The more conscientious the trader, the better the luck seems to be (I know that is a bit trite...so bear with me a minute). We have found that when traders have some "luck" doing a certain strategy, that they try to stick to it (with the misconsception that what happened was pure skill)...., and stick to it too long....ending up in disaster. When you combine the basic skills that do work, and work well (tape reading, premium/discount enty/exits, risk control, etc.) with the current market conditions (hopefully resulting from the trader reading the market correctly, which is something that takes some real objectivity) ...you will have a good chance for success. Newer tactics (I haven't seen anything truly new for a long time) will have a better chance to work only if combined with the basics. Ideally, we would like to be lucky AND good.
DT, what happens if you take one system tested on historical data then you use that system on out-of-sample data and find very similar results as far as % winners, average profit, risk factors, etc. Then you take that same system and test it on completed unrelated markets and produce similar numbers? Would you then say this system is based on luck?
dottom: Of course, it's recommended to test systems on many non-correlated markets. It's difficult to find out - whether my trading results come from skill or luck? Trading is not a game with known probabilities. In games like roulette, where there're no positive expectancy systems your profits depends only on luck. You can lose less, if you apply some money management strategies. So, you can limit losses with some skills. In games where the best systems can have neutral expectancy - systems results depends on luck. Some money management strategies (skills) can only limit your losses ( and profits, too ). In games where there're positive expectancy systems, your results depends on skills <b>and</b> luck. The higher expected profit system has, the lower impact of luck factor is. But how to account "system's expected P/L"? Market behavior is changing. You'll never know whether a market will move randomly in the future. You can't win in a random market. Trading can be compared to a following game: Screen can display only two colors: green or red. You place your bets. If you predict the next color - you win your bet, if you don't - you lose your bet. Nobody knows what's the probability of displaying each color. We can only account probabilities based on historical data. So, historical system's result is our expected result in the future. This method is not perfect, but IMO has some logic. We will never know how close our calculated expected P/L is to it's *real* value. DT-waw
This is a re-phrase of the random walk theory.... which I believe is bogus. As long as human nature is it is, there will be greed/fear/trends/ranges in the market and a system reflects such human nature and well-tested for expectancy/risk tolerance should work in the long term.
Yeah, dozu but what can happen if someday majority of money in the market will be traded by automatic systems, not by humans?
That day is still far far away, and there is still time for us to make a buck, and I am sure experienced system developers have seen a clear edge and are confident enough to trade them... If you see a system making money 10 months out of 12 for 5 straight years, you can be pretty sure it's not random.. And you don't need to have it working on multiple markets necessarily.... a trend follower will do well with Japanese Yen, but will suffer with Canadian Dollar, as the two markets are so different. When that day comes when the markets are dominated by robots, I will just invest in a mutual fund run by the smartest one.... for e.g. if the IBM Deep Blue team ever develops a robot, I will give them all my dough to trade
Well mate I would start looking at those funds... program trading and index arbitrage now accounts for 30% of all volume at the nyse. IMHO it is pretty much responsible for TA no longer working on the qqq, es etc. Mean reverting is the big boys' game - forget trend following - this is just retail hype to keep you going.
DT, when you say "luck" I read "probability". If I have a positive expectancy system, and I generate 1000 trades in one year, I have a high degree of confidence that the results reflect the profitability of the system itself, and luck is a pretty much a non-factor. The big question for system traders is will that system continue to work in the future? The truth is no one every knows, but the best systems are ones that where the trader understands the underlying rationale for every element of their system. I am a systems trader, but I will not blindly follow a system regardless of historical & out-of-sample testing. I need to understand what the system represents. What element of fear/greed/institutional buying/market sentiment/volatility is the system trying to capture? Then when I have a drawdown or bad year I can look to determine if the results are more likely due to probability distribution or is there something fundamental about the markets that have changed which affects my model. In reality, though, proof is in the profits. If my system is consistently profitable month-after-month, with only a couple down months in a year, then I'll keep riding it out. Maybe one day the system will stop working and I will lose 6 months in a row. Hopefully I will have several years worth of profits by then. Hopefully if the system stops working one can identify the fundamental changes and make adjustments. Think of systems trading as it as a trailing stop loss in a trending market. You identify a working system (trend). Now ride it...
DT-waw, despite the many excellent arguments posted on the 18 pages of this thread between your first post 2 months ago and today's posts, you have not changed your thinking, which of itself is not necessarily a bad thing. Nevertheless, I would like to leave you some points to ponder on for the next 2 months. (1) Do you really want to trade? I suspect the answer is no. (2) What exactly are you looking for? Assume Dottom is magnanimous enough to share his system with you. You do back-testing, forward-testing, intra-day testing, intra-market testing, intra-venous testing, intergalactic testing and find that the system has a statistically positive expectation with low variance - the mother of all systems. Which of the following would be your reaction? (i) Unbridled exhilaration accompanied by loss of bladder control (ie. you piss yourself with joy) (ii) You reject the findings outright, on the grounds that there is no guarantee the results can be duplicated in the future. (3) Exactly the same as (2) above with Dottom pointing his finger at a section of the graph 6 months in the past, looking at you and saying "Had you started my system here, you would have been up XXX today". (4) Let's talk about roulette, which you use as an example of a negative expectation game. Have you read "The Eudaemonic Pie" by Thomas A. Bass, about a bunch of computer wizards who tried to win at roulette? They went to Las Vegas and played roulette, in disguise, with computers in their shoes. Using physics, a computer can actually predict, fairly roughly, where the little ball will land. And the person with the computer in his shoe bets on the numbers all around that predicted number, and gains a fairly strong advantage. Where there's a will, there's an advantage. (5) Expend more energy on positive simulations and do some trading - you're too smart a guy not to find an advantage.
There are many other ways to trade besides "dumb man's trend following". Even within the general trend following wisdom, there are many ways to trade a trend. Following the trend works. It all boils down to determining when the trend is likely to resume. The Turtle techniques are what I call "dumb man's trend following". Jump on everything hoping that it trends. Profitable over time but big drawdowns in between. You can add a variety of different methods to identify when to take a trade in direction of the trend to increase your probability. Trend following seems to have worked pretty well shorting ENE. In fact, trend following strategies have done quite well in interest rates & currencies as well just to name two sectors But even for the overall markets, and specifically S&P and NQ, trend following strategies have done well. I'm not referring to buy breakout of X bars (which itself was profitable), there are better trend following strategies. For example, entering on breakout of congestion, entering during congestion with low risk entry, entering on momentum shift on pullback, correlating the last 3 with volume patterns, etc.