A great book on testing and optimisation - page 6
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
Valmars, I have an idea. I'm going to purposely, for fun, write out a complete list of what Pardo suggests here in this thread. It will turn out to be very impressive (I finished a preliminary reading of this book today). Let this list serve as a reminder to grail makers of how much they underestimate the current level of understanding in this field. Still, this list is, by the book's intention, at least some, not at all 100%, relative assurance that the strategy has a reasonable right to exist. The list will be here shortly.
Where's the list? .....öklmnoprst......
Yes, you are. We should put it out there. Thanks for the reminder, LeoV.
Something about the subject has flown by. It got pretty interesting. Waiting (I'm not the only one).
Yeah, it sure is. Should put it out there. Thanks for the reminder, LeoV.
Something about the topic just flew by. It got pretty interesting. Looking forward to it (I'm not the only one).
+1 The topic is very interesting indeed....
1. Forming a trading strategy in the form of a flow chart. Formulating TS in the form of rules in a pseudolanguage. Coding of TS.
2. testing:
a. Checking whether the code does exactly what was formulated earlier, on a short piece of data,
b. Getting a rough idea of the profit and risk profile - based on testing in different markets and time periods:
This phase is a rough estimate of how the system behaves under more or less reasonable parameters. If the system shows acceptable parameters, you can move on to optimization.
3. Simple optimization: Just what we do in the optimizer, setting parameter ranges and steps. At this stage we try to squeeze the maximum out of the system. We choose the options we like best.
4. Forward Analysis. This is what the author himself writes about its importance:
How exactly it is carried out is described on pages 28-31 of the book, as well as in Ch. 7.
5. System Trading.
6. Comparison of the profits obtained in testing and real trading [or on the demo, if there is reason to believe that the results on the demo will not differ significantly from the real - Mathemat].
7. Improvement of the system.
These are all just major steps, which need to be elaborated on. More to follow. In the next post I will clarify what the author understands by testing, and what the requirements are.
The first stage of testing is the selection of an adequate test window, i.e. the testing area. The test window should ensure 1) statistical representativeness, 2) relevance to a given TS and market.
1. Statistical representativeness.
Firstly, it is a sufficient number of transactions: if the number of transactions is N, then the standard error in the system parameters determination is approximately equal to 1/MathSqrt(N+1). Explanation:
The standard error is a notion applicable not only to the value of average gain, but to anything. For example - to the duration of trades. It would be desirable that profitable and loss-making trades would be evenly distributed within the testing area.
Next, the number of degrees of freedom of the system is estimated (p. 68-69). Roughly speaking, it is the difference between the number of signals and the number of rules defining the signals. A more or less reliable estimate of the minimum required number of degrees of freedom is ten times the sum of the number of rules and the number of conditions. If we have 5 input/output rules and 3 conditions to them, the number of degrees of freedom must be at least 10*(5+3) = 80. But this is the minimum which is desirable to exceed.
Further, it is desirable to cover as many real market types in the test window as possible. If the testing is performed only on the bull market, the system can obviously operate only on it.
2. The testing data should be relevant to the TS itself and the market characteristics.
The author's reasoning on this point is very vague. Their essence comes to the fact that during testing only data similar to the current trading conditions should be used.
In Chapter 5. 5, the author considers different methods of searching the best strategies (including genetic algorithms), but they are not of much interest for us here, because they have already been implemented in the tester. And then, from page 89, the author concentrates on strategy evaluation methods. Evaluation criteria given there are quite interesting, and not all of them are implemented in the MT4 tester. Most beginners usually look only at the gross profit, but this is not the most optimal strategy evaluation parameter.
Apparently, the author considers the pessimistic return on margin (PROM, see pp. 93-96) to be one of the best comprehensive criteria.
OK, let's stop for now to take a little break...
So what is this film about?
Nothing......
OK, LeoV, I can finish. Did you get bored?
P.S. Does everyone agree that we can go no further?
2 Korey: In the fifth league all very much want to get, it is obvious. But very much want to do it so that quickly and without much effort...
OK, LeoV, I can finish. Did you get bored?
P.S. Does everyone agree that we can go no further?
NO WE DON'T!!!
LeoV may know it all. But the rest of us( me in particular) would love to read.