|
PureBytes Links
Trading Reference Links
|
If you've recoded the Kelly formula in Easy Language, I'd like to see
it......I am still wrestling with all of the spreadsheet formulas. There
were several bugs in early versions of this. It's quite involved.
The problem with the Kelly approach is that key inputs are based on MONTHLY
equity changes and MONTHLY max loss tolerated. All that needs to be done is
to rework the inputs and related formulas to permit DAILY equity closed and
open positions. Then, it's "K-Ching"....you're off to the bank......maybe.
Kelly has a built-in leverage "governer" which makes it pretty conservative.
However, with one move on the keyboard, you can remove it.....and watch
those contracts go zoomin.
What's impressive about Kelly is it effectively incorporates %win rate,
win/loss ratio, and stop loss setting into the bet-size formula. Put the
formula into your system, and each trade as well as new parameters such as
stop loss value will result in an optimal bet for the next trade. At the
begin of each trading day, just feed the system updated winrate and win/loss
ratio and be sure to recalculate internally after each trade. Do this, and
you'll have a "turn-key" system.
> -----Original Message-----
> From: Bengtsson, Mats [mailto:mats.bengtsson@xxxxxxxxxxxx]
> Sent: Thursday, July 20, 2000 12:03 PM
> To: omega-list@xxxxxxxxxx
> Subject: Optimising trades using Kelly trade size formula
>
>
> I found this formula, and it looked logical and nice. So I did
> calculate it
> for a number of backtested systems, and then found problems...
>
> The formula does not care at all about how long time the trade
> takes. Making
> 1000$ in trades averaging 5 days should be more worth than making 1000$ in
> trades averaging 100 days. Is it just to exchange winperwinning
> trade and so
> on in the Kelly formula for return on account for winning trades? Or is
> there another formula that does something similar to Kelly but also taking
> number of days for the trades into account? (The Sharpe formula
> is something
> else, I am trying to get that into my system tests as well).
>
> --- Mats ---
>
>
|