For those interested in making money, and not imposing their model on the market, I'm re-posting Bundlemaker's answer...
Here's what to do: (1) Write down the important preconditions for the current price. It could be something that happens once you're in a trade, or while you are stalking, whatever. Simple example: the stock closed at a 52 week high. (2) Find and print out charts of all the occurances and what happened after, on the scale appropriate for your trading (3) Count how many did the "good thing" and divide by the total number. That's your probability. That process will give you something a whole lot more useful than "brownian motion with drift" formulas. (For that times that model is appropriate, I wouldn't want to be in the market anyway.) To get more advanced you could start making histograms of the distributions. I have found many useful things about the markets this way.
Pete, Good you did. Bundlemaker writes though: "I'd recommend studying statistics, probablity, and game theory..." So I presume that he is planning on modelling something related to the market. 'Imposing a model' is not what he talked about. Be good, nononsense
Some may find this useful: Computer-Assisted Statistics Teaching ~CAST It's the first link amongst the others.
The model assumes that the stock price S follows a geometric Brownian motion dS/S=m*dt + s*dz, and snorm is the cumulative normal distribution function (implemented by a known polynomial approximation). Regards
Read the John Bender piece on Stock market wizards. Very interesting piece on modelling your own prob dist. It is only about 2 paragraphs but really thought provoking. Main gist is that if stock has been in a range and suddenly it breaks out, the prob dist gets skewed tremendously to one side and that option spreads can be constructed to tilt the R/R odds in your favor.
After digging through two boxes that I haven't been through in five years since moving from Chicago, I realize that It never could have been in either since it was just published a few years ago nor have I ever onwed it. However, I did find my Cottle book.
for calculating confidence intervals which variables are better to use... a constant of 252 days or 360?? and also when using volatility is it better to use historical or Implied volatility for equities?? I am currently using 360 for day of the year and implied vols. If anyone has backtest these variables which ones turned out more promising ?? it might not be a major difference but just curious.
252 is better, 'cause that is the actual number of observations that you get in a year. As for volatility you can use both to get a better estimate.