I have built profitable trading systems and I can say that money management has very very little to do with the expected value each of trading model
I agree with your thinking. MM should just make an edge do better or prevent blowouts due to DD. I am actually thinking of trying some back testing software for futures because I have found some edges, but I need to be able to test them much faster. It is taking me 5 minutes per trades to manually test. My question is.............is it necessary to test with out of sample data? I have heard I should, but don't edges come and go as the market changes? Is it so important that your edge works in a different time frame or do you just work on a new edge when volatility changes? When you developed your system did you test with out of sample data? Does your edge hold up over many year and many trades? TIA Dan
Quote from maxpi: He says something like "it is trying to give you money all day". That is the correct view, the things about "zero sum, me against the market, me against all the other traders, I predict that....." are really forms of insanity. The insanity is believing what you said. The truth is the exact opposite to what you stated. If the market is trying to "give you money" then how come many large CTAs struggle to outperform the S&P return year after year, with leverage and large research departments at their disposal? You think that Joe Schmoe sitting at home can read a few Jack manuals and regularly reap all this money the market is trying to give you? Do you see ET littered with thousands of Jack Hershey students who are now rich?
ProgrammerGuy hasn't gotten back to the thread so I'll put in my two cents. You can verify results of backtesting by doing out of sample tests, there is nothing prohibiting you from doing that. 3D views of results are valuable as well. Multicharts has that feature, there are addons for that for Tradestation as well. The 3D view will show you peaks. You don't want to operate your system near those peaks, it means that a small change in parameters makes a big change in results and that means overly fitted to the data. Without the 3D viewer you can put your backtest results in Excel and simply sort and look. Try to find the sweet spot where changes in parameters don't affect the outcome all that much. It's not at all difficult and in fact, you will learn to see when there isn't a sweet spot and that means that the system is really invalid.
Thanks for the reply. I briefly tested AIQ for the purpose of backtesting and believe I noticed they had some sort of 3D, but couldn't get the support I needed at the time. Now I know what the 3D is for.
Random Entry Money Management Strategy 1 â Toss a coin for long or short entry: Heads long, Tails short. Enter in the middle of the bar; If you are using 3 minute bars then enter 1 ½ minutes into the bar. 2 â (for long, reverse for short) If the entry bar closes lower than the entry, cover immediately. 3 - If entry bar closes higher than the entry, keep position and place stop just below entry bar. 4 - If next bar closes higher than preceding bar, keep position. 5 - Keep doing this until price trades below the close of the preceding bar, and then close position You should make money over time. You cut your losses and let your profits run.
That is like asking, to build a car, are brakes all you need? It is essential, but insufficient. You still need an engine (edge) and other components.
There are two distinctly different schools of MM, one is the Van Tharp one where you are risk aversive to preserve your capital and your mental capital, and you can take a thin edge and make it more palatable. The other one is along the lines of the Kelly position sizing which applies after you have a solid edge worked out and you want to grow your account rapidly without risk of ruin. I like the concise nature of the Kelly math, you can characterize your strategy with just percent wins and reward/risk averages and some monte carlo work with account simulations to show you the worst / best case scenarios. You can do that and get a good idea of what will happen in the future without going into a terminal state of paralysis analysis. [and getting real cranky like some posters here]. There are some criticisms of the Kelly regarding how it was designed originally for a two outcome situation and trading is supposedly not just two outcomes... If I enter a trade and the method of exiting is in place, whether via TA or via some fixed number of points, and the method of stopping out is in place similarly then barring unusual events there are just two outcomes, win or lose... I don't see a lot of value in arguing that point at all. Unusual events just can't be taken into account in a system, forget it. They can better be handled by a stop based on account balance permanently held at the exchange if possible.
You can calculate optimal Kelly fractions with more than two possible outcomes. If x(i) is the outcome and p(i) is the probability of the outcome the growth rate is g(f) = Sum [p(i) * ln( 1 + x(i) * f)] An easy way to find f that maximizes above function is by using excelâs solver.