After working three years, I finish a quant trading system according to my approach which(see attachment)was found ten years ago. It is a pure quant trading system which is base on some new statistic rules which do not include The Large Number Law. In fact, if you find some probability relating to a pattern in financial data, this probability will change with time; it is a great challenge to develop an algorithm to deal with fluctuating probability. My basic view about financial market is: the financial price doesn¡¯t move in random way, but in the way that contains inevitability and random factor, and their relationship is complementally. So the probability of market direction fluctuates round 50%; you may bet the market direction at some favored chance, and may make profits by lot of these favor trading. Let me simply introduce my approach here: The market data is divided to different trends at certain levels, and a special trend ¨Can uncertain trend is defined, that would shift to an up trend or a down trend. What is concerned about is the probability p under which an uncertain trend shift to up (down trend), and f(x) that is target distribution of the up trend¡¯s end. There are two critical points: L1 and L2; if the price rise above to L1, that means an up trend is forming, and the price go down through L2, another down trend is forming. For example, an up uncertain is forming, current price is q, with f(x) the expectation e of the up trend target is calculated; if we decide to make a long position, we know that the profit is: (e-q)*p-(q-L2)*(1-p). According to probability theory, the ruin probability is calculated with these data that is risk for trading. Profit and ruin probability are most import parameters for decision making. By mean of statistics tool, p and f(x) is calculated with historical data, and the Bayesian Theorem is used as basic rule. Let P/L = (e-q)/(q-L2); P/L and p would determine ruin probability. Because ruin probability is not very exact, and the decision making theory is not very effective, the system uses minimal p and minimal P/L as basic trading parameters; and another trading parameter is allowed maximum loss ratio (that equals to loss/total capital) for every trading which could be chosen by user; the system assigns optimizing ruin probability for each trading decision. I collect FX week bar data that start from 1978, daily data from 1999, and hour data from 2006, and minute data from 2007, they are reconstructed to simulating data. The spread for FX pair is set to 6 pips and trading fee is 5 pips for each trading, the maximum loss ratio is set to 3%; p and P/L is optimized with data that occur after 2006. The trading rule is that the profit is not added to trading capital. The simulating trade result whose data occur after 2006.06.13 to 2007.06.15: USD/CHF: 40%; EUR/USD: 2.9%; USD/JPY: 22%; GBP/USD: 11.5%; AUD/USD: 30.9%; CAD/USD: 23.6%; If having enough data, the system work at any time; and some profit of a FX pair is more in this year and lower at other year; but if interval of data time is too long, some lower level trends would be ignored by the system; this is more different from real trade. Now I have integrated MB Trading API into my system, if they give me account that would not expire, I will test it soon in real time data. I know the importance of good records in financial field. The system would be applied to FX market currently; I think it could be applied to futures, and I will do this thing late. I am not in the field of financial market, so need some advise how to apply this system to financial market. And any one wound to use this system, contact me; let¡¯s seek a way to use it. EShin.Lou E_mail: yayasprite_zsu@hotmail.com
So basically all you have there is like a 123 pattern or a breakout deal and throwing in a bunch of math mumbo jumbo. Dude have you ever even traded before? You don't need to take so long to come up with a automated system using all sorts of quant stratagies to know when a trend is developing. I would suggest watching the markets for a while and learn to react to what is happening around you instead of trying to define it all with a mathematical model. Just my 2 cents.
This is interesting. But it remains just that: interesting. If it's meant to be a quant trading system, then the paper lacks the necessary validation and verifications. If it's meant to be a theoretical paper, it lacks foundation. Keep in mind that formal finance has moved beyond gaussian assumptions and naive market efficiency long ago by either (1) finding alternative characterizations, or show that (2) simplified settings do in fact describe the target systems sufficiently well even when misspecified.
Thank for your advice. Ten years ago, I was a trader; my intuition about direction of the market is very exact; but the work of executing order made me tired; and then I developed my approach; but there was no condition to develop an automated trading system. And then I leave this field and become a software engineer. Thousand of patterns would come from 3 basic patterns; the code of all software is consist of 1 and 0; Hi, sjfan, I think you are right; but you must know that I have to keep some things in secret. What I most concern about is if my system would work at current market, so I test it every week; and will test it with real time data. Would you give me more information about formal finance approach, event if I think that foundation of most these theory is wrong.