Machine learning in trading: theory, models, practice and algo-trading - page 3034

New comment
 
Aleksey Vyazmikin #:

I did exactly as you wrote.

N P.P. Balance 1 T.L.Reg. Deviation Correction Balance 2
1 0 4,075 -4,075 -2,8525 0
2 5 4,4929 0,5071 0,65923 5,15213
3 7 4,9108 2,0892 2,71596 7,62676
4 5 5,3287 -0,3287 -0,23009 5,09861
5 3 5,7466 -2,7466 -1,92262 3,82398
6 8 6,1645 1,8355 2,38615 8,55065
7 10 6,5824 3,4176 4,44288 11,02528
8 9 7,0003 1,9997 2,59961 9,59991
9 8 7,4182 0,5818 0,75634 8,17454
10 7 7,8361 -0,8361 -0,58527 7,25083
11 9 8,254 0,746 0,9698 9,2238
12 8 8,6719 -0,6719 -0,47033 8,20157
13 9 9,0898 -0,0898 -0,06286 9,02694
14 8 9,5077 -1,5077 -1,05539 8,45231
15 9 9,9256 -0,9256 -0,64792 9,27768
The points below the regression line should be made below the primary balance line.
For example, point 5 with a value of 3 should become not 3 + 0.82398 = 3.8239, but 3 - 0.82398 = 2.17602 i.e. 3 - deviation*0.3.
 
Forester #:
The points below the regression line should be made below the primary balance line.
For example, point 5 with a value of 3 should become 3 - 0.82398 = 3.8239 instead of 3 + 0.82398 = 3.17602.

I am correcting the deviation from the regression line -1,92262 , do you just want to correct the balance?

Then here's the balance.

1 0 4,075 -4,075 0
2 5 4,4929 0,5071 6,5
3 7 4,9108 2,0892 9,1
4 5 5,3287 -0,3287 3,5
5 3 5,7466 -2,7466 2,1
6 8 6,1645 1,8355 10,4
7 10 6,5824 3,4176 13
8 9 7,0003 1,9997 11,7
9 8 7,4182 0,5818 10,4
10 7 7,8361 -0,8361 4,9
11 9 8,254 0,746 11,7
12 8 8,6719 -0,6719 5,6
13 9 9,0898 -0,0898 6,3
14 8 9,5077 -1,5077 5,6
15 9 9,9256 -0,9256 6,3
 
Aleksey Nikolayev #:

Inventing their own Sharpe and Sortino) With blackjack and other necessary stuff)

How to calculate this outrage more easily, can you tell me? :)

 
Forester #:
evenness of balance)
What do you mean by evenness?
 
Aleksey Vyazmikin #:

I am correcting for the deviation from the regression line -1.92262 , do you just want to adjust the balance?

Then here's the balance.

1 0 4,075 -4,075 0
2 5 4,4929 0,5071 6,5
3 7 4,9108 2,0892 9,1
4 5 5,3287 -0,3287 3,5
5 3 5,7466 -2,7466 2,1
6 8 6,1645 1,8355 10,4
7 10 6,5824 3,4176 13
8 9 7,0003 1,9997 11,7
9 8 7,4182 0,5818 10,4
10 7 7,8361 -0,8361 4,9
11 9 8,254 0,746 11,7
12 8 8,6719 -0,6719 5,6
13 9 9,0898 -0,0898 6,3
14 8 9,5077 -1,5077 5,6
15 9 9,9256 -0,9256 6,3
Wrong again.
line2= line1 + deviation * 0.3
That's what it should be:
0 4.075 -4.075 0
2 5 4.4929 0.5071 5,15
3 7 4,9108 2,0892 7,63
4 5 5,3287-0,3287 4,89
5 3 5,7466 -2,7466 2,18
6 8 6,1645 1,8355 8,55
7 10 6,5824 3,4176 11
8 97,0003 1,9997 9,6
9 8 7,4182 0,5818 8,17
107 7,8361 -0,8361 6,75
11 9 8,254 0,7469,22
12 8 8 8,6719 -0,6719 7,8
13 9 9,0898 -0.0898 9.03
14 8 9.5077 -1.5077 7.5
15 9 9.9256 -0.9256 8.724
 
mytarmailS #:
What do you mean by uniformity?

A straight line corresponds to perfect uniformity. The difference from a straight line is unevenness. I want to minimise unevenness.

 
Aleksey Vyazmikin #:

How is it easier to calculate this outrage can you tell me? :)

I have nothing against smooth balance, on the contrary, with both hands in favour!) The problem is that your indicator depends on the order of sampling, and this is not typical for MO. Sharpe, for example, does not depend on the order of transactions. I should look for articles with similar loss functions, because my intuition suggests that it is not so simple.

 
Forester #:

A straight line corresponds to perfect uniformity. The difference from a straight line is non-uniformity. I want to minimise the unevenness.

Take the ideal curve and calculate the correlation with the balance...
That's it.
 
Forester #:
Wrong again.
line2= line1 + deviation * 0.3
That's what it should be:
0 4.075 -4.075 0
2 5 4.4929 0.5071 5,15
3 7 4,9108 2,0892 7,63
4 5 5,3287-0,3287 4,89
5 3 5,7466 -2,7466 2,18
6 8 6,1645 1,8355 8,55
7 10 6,5824 3,4176 11
8 97,0003 1,9997 9,6
9 8 7,4182 0,5818 8,17
107 7,8361 -0,8361 6,75
11 9 8,254 0,746 9,22
12 8 8 8,6719 -0,6719 7,8
13 9 9,0898 -0,0898 9,03
148 9 9,5077 -1,5077 7,5
15 9 9,9256 -0,9256 8,724

OK, you can do that - the result is the same by interpretation


 
Aleksey Nikolayev #:

I have nothing against smooth balance, on the contrary, with both hands in favour!) The problem is that your indicator depends on the order of sampling, and this is not typical for MO. Sharpe, for example, does not depend on the order of transactions. I should look for articles with similar loss functions, because my intuition suggests that it is not so simple.

I am not in favour of random mixing of all lines - only standard metrics can be used that way, without taking balance into account.

1...302730283029303030313032303330343035303630373038303930403041...3695
New comment