Hrmmm... that's true... I'm a moron... I'm not smart enough to know, before hand, if ideas are good or not. What do you suggest for Frosty? - Should he not test and do exactly what everyone tells him to do? - Should he start running tests that he has no objective knowledge about?
I don't get what you are saying. Please elaborate. Are you saying he should piss in his bed? And find causality in that????
I recommend you find a way to convince yourself that your results from random testing are stable and repeatable. (Convince yourself that the answer to the question "Have I simulated enough random trades" is a resounding Yes). Here is one way to go about the task. Doubtless you can think of other ways. Choose a summary statistic. LeBeau chose Winning Percentage. You might choose Profit Per Trade in points. Choose a number "T". This is the number of random trades you are going to test. Open the newspaper and choose ten 7-digit numbers at random. Maybe from the sports pages or the stock pages or the weather page. Call these S(1) thru S(10). These will be used as seeds to your random number generator. Let C = 1 Seed your random number generator with S(C) March through the price data and generate "T" random trades. Keep a running summation of the total #trades and the total #winners. Also keep a running summation of the total Profit and the total #trades so far. After each trade calculate %Wins_So_Far (= #winners / #trades) and also calculate Profit_Per_Trade_So_Far (= TotalProfit / #trades). Plot %Wins_So_Far versus #trades. Plot Profit_Per_Trade_So_Far versus #trades. Add 1 to C. If C is less than 10, go back to step 5. Now you've got 10 plots. Overlay them on top of each other. Did all ten sets of random trades, all converge to the same result? If so then you can feel comfortable you have simulated "enough" random trades. If not, increase "T" in step 2 by a factor of 5 and run the whole procedure again. At the end of all this, you will have a value for %Winners and a value for Profit Per Trade in points, that you believe. You will know with confidence, how well (random entries + N-bar hold times) succeeds, on this set of price data, for this interval of time. Of course you don't expect %Winners = 50.000% and you don't expect Profit Per Trade = 0.000 points. Why? Because you are testing Long Only. If the data has an upward drift, this will bias the results towards positive profits. If the data has a downward drift, this will bias the results towards negative profits. Furthermore, you are deducting slippage and commissions. This will bias the results towards negative profits. Therefore what outcome do you expect? The answer is: you don't know. That's precisely why you're performing the experiment in the first place: to find out.