Quote from rcanfiel:
advise avoiding optimal f. Thy Bob Spears, fixed fractional & possibly fixed ratio...
Advise using none of the above. Using backtested results
to determine any quasi-Kelly fraction is a recipe for disaster.
Ralph Vince has neither mathematical training nor aptitude
and Ryan Jones is a complete swindle. Avoid them like the
plague.
Start trading with a constant rebalanced portfolio where the
proportion in each system + the risk free asset (cash) is 1/n.
If you have three systems plus cash that is 25% initial allocation
to each.
This allocation can be made to converge exponentially towards
the optimal rebalanced portfolio by updating the weight vector
proportionately with a ratio of two integrals. In practice the
integrals are replaced by summations over a reasonably fine
grid over the n-1 hyperplane.
The procedure is given in Cover, 1991, and (revised) 1996.
Complete with detailed examples.
This is an online algorithm meaning it uses no past curve-
fitted data to achieve its results. It is orders of magnitude
safer than the suggestions above.
Good luck.