Alex – thanks for the link to your function. I was doing some GA and neural network myself a while back but had a heck of a time beating buy and hold no matter how many neurons I threw at the historical data! At best maybe I was just going after securities that were too “efficient” (a.k.a. SPY). At worst even the lightly traded small caps may be efficient!

-lumi

I’m actually trying to do backtesting sims in matlab using several TA indicators in combination with genetic algorithms for optimization and forecasting.
In case you’re interested and want to pull resources together just drop me a line.

I have a little question:
How can I pull the Date & Time correspondence to the moment I checked the stock value ? The time must come from Google at the same we check the stock value.

Thanks

I used to do a lot of backtesting sims in Matlab. My first ones were based on using common metrics to trigger buys / sells (like simple moving average, MACD, stochastic, Darvas boxes).

My structure was similar to what you described, start time, stop time, etc. But I also had a lot of for- loops so that I could vary the algorithm parameters over a wide range to see what worked best. (e.g. 2-day moving average, 3-day, 4-day, …, 500-day)

I’d keep track of profit and loss, including $9.99 trading commissions, and then compare the result to buy-and-hold for the same ticker. I’d also factor in taxes (my trading algorithm would usually incur short-term capital gains while buy-and-hold incurred long). Finally I’d compute mean and standard deviation of monthly returns just to get insight into whether a potentially profitable trading algorithm was taking a lot more risk.

I’d also plot the P/L of the individual trades just to see if the gains happened to come from 1 or 2 homeruns, or just a bunch of tiny mediocre trades that added up nicely.

To be honest I had a heck of a time finding any strategy that would always work, mostly work, or at least work more often than it didn’t work when compared to buy-and-hold. An efficient markets theory person would say “Of course!”.

But it’s fun to keep looking so lately I’m pursuing time series forecasting methods to see if they might bear fruit. Happy hunting!

-Lumi