LOLOL - No the maths was wrong, but not my much. I realised where the error was when I started to get the same result on the spread sheets no matter what the order. The reason it will always lose is, as I suspected, that every loss is made from a higher cash level than its corresponding winner. rgds Natalie
I think one of the reasons for the whole discussion lies on the disadvantages of thinking in percentage terms. When I read the other thread of acrary, I was stuck by his proof that thinking in % reduces your overall performance (very simplified, I know). I think this is due to the fact that people think in % and not in log returns. I attach a spreadsheet on this. I can imagine that this sheds light on the basic problem of order, once you use percentage terms. The problem IMO is that neither percentage nor log returns fulfill all intuitive requirements. Log returns have some features, that are quite disturbing once you are used to precentage thinking. For myself I conclude that no concept gives all answers: 1. absolute dollars: you cannot bet the same amount once you have ten times the size. 2. percentage: it is ridiculous to say that an equity curve of 100 - 150 - 100 has won money. 3. log returns: it is very unintuitive to be faced with returns that can fall below -100% I think the real trouble starts when you want something from one way of thinking that this way cannot provide. 1. absolute dollars: it is fine as long as you do not try to compare two points in time where your capital differs substantially. If you do so, you compare very different risk features. Consider a trader who started with 250k and falls back to 50k, but still bets the same amount of money in absolute terms. If that does not bother you consider another trader who just started with 50k. Both have the same money, but trade different amount, just because one has a history. Cannot be. The market does not care where you were last year. 2. percentage: if you do want to judge what you made over a single period, it is fine. Whenever you start averaging and compounding you get in trouble. 3. log returns: fine for averaging and compounding. Gives a distorted, unintuitive picture for one period (100 to 150 is not a 50% gain) To sum up. I think that all of us probably do not suffer too much from the effects described in actual trading, since we usually move in low percentage area, where the differences an weaknesses do not effect to badly. But it is a huge problem IMO for testing strategies, thus for our strategy choice. Sorry if I am wrong and bore you with that stuff. I am not completely sure if this really relates to the subject, but I guess it could. In any case - I like this thread. peace
Would you mind summing up the findings once we will finish this thread? With all agrees and cons? I always find it a little frustrating when threads just - die. (I do not assume it is already over here - just came to my mind). peace
I think you have made some good points there, especially the one about testing out strategies before using them. Knowing and understanding what works and why, and where the break even points are, seems to me to be one of the most important elements in finding strategies to trade. The original question was about bigger less frequent winners vs. smaller more frequent ones, which has raised the question of where do you actually break even with no 'Edge' in the market, and 50/50 clearly isn't the answer. To me, it now enlarges the question not only to bigger vs more frequent, but also to include money management (that does seem to keep coming up and does seem to be hugely critical to success ) I'd love to hear views about money management approaches and how they relate to this. Rgds Natalie
with all respect to the initial topic of that thread, I would like to sum up what I think on MM. "The amount of money bet is exactly as important as the direction of the bet." Single market: 1. first place the stop, then calculate the number of stocks/contracts according to the risk derived out of current price and stop. (this is basic, just to define starting point) 2. see the amount of money in the same way as you see a filter for the signal. It is just not binomial (1 or 0), but kind of semipermiable, thus you can bet less when expectation value of a possible trade is positive (= signal passed the binomial question), but you can vary your amount to the distribution lying below that positive expectation value. For futures trading to use volatility as a determination for the amount to bet is quite common. Bet more when vola is low, some add bet more when vola is extremely high as well. In essence that means that the distribution of the expectation value is connected to volatility. By means of multiregression, one can use any kind of variable to determine the influences on the distribution of expectation value. In essence one will be interested in the fat tails of that distribution. Break Outs will yield more (not necessarily more often!) in line with trend, not against it. RangeReversion will yield more when vola is high. A gap-fading play in the NDX will yield more when the SPX did not gap at the same time. You get the idea - analyze what circumstances relate to the fat tails of your tradeResult distribution. 3. managing a trade. once everything is set in place there are numerous ways to play a trade. Pyramiding is one of them. IMO pyramiding can best be seen as a additional trade, which is by nature trend following. I used to run models with different time frames and the effect for that market is like pyramiding with each time frame in essence operating separately. 4. portfolio. most important is to realize that you cannot overestimate the effect of adding stocks and strategies. Three medium strategies combined can make a bank. 5. portfolio optimization. a thing that takes volatilities and correlations into account is VAR, when you base it on historic simulation and not on correlation matrices. we programmed a thing that can test the effect of keeping a portfolio at constant VAR. Sorry, again a long post. peace
sorry for the trouble to read my last post - many spelling mistakes. should take more time for this ... (I hate to read my own posts ...) peace
pretty much, yes- leveraging profit rather than initial capital. you must be joking. the entire trading process is contradictory in the sense that risk and reward are contradictory; balancing them effectively is the critical issue for any trader. This doesn't make much sense to me, as when you add to a position, the initial part of your position is still in play and so there is no progression from one trade to the next, simply a continuity of growth. From a risk management perspective it is still one trade as well, as the profit of the initial position is used to lay off the risk of the addition, thus keeping risk level the same or even decreasing it (adding in such a way that you are break even if you bail)- again, this implies continuity rather than a fresh play. Last but not least, generally the only reason you add to a favorable position is because you planned for it from the start, thus suggesting your initial goal was to target a unified move in stages, like a football game has quarters or a baseball game has innings. Time frames don't really move "separately" either because they are all part and parcel of the same continuum. But hey it's just opinion, so whatever blows your hair back...
For a graphic display of how w/l and win % affect your equity curve go to the following web site. http://www.hquotes.com/tradehard/simulator.html Norm
I had spent a good part of my waking hours mulling over the differences between your statements and figures and my simulations and intuition. And then it hit me. You are not presenting the full picture. Let us take the "breakeven" trader from my earlier post who starts off with $100,000 and executes 2 trades with 3% risked per-trade. As before, % win = % loss = 50%, with win size = loss size = 1. Your method of depicting this case (and others) does not include all possible scenarios and is therefore flawed, IMHO. For this particular situation there are 4 scenarios, each with equal probability of occurring : {Win Loss}, {Loss Win}, {Win Win}, {Loss Loss} As can be seen from the attached Excel worksheet, the expected ending equity is $100,000. The "drag" you isolated is made up for by the "boost" in {Win Win}. The negative expectancy you allude to is counterbalanced by the positive expectancy you have chosen to ignore. The expectancy of the strategy is zero and remains so even with a %-of-equity-based sizing strategy. Similarly, if you include all possible scenarios in your earlier 40% win rate vs. 70% win rate example (see link below), http://www.elitetrader.com/vb/attachment.php?s=&postid=193181 and weigh them by frequency, you will end up with equal ending equities for both strategies, as indeed I did in my simulations. It will not have a higher return - the returns are identical.