I guess a number of people have looked at cycles and found that they cannot rely on them. The reason of course is that they set out to find fixed cycles and also the fact that unexpected events create havoc with one's projections. That's life. After all, the market is dynamic . The only way to handle that is to analyse often (i.e. at least on a daily basis) and to forget about the idea of fixed cycles. Having faith in the existence of fixed cycles is dangerous to one's financial health. I nearly said 'Look and you will find' except for the fact that that smacks too much of a bible quote. BTW, W.D.Gann you talk about charting by hand. That would limit you to only a small number of charts doesn't it ? That would therewfore eliminate a lot of related information. So how do you handle that ? Jack
I also have the Time Series package, the Wavelet package, and the Signals and Systems package. However, 99% of the stuff I have done with Mathematica realted to trading is just doing some programming in the language... nitro
I know several that were offered these kinds of positions and declined them. For them it is really boring to go from Particle Physics of Algegraic K-Theory to go to a job where finding a better volatility model is the hilite of their day. What is interesting about these "Wall Street Firms" is that from all that I know about Quants, they aren't all that interested in these guys because of their _trading_ talent, but because they simply want them to apply their skills to finding more and more ways to stay _market_neutral_ and pick up nickles and dimes with little or no risk. This is _not_ trading (although I wish I could do it ) nitro
Yes K theory is real. Without going into the details - I am not sure i could explain fully without generating an endless string of jargon .... Essentially a way to catagorize and decompose mathematical spaces of functions. The physics part of it is that in certain theories you need to know which types of mathematical structures allow your ideas to be expressed. Algebraic K-Theory is a type of union of algebra and analysis ... Probably enuf said. If you really want to know more there are lots of introductory books but they require a fair amount of background ....
Worked as a type of rocket scientist in these areas .... my recollection is that the quants were never really in charge. I dont view trading skills as being commensurate with your modeling skills: the modelling / quantitative stuff helps but I think my trading relies on a lot more general types of knowledge. If it were just math then I could automate everything and walk away
same here .... The time series package is somewhat useful. ...Almost all of my use of mathematica is using custom modules, packages, and workbooks that I have developed over time and that hook into some external programs as well ......
http://www.math.rutgers.edu/~weibel/Kbook.html http://www.math.niu.edu/~rusin/known-math/index/19-XX.html nitro
Lots of good advice from nitro and others. I only have a few more things to add: You can do quite of bit of analysis in Excel. After going down the path of genetic algorithms, chaos theory etc., I've eventually come back to "simple is good". I was never really able to find much truly profound information with the more advanced mathematics; hence I've stopped using it. I'd recommend simple things like linear regression for trend, volatility to determine mode and some combination of the two to determine extremes. That's really all I use.
I like your idea of testing all sorts of patterns and pattern combinations to see if any have predictive value. In my testing though, I've found that if you include enough patterns to combine, you very quickly start to find random combinations that perform extremely well in sample, but fail out of sample. For example, I just ran the following test: Looking at SP data from 1983 to 2001, I looked for bar pattern combinations that in each year exceeded 75% of random entries/exits (a test that you previously suggested) when buying at open and selling at close. The pattern matcher came up with 10 patterns, of which 6 did decently well over 2002-2003, and 4 fell apart completely, which is in line with what one would expect of any ten random patterns. How do you separate the wheat from the chaff, when looking at these patterns and decide which are statistically significant, and which are just random flukes? It seems that no matter what test is applied, some meaningless patterns always qualify simply as a matter of chance. -bbc