A single quality indicator for the strategy

New comment
 

I propose to come up with/develop a single coefficient showing the quality of the strategy, which will take into account its many characteristics (profit, drawdown, number of trades,...). In MT5 it is possible to use this.

This kind of task can be solved graphically:

Profit and drawdown are indicated in stop loss. The three values used in the functions shown in the graph do not contain such important information about the strategy as "growth stability", in part only the drawdown.

The result of these functions can simply be multiplied, resulting in a single number, which can be used to judge the overall quality of the strategy.

 
Aliaksandr Hryshyn:

Big scientist.

I don't use it at all, because I don't need it. And, by the way, I don't understand your enthusiasm for these indicators.

 
Yuriy Asaulenko:

Big scientist.

I don't use it at all, because I don't need it. And, by the way, I don't understand your enthusiasm for these indicators.

Automation of strategy evaluation. You need only one indicator that describes the overall quality of the strategy. If you have ideas, suggest them).
 
Aliaksandr Hryshyn:
Automate the evaluation of the strategy.

Also to automate this. What's the point?

Like Voznesensky. :)

 
For the strategy generator).
 

There is a suggestion:

drawdown - 3 to 5% in the testing period - 10 points

drawdown from 6 to 15 % - 6 points

drawdown from 16 and above - 4 points

And so on all parameters to derive a point system.

And then summarize, who has more points-a better system.

Evaluate the drawdown, profitability, and so on.

So on. What ranges of drawdowns and profitability and other parameters to take and what points to award them - that is the question)

 

It is advisable to use smooth functions, otherwise, it turns out that 3% drawdown and 5% drawdown are the same thing.

And the number of trades? Three profitable trades on the history does not say that the strategy is good, although the drawdown will be zero.

 
Yes, you can come up with a lot of such complex indicators. For example, I once used an indicator directly proportional to profit, profit factor, number of trades, evenness of profit distribution among trades and ideality of the regression line for the balance curve (the higher the slope and the smaller the spread, the better), and inversely proportional to the drawdown. But I haven't made any proof that a complex indicator is better than a few standard ones. The main advantage is that it is one and removes the issue of choosing parameters between several options when the standard metrics diverge in their estimates. The choice of function is highly subjective. There are some fans of the recovery factor, whatever...
 

"But there is no evidence that a comprehensive indicator is better than a few standard ones." - the only question is the approach, if such a method of evaluation is not suitable, it automatically means that the standard methods are not suitable either.

"Feature selection is highly subjective." - The level of subjectivity is no higher than the standard approach, the model is treated as a "black box" here too.

The significance and impact of the main features of the strategy on the overall indicator can also depend on other features of the strategy, and this is the standard approach.

 
The graphs do not take into account the stability of the balance line movement.
 
everything has already been invented a long time ago
123456789
New comment