Help with Correlation
- Thread starter progers82
- Start date
Pat,
It looks excellent to me. Basically the ideal correlation is as close to 0 as possible, and most cells in the matrix are around 0.2 which is very, very good (uncorrelated). There are a few rows in the .5 - .6 range, but it is hard not to have a few like that with 7 systems.
1.0 means perfectly correlated and -1.0 means perfectly inversely correlated, both are bad for different reasons; if the system is close to 1.0 with another system it is multiplying risk/leverage to some degree with no diversification benefit, and if it is close to -1.0 then it is canceling out the trades of another system too much.
But in this data everything looks great. Nice work.
-Taric
edit: If you want to rank the systems (just based on how uncorrelated they are), maybe a simple way is for each row just take the average absolute value. The lower # the better. This doesn't include returns however so this alone doesn't tell you which systems are better/worse, but does give you a good idea of how well each system is behaving as a "team player" in your portfolio so to speak.
It looks excellent to me. Basically the ideal correlation is as close to 0 as possible, and most cells in the matrix are around 0.2 which is very, very good (uncorrelated). There are a few rows in the .5 - .6 range, but it is hard not to have a few like that with 7 systems.
1.0 means perfectly correlated and -1.0 means perfectly inversely correlated, both are bad for different reasons; if the system is close to 1.0 with another system it is multiplying risk/leverage to some degree with no diversification benefit, and if it is close to -1.0 then it is canceling out the trades of another system too much.
But in this data everything looks great. Nice work.
-Taric
edit: If you want to rank the systems (just based on how uncorrelated they are), maybe a simple way is for each row just take the average absolute value. The lower # the better. This doesn't include returns however so this alone doesn't tell you which systems are better/worse, but does give you a good idea of how well each system is behaving as a "team player" in your portfolio so to speak.
if you want to get more feeling for the portfolio effect of different system take a spreadsheet, calculate the sharpe ratio for each of them and then calc the sharpe ratio of the mean of the seven daily returns.
correlation is fine, but it does not tell you about the joint movement once thighns get rough.
peace
correlation is fine, but it does not tell you about the joint movement once thighns get rough.
peace