LOLOL! I saw this thread title and the first thing I thought of was you db'! LOL, oh man, I bet you rolled your eyes and said just what you posted. You just gotta love this place.
Here is a snapshot 2 weeks into the experiment. Still positive and seems to be holding an edge even though the system went though a significant drawdown. Samples: 40 Wins: 15 Losses: 25 Win Ratio: 37.5% Longs: 19 Shorts: 21 Ratio: 47.5% Longest consecutive winning streak: 3 Longest consecutive losing streak: 7 Net pips: 121 (net of interest, spread, and slippage) Attached is the screen shot of the chart with entries and exits. I had to move the timeframe to 3hr candles to capture the past trades.
Honestly... I think it'll outperform a newbie... Considering this will as a random entry, set exit system, I'd say with the right Money Management and some outside exit criteria, it'll be great. Non-biased entry, biased exit... psychologically, it's easier and efficient to react to the situation (exiting a position) than to entry...
You flip a FAIR coin ten times you can get 6 tails for example so that the REALISED FREQUENCY of this PARTICULAR experience is 6/10 so above the 1/2 (THEORICAL PROBABILITY) but the EDGE is still NULL because the EDGE doesn't concern a REALISED FREQUENCY but the THEORICAL PROBABILITY and the THEORICAL PROBABILITY is not changed by the PARTICULAR experience and so the EDGE either. And why the realised frequency is not always equal to the theorical probability is obvious: if it was always equal, one wouldn't need a probability theory . This seems trivial but look at people who play lotto: they think that some numbers have more occurence than others because there is something special (or special cause) about these numbers. Now I won't baffle to try to do that kind of experience because simulation is a good teaching tool. Nevertheless one must always remind the spirit of experiment in scientific approach: an experiment cannot be interpreted without a knowledge theory. See Henri Poincaré, one of the greatest scientist of all time, about Science and Hypothesis http://www.utm.edu/research/iep/p/poincare.htm#Science and Hypothesis "Experience suggests scientific theories; but experience does not justify them." Or Deming (the Statistician Guru in Quality) http://www.maaw.info/ArtSumDeming93.htm "Without theory, experience has no meaning and there is no learning. Copying examples without understanding the underlying theory may lead to disaster."
Theory: Random entry with a wider target vs a shorter stop. When this theory is tested in a random environment, it breaks even. This is known. What is not known is if it profits in a Forex market. If it did, there must be something non-random about that market. A difficulty arises with assessing a short sample such as 100. The profit would need to be obvious and steady. As a rule of thumb, given a 20 point target and a 10 point stop, the win to loss ratio would be twice as many losses as wins for break even. Similarily, a 30 point target vs a 5 point stop would produce six times as many losses as wins for breakeven. There is a margin of variance that is wider as the sample is shorter. We are now at a sample of 40 with a 25 target vs 10 stop. Theoretically we should have the following results: 25/10 = 2.5 losses to 1 win. This is about a 71% loss rate such that a sample of 40 would yield 28.5 losses and 11.5 wins, if there was no variance. So in this case, with 15 wins and 25 losses, we have 3.5 extra wins and 3.5 fewer losses (25*3.5 + 10*3.5 = 122.5) than expected, accounting for a net of 121 pips kept. This would be a 62% loss rate for a 9% variance over our sample of 40. For arguments sake, we will say 122.5 = 121 pips netted so far. We still need to account for th 2 pip per trade vigorish: 40*2= 80 pips that must be paid from the 121 = 41 pip profit. So the question is, can a variance of 9% over 40 be expected? Said another way, is 3.5 extra wins, or visa versa common enough to be still considered random? The way to determine that would be to run many samples of 40 and look at a bell curve of the variances. If it is rare, we may be seeing some non-random activity. When paying 80 pips to make a possible 41, we would want to make very sure the profit is certain and steady. In my experience with randomness, one would have to run several thousand samples of 40 to determine the certain rarity of a 9% variance. This is where the difficulty arises, when trying to collect data from the market, especially walk-forward. To summarize, 3.5 extra wins must be extremely rare over a sample of 40 in order for this experiment to suggest non-randomness. You must maintain a 9% variance in your favor to profit here, and it must be absolutely certain. Over a larger sample set, the variance will tend toward 0%. If the variance slips to only 7% you get eaten up after vigorish. Thus, a 2% different can kill the deal. A sample of 100 would likely not be enough to determine the certainty of maintaining a 9% variance/"edge". If, in 8 out of 10 samples of 40 you got 3.5 extra wins, only then would you - maybe - be dealing with a profitable Forex system. It is suggested that you expand your walk forward to a set of 400, requiring 8 out of 10 to profit before comming to a conclusion. JohnnyK
Merely a thought: I would believe Deming's systemic (i.e. totality/wholistic, instead of systematic) concept could be applicable to many things including this thread. Likely an entry should be just one of the many sub-systems in a trading system, and optimising only this entry sub-system would not be the total answer to the problem. Furthermore, basically we all would want to increase our edge (i.e. probability) and expect a (say) 55% (or any comparatively higher) win-rate during a trend market. Then who would prefer to rely on a random-based system expecting an average 50% win-rate?
Here's the data at 58 samples. Samples: 58 Wins: 17 Losses: 41 Ratio 29.31% Longs: 27 Shorts: 32 Ratio: 45.76% Net: 6 pips (net of interest, spread, and slippage) Attached is the equity curve through 58 samples. If this strategy (through longer sample sets) breaks even like it is theoretically suppose to do in a random environment, then doesn't that give weight that price action is random? Said another way, does this experiment confirm the Random Walk Theory?
You have no idea the importance of this statement. It is the cornerstone of all such analysis, and if you like, it is the _only_ piece of data that you cannot leave out when doing any kind of high frequency analysis. Do you know that if it weren't for friction, I would never need to buy another drop of gasoline? nitro
Hi All, Popular wisdom has it that very few people know how to play the market right. If this is true, then many people must be trying all kind of tricks but nevertheless will fail. Random entry is probably best relegated to the class of tools used by the ineffective speculators. We should allow for a few artists knowing how to make a profit with random entries, but these are not going to let us in on their tricks. Random entry will never enable the many failures to shift over to the category of the wise. Of course, in theoretical probalistic models, something can definitely be said about outcomes of random and non-random policies. However, talking about markets, ever since the word "random" came into the public domain, all kind of people are continuously advancing theories, pro and con, on this. Also let's not forget that all these "markets" could be wildly different beast among one another. nononsense