A single quality indicator for the strategy - page 6

New comment
 
"The lot-to-deposit ratio is also indirectly related to risk. Prescribing lot size from deposit is a matter of three seconds. But these are MM methods, and they have very little effect on the quality of the system itself. At the same time there are a lot of MM methods, including hedging, locking and adding. Will you take them into account? And how are you going to take stop-losses in cases where the loss is PLANNED, for example in the case of pair trading? - If you want to take them into account when you have a planned loss, like in the case of pair trading, for example, we have a simple))), there is an entry condition, exit condition, stop loss and take profit, the latter can be replaced with a trawl or removed altogether.
 
"That said, by and large the risk is SEVERAL to the net profit. Why? Because its significance is shown on a large number of attempts. That is, if you take a large enough number of trades, then the riskiness of the system will already be factored into the profits earned. It is not possible to earn long enough on a risky system. All other things being equal, a riskier system will simply earn less. Therefore, for a correct comparison of quality we must simply provide these "other things being equal". I repeat - our goal is not an infinite reduction of risk, our goal-an infinite increase in earnings. Risk assessment is just a tool for a particular system." - Without these "other equals" there is no way to evaluate the strategy, evaluating the strategy and evaluating its results are different things...

"The bottom line is that if you need one parameter, you can't find anything better than net profit. As an option - the profit multiplied by the profitability. ( You can also use the beauty of a chart, but you can hardly express it in one figure.)" - strangely enough, the beauty is quite well estimated, the single coefficient is quite consistent with the intuitive estimation. The coefficient is also slightly influenced by the complexity of the strategy itself.

I completely agree with you that the evaluation of a strategy reflects the behaviour of the strategy in a particular environment. But let's not lie to ourselves, some characteristics of the environment are determined not by the market itself, but by the strategy tester, which simulates some market behaviour)), although everything can be fair (analysis of executed trades).


 
Aliaksandr Hryshyn:
- I just have this problem, tighter strategy evaluation, simpler MM, necessary for strategy generation, the "tricky" strategy finder can "take advantage" of the strategy tester features and show a nice equity, which is not good.

What do you mean by taking advantage of the features of a strategy tester? What do you mean, take advantage of?

Just don't backtest with beautiful equity, but with normal, out -of -sample plots and then beautiful or ugly equity will be an indicator of the system's efficiency.

I.e. you won't succeed without normal volking forwards, just like your predecessors didn't.

I think it's obvious - the toughest assessment of the quality and effectiveness of ANY business model is the RESULT. And it's the result AFTER training, not in the hothouse conditions of back-testing.

In other words, it doesn't matter what's inside the system, all that matters is how much output it produces. We cannot wait for years in real trading, so for now the only way to objectively understand something about an EA is to check out -of -sample on history.

 
Aliaksandr Hryshyn:
Evaluating a strategy and evaluating its results are different things...

Of course, I realise that all this fuss about forex, shares, etc. has long become a kind of subculture, a social environment with its own myths, its own features, etc. But personally, besides all sorts of side-effects like entertainment, self-development, etc., I would like, first of all, to earn money.

From this point of view, of course, you can evaluate the code of the programme from very different perspectives, but I prefer to evaluate it primarily in terms of results.

 
Aliaksandr Hryshyn:
I completely agree with you that the evaluation of a strategy reflects the behaviour of the strategy in a specific environment. But let's not lie to ourselves, some characteristics of the environment are determined not by the market itself, but by the strategy tester, which simulates some market behaviour)), although everything can be fair (analysis of executed trades).

You have just voiced a popular myth. Of course, the tester is not perfect, but what it does - it does REALLY well. I purposely took measurements and you too can repeat the simplest experience. Take some Expert Advisor, put it on a real broker's demo instead of the Metakvot terminal, wait a month, and then compare the real profit chart with the tester one on a minute chart. The coincidence will be 90-95%. This is quite enough to compare EAs with each other and improve the strategy.

And if more accuracy is needed, MT5 already supports tick history. I personally do not use ticks because losses in time are not worth the small increase in accuracy they provide.

By the way, many scalper creators will fail in their expectations when the tick history becomes generally available - so far it seems to them that their strategies are inadequate on ticks because of poor history, but not because of the quality of the strategy itself.

 
Youri Tarshecki:

What do you mean by taking advantage of the features of a strategy tester? What do you mean, take advantage of?

Just don't backtest with beautiful equity, but with normal, out -of -sample plots and then beautiful or ugly equity will be an indicator of the system's efficiency.

I.e. you won't succeed without normal volking forwards, just like your predecessors didn't.

I think it's obvious - the toughest assessment of the quality and effectiveness of ANY business model is the RESULT. And it's the result AFTER training, not in greenhouse conditions of back-testing.

In other words, it doesn't matter what's inside the system, all that matters is how much output it produces. We cannot wait for years in real trading, so the only way to objectively understand something about an EA so far is to check out -of -sample on history.

"What does it mean to take advantage of the features of a strategy tester?" - imperfection of the tester, because of this there can be overstated results, or even show super profits at all, in MT4 once there was such a thing. I had to make my own strategy tester, so the strategy generator would search "honestly", I had to introduce certain limitations. Then I should enable to use tick data...

You are right, I totally agree with you)).

 
Youri Tarshecki:

Of course, I realise that all this fuss about forex, shares, etc. has long become a kind of subculture, a social environment with its own myths, peculiarities, etc. But personally, besides all sorts of side-effects like entertainment, self-development, etc., I would like, first of all, to earn money.

From this point of view, of course, you can evaluate the program code from very different angles, but I prefer to evaluate primarily in terms of results.

You are missing the point).

Evaluating the strategy is an evaluation of the opportunity to make more money (future).

Evaluating profits is an evaluation of past results.

One profit alone is not enough to assess the possibility of earning in the future, at a minimum, we must figure out where this profit comes from (an evaluation of the strategy / opportunity). If a person won in a casino, it absolutely does not mean that he will win the same money tomorrow, i.e. we must think about why we were able to win (an assessment of the possibility/strategy).

 
Youri Tarshecki:

You've just voiced a popular myth. Of course, the tester is not perfect, but what it does, it does REALLY well. I specifically measured it and you can repeat the simple experience too. Take some Expert Advisor, put it on a real broker's demo instead of the Metakvot terminal, wait a month, and then compare the real profit chart with the tester one on a minute chart. The coincidence will be 90-95%. This is quite enough to compare EAs with each other and improve the strategy.

And if more accuracy is needed, MT5 already supports tick history. I personally do not use ticks because losses in time are not worth the small increase in accuracy they provide.

By the way, many scalper creators will fail in their expectations when tick history becomes generally available - so far it seems to them that their strategies are inadequate on ticks because of poor history, but not because of the quality of the strategy itself.

For me this "good enough" may not be enough, it's because of autogenerator, it can "squeeze the last out" with 5-10% imperfect strategy(historical data), I care about high quality.
 
Aliaksandr Hryshyn:

You are missing the point)).

Evaluating the strategy is an evaluation of the opportunity to make more money (future).

Evaluation of profits is an evaluation of the results of the past.

A profit alone is not enough to assess future earning potential, at the very least we must estimate where the profit came from (strategy/opportunity assessment). If a man won in the casino, it absolutely does not mean that he will win the same money tomorrow, ie we must think about why we were able to win (an assessment of the possibility / strategy).

You carefully re-read what I am writing. I am just writing about profits in the future, not in the past. When testing in wolfing forward mode, the Expert Advisor is just in the unknown future. And it is profit from this future that I suggest to take as a criterion of estimation.

I mean, once again,you will not succeed without a normal volking-forward, just like your predecessors did not succeed.

 
Aliaksandr Hryshyn:
For me this "good enough" may not be enough, it's because of the autogenerator, it can "squeeze out" the last 5-10% of imperfect strategy (historical data), I care about high quality.

To compare the two strategies, even a 60% market fit would be enough to make a choice in favour of one or the other. You will just have to increase the sample size.

For real code work the available 95% accuracy of simulation of the environment at 1m is very good . And a tick history will give all 99% .Many scientific and practical environment models cannot boast of such accuracy .

I do not see any reasons to believe that the automatic method of strategy creation is somehow miraculously different from the manual one in terms of history quality dependence.

If the history is bad, it is bad for all Expert Advisors.

In other words, the matter of choosing the environment is not directly related to the criteria of system evaluation.

123456789
New comment
Reason: