Why does any strategy only work successfully for a limited time and then stop working? - page 11

 
wise >> :

.....No strategy with negative MO will pull out any "right" MM.

You don't have to put all the power of MM on MO. There are a lot of interesting things in TCs besides the MO, for example Z-score. If it is stable and different from zero, an appropriate MM can turn a negative MO into a positive one.

 
coaster >> :

You don't have to put all the power of the MM on the MO. There are many other interesting things in TCs besides MO, such as the Z-score. If it's stable and different from zero, the appropriate MM can turn a negative MO into a positive one.

It would be more convincing with examples. =)

 
wise >> :

It would be more convincing with examples. =)

You won't need convincing if you're familiar with such concepts as system predisposition to series (losses/profits) Z-score. Or vice versa: a predisposition to repeat results of trades. It will be better for you and it will save me a lot of time. :)

 
By the way, does anyone know why the Z-score depends on the number of trades? The more trades, the more the Z-score can deviate from 0. Has anyone paid attention?
 
benik >> :
By the way, does anyone know why Z-score depends on the number of trades? The more trades, the more the Z-score can deviate from 0. Has anyone paid attention?

As the series of trials increases, the reliability of the results increases. The number of tests in a series should tend towards infinity. This applies to any kind of test.

 
joo >> :

As the test series increases, the reliability of the results increases. The number of tests in a series must tend towards infinity. This applies to any type of test.

Yes, but in my tests sometimes Z-score exceeded 300. On the average, with a number of tens of thousands of deals Z-score varies from 10 to 100. Can it be like that? Or it is a mistake in calculations (although I have double-checked everything 50 times and have not found any error).

 
benik >> :

Yes, but in my tests sometimes the Z-score was over 300. On average, with a number of tens of thousands of trades, Z-score ranged from 10 to 100. Can it be like that? Or it might be a mistake in my calculations (but I've checked it 50 times already and have not found any error).

Actually, I have no idea what a Z-score is. :)

But my advice still stands. The larger the number of tests, the less the overall result is affected by single incorrect data.

 
Here, for example.
 
coaster >> :

You won't need to be convinced if you become familiar with such concepts as system predisposition to Z-score series (losses/profits). Or vice versa: a predisposition to repeating the results of trades. It will be better for you and it will save me a lot of time. :)

No, that's not what I mean. It would be interesting to see a report of such a system practically. Even two reports. One with predisposition and one without. To see if the predisposition is justified. =)

 
benik >> :

Yes, but in my tests the Z-score was sometimes over 300. On average, with a number of tens of thousands of trades, the Z-score ranged from 10 to 100. Can it be like that? Or it might be a mistake in my calculations (but I've checked it 50 times already and have not found any error).

Error. According to the link to Rosh's article, Z-score is interpreted as the number of sigmas of standard normal distribution, by which the real result deviated from conditionally random. A Z-score of 5 is about the same as a Z-score of 100. In both cases, the chances that the real sequence of trades is random are negligible.