![MQL5 - Language of trade strategies built-in the MetaTrader 5 client terminal](https://c.mql5.com/i/registerlandings/logo-2.png)
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
OK, let's keep it simple - two TS.
One in a period of 1.5 months makes 179 trades and shows 240% profit
The second one for the same period - 40 deals and 52% of profit.
"Detrended" balance charts - the first one has worse balance values than the second one.
What do we do, throw out the first TS?
I always thought that the analysis of model's residues serves the task of predicting the behavior of series using this model - whether the model is adequate or not, etc.
And what about the balance chart? Are we making a profit forecast?
There, at last. This is exactly the topic of the topic.
You can't compare 240 with 52. For me it is a matter of principle. The point is that 240 and 52 are realisations of random variables. That's how it fell out. And the basic question is: what is the probability that in the future it will be the same, or almost the same, or not at all the same?
The residual analysis must answer this question and it is far more important than the size of the test profit.
Yes. We got a linear regression, the residuals are normally distributed - the model is adequate. What's next? Predict profits?
No, why predict it? The robustness of the system will be checked. In particular, we check that mo does not change or changes slowly. This is the model: equity=MO*N + noise (N number of trades). If the residuals (noise) are normally distributed with mo=0, then the model is adequate. I.e. the TS has unchanged Mo.
Yes. We got a linear regression, the residuals are normally distributed - the model is adequate. What's next? Predict profits?
Nothing of the kind.
I removed the trend in order to be able to analyse the residue statistically. This position is a position of principle.
This has been shown above. Statistical analysis of quotes is only possible on detrended (smoothed) series.
This is a position of principle.
so it's not about analyzing quotes anymore, but why detrended equity
No, what the hell is there to predict? What is being tested is the robustness of the system. In particular, that mo does not change or changes slowly. This is the model: equity=MO*N + noise (N number of trades). If the residuals (noise) are normally distributed with mo=0, then the model is adequate. I.e. the TS has unchanged Mo.
OK, good.
We find that the model is adequate and has an unchanging MO. What next? What did we build the model for?
What if the model is inadequate? OK, we find a non-linear regression model and it's adequate. And then what?
Let me tell you another secret - regression analysis is a forecasting tool. What are we predicting here?
OK, all right.
We find that the model is adequate and has an unchanging MO. What next? What did we build the model for?
What if the model is inadequate? Well, let's find a non-linear regression model and it will be adequate. And then what?
Let me tell you another secret - regression analysis is a forecasting tool. What are we predicting here?
if the MOI is unchanging, that's good. The system can be traded - robustness. If not, there is no confidence in past results and the system cannot be traded.
Wait for it. You won't get stationary. Forget about normality altogether. The world around us is not normal, it's residuals from quotes.
Who said such a stupid thing? These are the results of the TS - they may well be stationary.
Take the TS draining about the size of the spread - trading results will be stationary
if the MO is unchanged, then it's good. The system can be traded - robustness. If not, the past results are not reliable and the system cannot be traded.
)))))))
OK, now I understand - if the results of the TS, for example, grow abruptly and show a crazy profitability over a long period, but the regression residuals do not have a normal distribution - the TS should be discarded))))))))
Normality of residuals distribution as an indicator of TC performance))))))))))))))0
)))))))
OK, now I get it - if the results of the TS, for example, grow abruptly and show insane profitability over a long period, but the regression residuals do not have a normal distribution - the TS should be thrown out))))))))
That's why we need to analyze only normality in the negative zone, that econometricians don't take into account, but ordinary traders do ;) Thick tails in the positive zone are normal, e.g. for trendfollowing
Well that's why you only need to analyze normality in the negative zone, which econometricians don't take into account, but ordinary traders do ;)
What is "normality in the negative zone"?