Well I had a great holiday period probably running about 10 or 15 thousand runs on veritrader and as a result I have some problems/ queries .. Some of these may have been talked about elsewhere but I'm sure theyr'e important enough to re - open
If a system is Robust then it should work across a broad spectrum of markets ... possibly random to the point of a mixed selection of uncorrelated liquid markets.. say 2 or 3 markets from each sector in futures , currencies , etc
I dont see how one can put together a portfolio based on testing of past results since who knows where the next big trends will be .. therefore I start with the assumption of a broad base of markets .. all liquid and diversified.
Based on this I fing its extremely difficult to achive any results .. I hardly saw an MAR above 0.6 let alone 1 2 or 3.
However obviously If i took certain markets out and put others in , the performance dramatically improved .. however this is optimastion.
I also believe that if a system is robust and gives you a small edge over the long term then the more markets you trade the better the results .. compare to roulette .. if I have an edge over the house then I would play on 100 tables , not just 10. However I find again that results seem to diminish with application to more markets ?
I tested a handfull of random portfolios , each consisting of 20 - 30 instruments and saw nothing impressive at all.
When I optimised the portfolio to 18 markets that seemed to behave well then I had a much improved performance.
However I most certainly wouldnt apply this to the real world since , for example that ''optimised portfolio excluded products like live cattle and soybeans .. which without doubt have been the best trades this year.
I feel a little stalled in my analysis then since portfolio selection based on past performance is defunct by nature ?
Any views or help would be greatly appreciated.
Regards
Chris
Some Testing Results/Queries
Chris, you say
"I hardly saw an MAR above 0.6 let alone 1 2 or 3.
However obviously If i took certain markets out and put others in , the performance dramatically improved .. however this is optimastion."
This is my experience completely. yp and the guys at Castle had somebody do a pretty good job of programming in the Turlte Rules as c.f. has them into Trading Recipes. They said about the highest they were seeing was a MAR of 1.5 on a non-optimized portfolio. This is in line with the real life performance of many of the "turlte" CTA's too.
No offense, but I think this optimized nonsense of MAR's of 7 and 8 etc. is such BS. Call yp, ask him about the Trading Recipes results. I appreciate this forum, but it has almost started to borderline on hype in some cases!
Tests should be based on non-optimized large portfolios with identical parameters and tons of data. Those are the results to pay attention too. If anybody was getting MARs of 7 8 or 9 on the above paramaters they would have more money than bill gates, or be in jail for lying.
Trading Recipes can come pretty darn close to duplicating the rules as c.f. has them and the results are NOWHERE close given the "correct" way to test.
"I hardly saw an MAR above 0.6 let alone 1 2 or 3.
However obviously If i took certain markets out and put others in , the performance dramatically improved .. however this is optimastion."
This is my experience completely. yp and the guys at Castle had somebody do a pretty good job of programming in the Turlte Rules as c.f. has them into Trading Recipes. They said about the highest they were seeing was a MAR of 1.5 on a non-optimized portfolio. This is in line with the real life performance of many of the "turlte" CTA's too.
No offense, but I think this optimized nonsense of MAR's of 7 and 8 etc. is such BS. Call yp, ask him about the Trading Recipes results. I appreciate this forum, but it has almost started to borderline on hype in some cases!
Tests should be based on non-optimized large portfolios with identical parameters and tons of data. Those are the results to pay attention too. If anybody was getting MARs of 7 8 or 9 on the above paramaters they would have more money than bill gates, or be in jail for lying.
Trading Recipes can come pretty darn close to duplicating the rules as c.f. has them and the results are NOWHERE close given the "correct" way to test.