What's your wishlist for an ideal backtesting program?
- Thread starter krosny
- Start date
More sophisticted tests of statistical significance than Z-score. (My apologies to ET for playing straight man to this spammer, but those are good questions. I doubt that his tout (haha!) has what I asked for.)
Carry on, good spammer!
Carry on, good spammer!
My wishlist pertains to the ideal programmer, not the actual testing itself.
I'd love to meet a programmer who isn't a failed trader and who understands that I don't CARE about his creativity with regard to the market - who just does what I want done and keeps his mouth shut and ridiculous ideas to himself.
Oh, and who doesn't try to basket test everything - I don't expect/want my strategy to work for the entire Russell 2000. I wish these guys would realize how really trite they are.
Sorry for the rant
I'd love to meet a programmer who isn't a failed trader and who understands that I don't CARE about his creativity with regard to the market - who just does what I want done and keeps his mouth shut and ridiculous ideas to himself.
Oh, and who doesn't try to basket test everything - I don't expect/want my strategy to work for the entire Russell 2000. I wish these guys would realize how really trite they are.
Sorry for the rant
1. Make it as easy to modify the composition of a portfolio, as it is to modify the value of a parameter. Treat the portfolio composition as an optimizable variable.
2. Make it easy to systematically withdraw cash from the account using some user-specified algorithm. Handle the common situations of income taxes, management fees, and profit incentive fees.
3. Provide primitive builtins which return (A) # of trading days till end of this calendar month; (B) # of trading days till First Notice Day; (C) # of trading days till Contract Expiration.
2. Make it easy to systematically withdraw cash from the account using some user-specified algorithm. Handle the common situations of income taxes, management fees, and profit incentive fees.
3. Provide primitive builtins which return (A) # of trading days till end of this calendar month; (B) # of trading days till First Notice Day; (C) # of trading days till Contract Expiration.
1. The ability to apply position sizing based on a partial-Kelly criteria with the parameters (%wins and W/L ratio) calculated off of the last N trades.
2. Integrated Monte Carlo simulator.
3. Apply indicators to one equity curve (already mentioned by another poster), and use the output to filter or size trades to generate a second equity curve.
2. Integrated Monte Carlo simulator.
3. Apply indicators to one equity curve (already mentioned by another poster), and use the output to filter or size trades to generate a second equity curve.
A program turning your screen to red if it's bad;
turning your screen to green when it's good;
turning you screen to flashing yellow if it's up till now never test Jack stuff.
turning your screen to green when it's good;
turning you screen to flashing yellow if it's up till now never test Jack stuff.
Quote from FaderTraderr:
My wishlist pertains to the ideal programmer, not the actual testing itself.
I'd love to meet a programmer who isn't a failed trader and who understands that I don't CARE about his creativity with regard to the market - who just does what I want done and keeps his mouth shut and ridiculous ideas to himself.
Oh, and who doesn't try to basket test everything - I don't expect/want my strategy to work for the entire Russell 2000. I wish these guys would realize how really trite they are.
Sorry for the rant
The only way I found to solve that problem is to become a programmer yourself!!! It takes a lot of time, but it has worked wonders for me and my trading.