat each EA like an employee, and watch their performance, and only let the good performing EA's have live money. By spreading your account over 20 EA's it forces you to use better money management, and also smooths out your performance, and reduces your account drawdowns as often when some EA's are starting to do badly, others will be perking up and make up for the losses. Google all the terms I mentioned, and learn as much as you can, as there are lots of techniques for all these. But rest assured, that NO strategy can perform well all the time, BUT a PORTFOLIO of strategies CAN do well most of the time. The next level is to watch the performance of your PORTFOLIO. Then you graduate from just being a trader to being a fund manager, managing your own fund (which is your account)! Good luck, keep learning, and don't worry. Just don't over-leverage your account on a single strategy (EA) : - ) Trader 5of7 @ TheCollectiveFX.com
RoughTrader is correct in many aspects he mentioned with respect to Tradestation. Clearly someone who has been around : - ) Consider also that some of the issues go away if you trade Forex. In fact we switched to Forex to eliminate many of those issues. Forex gives us near bottomless liquidity, and market orders that fill 100% of the time until you get into very large sizes. Given essentially no slippage, and near perfect execution, we have found that our assumptions do bear out in live trading. But RoughTrader is correct in that you can't make those assumptions with futures and stocks. Certainly not options. Here's a futures example: FUTURES: - 1 Tick Spread (Difference b/t bid & ask) costs you $10 per trade. - Occasional slippage of 1/2 Tick costs you $5 per trade. - Commission of $5 per trade. = Net Cost of $20 per trade you take that must be deducted from your trade profit. FOREX - 0-.5 Pip Spread (costs you nothing to $5 per trade.) Let's say $5. - No slippage until you get to very large orders. - No commission. = Net Cost of $5 per trade you take that must be deducted from your trade profit. So you can see that with futures there are more considerations that must be made, and even more so depending on the liquidity of the market, and time of day that you are trading. I'm not making the case for Forex, I am making the case for what RoughTrader said. Keep in mind the considerations when transitioning to live trading. But as for learning, developing and getting a feel for system/autotrading, nothing beats Tradestation. We did follow RoughRider's route eventually, since once you learn the ropes. you will begin to feel Tradestation's limitations. Unfortunately we too ended up custom coding our entire system with our own internal statistics and management, order execution and oversight, etc. but but but, there are people who make money using Tradestation alone autotrading, so this isn't to say you MUST eventually custom code your own platform. It IS the defacto learning tool for sure. When you get to the genetic optimization, walk-forward, swarm optimization, portfolio optimization, fancy stuff, there are lots of add-ons for tradestation to do it all. In the end RoughRider is correct, there is some learning to be done before you jump to live trading. Trader 5of7 @ TheCollectiveFX.com
Thank you very much for the detailed explanation. I have partly already implemented your ideas, and what you said confirmed to me that I am not the right path. One important idea you explained that I am totally following is the multiple systems - I have 30. Another part I liked is that the fewer things to optimize, the better is the system. I think I have no parameters, to optimise. I learned this because I used to be a big overoptimiser, because I wanted my system to never make a mistake. The part I am not yet understanding is how you should or just could (possibly, as an option) trade only when you are having a sequence of positive trades and stop when you start having a few bad trades. Do you believe that your system's equity line has trends comparable to the ones in the markets? It sounds as if one could trade his own system with a moving average bullish crossover, but then again does that method work in the markets themselves? What about ranges in the equity line? It gets so complicated that it seems better to me to set it aside and only worry about having a bunch of good systems as we said earlier.