No, my assumption (perhaps incorrect) that folks used to working with scientific method use standard pre-processing to tag outliers that indicate bad data. You can't do stats with bad data. It may not be mentioned as it's something so obvious that it doesn't need it. Like if you were to work with precipitation statistics when the average temperature is less than zero. The rain falls as snow and has to be divided by 10 to get precipitation measured as rain If they don't understand the domain basis of the data then you are right, they deserve a good laugh.
Market data is not random. If you want to see truly random data please look at the test symbols, ZVZZT, ZXZZT, etc... those print at random prices, one print comes at 5000 the next at 12.98 then 1500 then 4000... there's absolutely no patterns in this symbols as there is no money to be made trading them and there is no underlying asset to price [no stock, no commodity, nothing.]
Some of you probably have seen this in the Futures market. I'd like to post it here since this thread has some intelligent discussions in it. The trader profiled has been around a while and he mentions couple of things which I think are are vital to long term survival. Enjoy
Thanks for posting, I like his bet sizing. BTW. If anyone wants to try to replicate some of the arguments here. This is a pretty good simple paper. which also discusses known problems with random walk (pos and neg serial correlation, heteroskedasicity (ski daddy),fat tails, etc.). But points out some reasons about why it is a useful model. It's a step above the coin toss as they use actual Gaussian distribution modeling. http://homepage.mac.com/j.norstad/finance/ranwalk.pdf When you are really ready to be blown away, start learning to simulate stable distributions. It's an eye opener.
Chinook, Any idea why the concept of one size fits all...one architecture and parameter set to for all markets, is so popular? To me the idea is about as logical as buying only one type of shoe and then trying to go the gym, a dinner party and skiing with that one shoe. Specific market instruments have different characters, although very similar one may fit the same size shoe. If they were people they would have different weights, heights, hair color, and preferences. I could see the logic if you are trying to market your system to people. Most people probably think the idea of it "works on everything" so it's a kind off Swiss Army knife is a positive. However the profits cited are pretty anemic, at least by Futures and Forex standards Jerry030
remember, my anti-trend rants are directed at equities, there is some evidence of trend behavior in some commodities. just wanted to clarify. regards, surf
I think in terms of parameter fitting if your system works reasonably well with one parameter set on many markets, then most likely you have a robust idea. I believe that most of it will depend on how you factor in volatility, how you position size and manage your stops and profit targets. In terms of his profits, yes they are not eye popping. However he's been around a while. And most importantly it depends of his risk adjusted returns. For instance if he can return 10% a year with only 1 % drawdown, then that's a GREAT return. Also with 10% returns, if he is managing $50 million then he's returning $5 million out of which he gets $1 million a year as incentive fee. I'll take that
yes. but when your professional hedge fund investors are expecting greater returns, get ready for the redemptions! surf