I have been researching price movement modeling recently and noticed that there is notable difference between daily and intraday modeling techniques, seemly due to the increased volatility inherent to intraday price movements. Geometric Brownian motion seems to be a frequently discussed method, but from what I understand it has its limitations as well particularly in regard to intraday. I am still learning the subject but was hoping someone may be able to point me in the right direction when it comes to more appropriate intraday methods. A lot of what I have seen appear to work off of a set starting price and then take into account various static factors that could affect the outcome (sounds logical) and then this is run through something like a Monte Carlo simulation. What I haven't been able to find as much information on is methods that take expected movement ranges and then factor in conditional changes as the price movements unfold as the day progresses. Essentially something more along the lines of a combination of expected value and conditional probabilty If anyone knows about this type of work could you possibly point in the appropriate direction so I can further research the subject ?
For understanding how conditional probability works in poker I would suggest 'The Signal & The Noise' Chapter 10. This is a great foundation for the challenges that lie ahead in applying 'conditional' thinking to trading. I would encourage developing an orientation towards understanding the behavior that causes price to move - behavior of both traders as well as of systems. Once you get the underlying principles, you can slice and dice the numbers all you like. For systems - 'More than you know' by Michael Mauboussin http://www.traderslaboratory.com/forums/wyckoff-forum/15385-wyckoff-auction-markets.html
The shorter your time frame, the closer the fractal dimension is to 0.5 i.e. random motion. Some very short period data can even be anti-persistant i.e. reverts more than a normal Gaussian times series would.
There seems to be variations that were intended to track/predict the more nuanced movements of an instrument and then kind of at the other end of the spectrum methods more focused on the probability of an expected value being achieved. I personally am more interested in coarser approach of the probability of an expected value being achieved ... I was just hoping to find a way that would take the evolving of the process to achieving that particular price into context to help determine if events along the way were undermining the probability of such a value being achieved or increasing the likelihood.
ex: Calculate the number of standard deviations "x" that a future price target X of 450 is from the current price S of 444.27 in 35 days, with a volatility of 11.61%. x = log(X/S)/sigma*t = log(450/444.27)/0.1161*sqrt(35/365) = 0.3565 in Excel. = normsdist(0.3565) = 0.64 or in other words a 64% probability of the price ending below 450, 35 days from now. Assumes a perfectly Gaussian return distribution with no skew or kurtosis
OP You are basically screwed I acknowlege that you can see "stuff evolving". That is you personally mention the concept of something going from A to B. Harken back to some math classes you may have taken. Notice that if you followed that learning you could reckognize that you are using the wrong variable to do your market analysis. Yuor respondents so far have the came, unknown problen, to them. Probably you better consider doing some other kind of work where you are paid to follow doing what you are told to do.
The probability can be 99% but the prices can still make a U turn or keep going in certain direction. Mathematically probabilities can only be a confidence factor for the trader. Learning the price behavior is more important as to what is causing the price to act in a certain way in certain security. That is still subjective IMHO and at my level of only little mathematical background and nearly nil in maths based modelling.