"Symmetry" stat is currently not the best way for identifying trades symmetry
Currently, this stat only seems to consider short PL vs long PL. Sometimes 99% symmetry is reported for strats that traded 200 long trades and 1000 short trades for instance just because longs and shorts had the same profit. Can we redefine this stat to be more useful?
I think number of long trades vs number of short trades should carry a fairly heavy weight maybe as much as 50% of the weight and I think the other 50% of the weight should be from a fitness algorithm.
I think number of long trades vs number of short trades should carry a fairly heavy weight maybe as much as 50% of the weight and I think the other 50% of the weight should be from a fitness algorithm.
OR maybe two or three separate columns would be the most practical: P/L Symmetry (as it is now), Number of trades Symmetry and maybe eventually a Fitness Symmetry. Then it wouldn't mess with peoples existing filters.
Attachments
No attachments
-
Votes +2
-
Project StrategyQuant X
-
Type Feature
-
Status Moved
-
Priority Normal
-
Assignee None
History
b
b
KL
kainc301
19.08.2019 16:45
I think the second option you presented is more practical where you would have different symmetry stats. One for P/L, one for # of trades and so forth. It would give more options to compare the long and short side more easily.
KL
b
bentra
19.08.2019 21:49Description changed:
Currently, this stat only seems to consider short PL vs long PL. Sometimes 99% symmetry is reported for strats that traded 200 long trades and 1000 short trades for instance just because longs and shorts had the same profit. Can we redefine this stat to be more useful?
I think number of long trades vs number of short trades should carry a fairly heavy weight maybe as much as 50% of the weight and I think the other 50% of the weight should be from a fitness algorithm.
I think number of long trades vs number of short trades should carry a fairly heavy weight maybe as much as 50% of the weight and I think the other 50% of the weight should be from a fitness algorithm.
OR maybe two or three separate columns would be the most practical: P/L Symmetry (as it is now), Number of trades Symmetry and maybe eventually a Fitness Symmetry. Then it wouldn't mess with peoples existing filters.
Type changed from Bug to Feature
b
KD
kejdi
04.11.2019 09:01Attachment Columns-Databanks-SymmetryNT.sxp added
Attached snippet that shows the 'Number of Trades Symmetry' I created for myself - maybe you will also find it helpful. Enjoy.
KD
kejdi
04.11.2019 13:50Attachment Columns-Databanks-SymmetryNT.sxp added
Updated version - fixed dependency to increase reliability.
MF
Mark Fric
20.10.2020 14:50
Task moved to project SQ Programming, new link to task: https://sq.projectpanel.com/tasks/sqp_0112
Votes: +2
© Copyright. All rights reserved. ProjectPanel.com
Rate the Longs_vs_Shorts_fitness_comparison_score from 0 to 1.
Symmetry = 1 * Longs_vs_Shorts_Total_trades_comparison_score * Longs_vs_Shorts_fitness_comparison_score
That way a very low score in either one results in a very low Symmetry score. What do you think?
Or something more basic like 2 or 3 different separate symmetry stats: P/L Symmetry (as it is now), Number of trades Symmetry and maybe eventually a Fitness Symmetry. Then it wouldn't mess with peoples existing filters.