[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: puzzling probability and roulette a la Mark Brown



PureBytes Links

Trading Reference Links

Gary up'n sez:
> The problem with applying this logic to trading systems is that you 
> assume each trade is totally independent from the previous ones.  
> Many people say that's a valid assumption -- e.g. Monte Carlo testing 
> is based on that assumption -- but in my opinion that's not always 
> true.

I don't think that's an accurate characterization of how Monte Carlo
works. You need a large number of samples for the Monte Carlo to work
properly. The assumption is that *on average* over a *large* number of
samples, the system results can be simulated *as if* each trade were
independent. Big difference.

> E.g. look at the market for the last few weeks.  It's 
> been stuck in a narrow range for quite a long time.

Sure, and there have been other periods where it trended like crazy. If
your samples cover both kinds of markets the MC is a perfectly valid way
to get an idea of how variable your results will be in the future. It
just lets you play best_case/worst_case/average_case games. We know the
future will probably be nothing like the past so it's better to shuffle
the trades a few thousand times and run some stats than to just look at
a single series of numbers that may be the result of random chance.

-- 
  Dennis