Neural networks, genetic algorithms, swarms etc. etc. have their day and then fade into obscurity. Only a few hard core devotees keep plugging on with them. Do they make huge profits ? Wish I knew Still looking for someone to try my biological semi-intelligent computer set-up ??????????? ( prolly fail tho )
'I recommend 2 semesters of college-level statistics and 1 semester of calculus.' Agree, and with these at hand one has to be able to program atleast C+(s) and Easylanguage also otherwise applying the stat and math logics into a model might become *****extremely***** cumbersome.
Mathematics don't give you that much of advantage... I know it's a "zero sum game" in most cases and most people are not even ready to share such information. To my knowledge some of the most used and perhaps best trading models/strategies/theories are as follows: - CAPM - Heston - Kelly criterion - Fama-French - Black-Scholes - BOPM - ARMA - Neural Networks My personal choices are Heston and Neural.
Of 5 people I know personally who trade, one being me, none are math or science majors. 4 of us for sure make a living or better, and the other I can only assume as he doesn't ever give insight into how well or not he is doing. 2 of us are mostly futures daytraders, 1 equity daytrader, other 2 are mostly equity swing traders and incorporate options. I don't know about the others, but I was straight A's in 3 semesters of Calc, and one Discrete Math course, but I don't really feel like it is anything I use other than perhaps the basic logic of math training. I actually believe that success in this field as most any involves the work and discipline you chose to put into it. Other training will help, but not as much as study, study, study.
For instance...: I was thinking recently about the St. Petersburg paradox and the EMH...and the Kelly Criterion. I hope you are familiar with those concepts - they aren't that hard to understand, although the logic behind the St. Paradox is a bit conterintuitive and hence you need some decent logical thinking to understant it at first sight. What, I was thinking for is a very simple model that one can use to generate high income this way. Now...since it's a weekend almost i won't go further into more scientific terms, not to mention formulas or computer code(thogh believe it or not I can do both ...the formulas and the programming code...). Anyway, here is the model: An investor enters the market. She decides to buy Nasdaq 100 for example. On every deal she puts 30 pips limit order and -20 pips stop loss order. She uses this step many times...for example thousands of times....which is clearly a whole year - but... What happens actually, using this extremely easy model of trading? The idea is simple. In the St. Petersburd paradox, Nicolos Bernoully describes a game where a player doesn't bet anything but simply flat fee to enter the game. Then a fair coin is tossed and the player wins each time a tail occurs. When head occurs eventually the game ends and the player is left only with her gainings from the previous tails. I hope you got the simple logic. The idea however is that, surprisingly this game is so promising that the player should enter the game at any price that is offered - because the potential of income is infinite. Now, how this resemebles the stock market? Suppose, every 20 pips are equal to one coin toss. This if the investor wins 20, she wins one "tail". If however, she wins "40" - this is "two tails". Because the EMH is right however - the investor has a chance of 50% each 20 pips. Keep in mind that 20 pips is randomly chosen value, but even if the value was 200 pips, then the proability is yet 50%.
For the real world you HAVE to add in commission. slippage and the fact that mms hunt stops. Taking those factors into account, I regret to tell you your system hasn't a chance
The ICAPM Adds More Realistic Assumptions About Investor Behavior The ICAPM contains many of the same assumptions found in the CAPM, but recognizes that investors may wish to construct portfolios that help hedge uncertainties in a more dynamic way. While the other assumptions embedded in the ICAPM (such as complete agreement among investors, and a multivariate normal asset return distribution) should continue to be tested for validity, this extension of the theory goes a long way in modeling more realistic investor behavior, and allows more flexibility in what constitutes efficiency in markets. The word "intertemporal" in the theory's title refers to the fact that, unlike the CAPM, which assumes investors only care about minimizing variance in returns, the ICAPM assumes investors will care about their consumption and investment opportunities over time. In other words, the ICAPM recognizes that investors may use their portfolios to hedge uncertainties relating to future prices of goods and services, future expected asset returns, and future employment opportunities, among other things. Because these uncertainties are not incorporated into the CAPM's beta, it will not capture the correlation of assets with these risks. Thus, the beta is an incomplete measure of the risks investors may care about, and thus will not allow investors to accurately determine discount rates and, ultimately, fair prices for securities. In contrast to the single factor (beta) found in the CAPM, the ICAPM is a multi-factor model of asset pricing - allowing additional risk factors to be incorporated into the equation. An interesting article by N. Riley
Looks like Greek to me but if you are into maths Finite Difference Schemes for Heston Model Lin, Sensen (2008) Finite Difference Schemes for Heston Model. Masters thesis, University of Oxford. Preview PDF 712Kb