don't see how the size of the stop is related to the long term profitability of the trade It is most definitely related. Take the lowest-low of last 10 bars on one trade as your stoploss. Now take the lowest-low of the last 30 bars as your stoploss. You will most definitely get different results. Same applies for fixed stopped loss.
I should have been clearer. Of course, if you have a 5 dollar stop loss and take profits after the first 1/2 point, it will skew the results. What I'm saying is this. Can it be proven, that any betting methodology can predictably enhance the return of any system? Again, I'm speaking of methods other than the maximum efficent use of capital as in a Kelly system. If a method returns 10% on invested capital, how do you get more than that out of it? All you can do is invest the maximum amount possible over time without going broke during losing streaks. Any other fiddling will either introduce additional RISK to the method, which may enhance OR reduce profits. If there's some other holy grail of money management, I haven't yet absorbed it.
Unless his method results in greater stops on trades that have a lower probability of becoming winning trades (in which case adjusting size on those trades is not what he should be doing,) this makes no sense. You can optimize a positive expectancy system, you can't 'optimize' a loosing system into a winning one (not unless his system is just barely positive and the profits get eaten by commissions.) Not by varying size. voodoo
G'day everyone, first post here. I have been lurking this board for many months now, and wish to thank you all for the immense amount of information I have gathered while here. I hope to be able to participate more now (though I'm certain my questions will far outnumber my answer and help). Please note that I have not yet traded anything for real so far, though I have been studying and paper trading as much as possible since about April, (starting in stocks, now on futures since the SEC $25k rule changes), and will begin trading in about a month. So I have no experience with some concepts, and therefor I may only be able to fully realize them through my own trading. As for my question, it is regarding the examples of drawdown cited in some of these articles. I assumed that when trading there would be winning streaks, and losing streaks. But to lose over half of your portfolio? And from the time range he was talking it seemed he was saying this is over many months (though perhaps I am mistaken). Is there a logical reason such a large drawdown would occur? Obviously market conditions change, but I assumed that you would then alter your strategy in order to stay profitable. I was under the impression that after only 3 losing trades you should sit back and evaluate the situation until you find the problem and are able to solve it. I would assume this would be all the more important if you were to have maybe 3 losing days in a row. Is this just a part of trading then? I would think it would be better to just think of such a loss to be unacceptable, and after losing a certain percentage (10-20% maybe), just sit back and paper trade until you find your mistake, or the changes that need to be made in order to profit from the situation. If I missed something in reading these articles, or if anyone could clarify this issue for me, please do so that I may be better prepared for what is to come. Thank you, Ben
stock777: Can it be proven, that any betting methodology can predictably enhance the return of any system? Again, I'm speaking of methods other than the maximum efficent use of capital as in a Kelly system. Yes, there is a methodology to enhance the return of any system with a positive expectation, assuming that the probability of success is known or can be approximated. For example, in Blackjack you bet more when there are more higher cards remaining in the deck, and bet less when there are fewer cards in the deck. Both Kelly Criterion and optimal f are methodologies to accomplish this. If you applied this to BlackJack where the probabilities are known at all times, you will receive the same statistical result over time. The difficulty with applying any sort of variable position/bet sizing strategy to trading is that you are never sure what your probability of success will be on your very next trade. The best that you can do is to approximate based on historical results. This is the heart of the problem. If a method returns 10% on invested capital, how do you get more than that out of it? All you can do is invest the maximum amount possible over time without going broke during losing streaks. Any other fiddling will either introduce additional RISK to the method, which may enhance OR reduce profits. Another technique that is used I mentioned previously is to use 'optimal f', an offshoot of Kelly, and then factor 'optimal f' to your preferred risk tolerance. You do not need a mechanical trading system to determine optimal f. You can use the results of your discretionary trading over time because optimal f is just looking at your risk profile. For example, if the 'optimal f' on your system is 0.4, and you have $100k equity, it would be obviously foolish to bet $40k on your next trade. If your risk tolerance is 4% equity risk on any one trade, then use a factor of 0.04. So at $100k you would be $4k, if your account grew to $120 then bet $4.8k, if it dropped to $80k then bet $3.2k. However, keep in mind that optimal f is designed for normal and stable distributions. If your trading results in the future are not likely to be anything similar than your trading results in the past, don't use optimal f. While no one ever knows if future performance will be anything like past performance, with many systems trading approaches you can be reasonably assured that your results will be similar within a certain confidence level. Just like trading is all about probabilities, so is using a variable position/bet sizing strategy. If there's some other holy grail of money management, I haven't yet absorbed it. You're right, there isn't a holy grail of MM in trading because there is no holy grail way to know your chances of success on any one trade. The more accurate you can forecast your risk factors for each trade, the better you can improve your trading results. Hope that helped.
voodoo: "You can optimize a positive expectancy system, you can't 'optimize' a loosing system into a winning one" I give an example. Your trading system produce following results in points in 5 periods: +20, +20, -20, -20, -6 Total: -6 pts. This is negative expectancy system. If your bet size is always 20, you'll lose 120 pts. Let's say your starting bet is now 10, and if you win, you reduce your next bet size by 2. If you lose, you adjust your next bet size by 2. Results: (+20x10) + (+20x8) + (-20x6) + (-20x8) + (-6x10) = +20 points!!! We just optimize losing system into a winning one! It is possible in some rare situations. Let's say you have a positive system: +20, +20, -20, -20, +6 The same MM strategy gives: (+20x10) + (+20x8) + (-20x6) + (-20x8) + (+6x10) = +140pts which is higher than 120 pts if your bet size was always 20. But in general, any MM strategy ( betting less than 100% of your money ) will give lower profit for positive expectancy systems, and will give lower loss for negative expectancy systems. The only adventage of MM is - in my opinion - giving you a chance to stay in the game longer ( than without MM ) when your system has a expected max drawdown equal or higher to your account size. DT-waw
Earlier on this thread I wrote, that "The market behavior is very complicated, and it's impossible to describe it with few simple parameters. It can only be done by human brain". As you can see reading my last posts, I've changed my mind. It's possible to describe historical prices behavior with few stats. Simple, automated trading system can earn money. DT-waw
This is negative expectancy system. If your bet size is always 20, you'll lose 120 pts. Let's say your starting bet is now 10, and if you win, you reduce your next bet size by 2. If you lose, you adjust your next bet size by 2. Results: (+20x10) + (+20x8) + (-20x6) + (-20x8) + (-6x10) = +20 points!!! While this particular sample size of 5 bets resulted in a win using this betting progression, the overall expectancy of the system is the same: negative. Similar to many a random chance betting progression studied such as coin flipping/BJ, what happens with the above is if you alternate wins/losses in a short-time frame you show a profit, but you will lose in the long run because you win a lot less with consecutive winners vs. consecutive losers, not to mention you only win 40% of the time. Monte Carlo is another example. I think if you used a more realistic trading-related example rather than the traditional games-of-chance exercises, it would be more applicable.
This discussion is interesting, but theoretical. No one can trade optimal f, it's way too risky. Also, it's conceptually flawed. Your risk is not determined by past statistics but by current exposure. Trading is not playing blackjack, you can lose way more than you wagered. They can close the market for a week and if your exposure is too high, you're out of business. Your number one money management goal is always to avoid blowing up. Everything else is secondary. Ralph Vince is not going to pay off your mortgage if you crap out with optimal f. The far more practical money management issue is determining where to place your protective stops. This has a direct effect on profitablility. I have found through experience and testing that my winning trades almost never go against me by more than a certain amount. There are other factors but that maximum adverse move is the failsafe point. Many books advocate "natural" S and R points as stops. This is foolish nonsense. Pay attention to those points, but don't put your stops there. Placing your stops at highs and lows is just giving your money away. The floor traders spend all day shooting for them. That is their job and they are good at it. Or maybe you think it was just bad luck that the morning run up in the ND's just barely took out yesterday afternoon's range? Or that yesterday's lows just nipped through Monday's lows?
Nebeno, 40 and 50% drawdowns are common for system traders. CTA's try to avoid big drawdowns by using diversified portfolios and by trading much smaller size than individuals do per account equity. If you have a fully backtested system, I don't know why you would stop trading it after a few losses. In fact that is a classic reason system traders do worse than the vendor's record. For a discretionary trader, stepping back after a string of losses or at least scaling back is not a bad idea. The market is trying to tell you something. Usually it's a good idea to listen. My advice to newbies is take half your trading account and put it where you can't get at it. If you run through the first half, at least you have to do something to get at the remainder. You can always recover from the financial damage if it's not too bad. The psychological damage you suffer if you blow out your account is harder to deal with.