The automation is engaged and as I said I won't claim anything yet as long as it is not finished although I can give an idea as already posted here:
http://www.elitetrader.com/vb/showthread.php?s=&threadid=26107
"Win/Loss ratio > 3 to 10+, only 3 points stop and %>75 with AUTOMATIC TRADE ?
Yeah I know that seems totally impossible especially with 3 points of stop on dow yes you read it 3 absolute points not 3% of course . This is not a claim yet, what I will do is next year launch the programming of the Baby Trader, why the "Baby" because he will be capable to auto-learn. At the beginning the stop will be 10 points instead of 3 points but I know that statistically 3 points is possible, nevertheless I will let some room during his learning process as he will commit some errors at the beginning but as he will improve the stop will be tighten. Moreover he must be able to pick the exact top and or bottom let's say 50% of the time at the beginning of his learning and only the top or the bottom and over 70% of the time for both the top and bottom after a few weeks/months I don't know yet."
http://www.elitetrader.com/vb/showthread.php?s=&threadid=26107
"Win/Loss ratio > 3 to 10+, only 3 points stop and %>75 with AUTOMATIC TRADE ?
Yeah I know that seems totally impossible especially with 3 points of stop on dow yes you read it 3 absolute points not 3% of course . This is not a claim yet, what I will do is next year launch the programming of the Baby Trader, why the "Baby" because he will be capable to auto-learn. At the beginning the stop will be 10 points instead of 3 points but I know that statistically 3 points is possible, nevertheless I will let some room during his learning process as he will commit some errors at the beginning but as he will improve the stop will be tighten. Moreover he must be able to pick the exact top and or bottom let's say 50% of the time at the beginning of his learning and only the top or the bottom and over 70% of the time for both the top and bottom after a few weeks/months I don't know yet."
Quote from harrytrader:
"I have to decide ..." you're the King of Yourself I will accord you the right to decide what you want
"my experience in trading is that whatever sounds great usually does not work": you're right it is a conservative approach now to
pretend that your experience is universal is another thing. And why does it sounds great ?
Did I claim so ? Just look at the model's number yourself (I remind they are calculated 24h in advance so no curve fitting
even if you try to curve fit I wish you good luck
That's why my model can't be a stochastic model but a deterministic one. Moreover what my model pinpoints implicitly is that the BELIEF by many into a CHAOTIC STATE is a MYTH because it was in such state a deterministic model won't be able to make prediction. In fact the Lyapounov exponent has been demonstrated including by academic researchers to be too low for exhibiting chaos state so this confirm my claim above (a deterministic but non linear system is not obliged to be in a chaos state see the short tutorial on attractor in chaos theory http://www.elitetrader.com/vb/showthread.php?s=&postid=335367&highlight=attractor#post335367). But they have made the false conclusion or rather hypothesis, although reasonable at first since they don't know the true model, that this could mean that the market is not deterministic because they thing that such volatility can only come from chaos state but it doesn't come from chaos state it does come from the intrinsic deterministic equation itself. And so they continue with stochastic models with poor prediction power although they included long term memory effect in model like arima's model.
"tell me the sharpe ratio": first the sharpe ratio is just the basic CV in quality control - since I've been statistical process control engineer (or SPC). CV = standard deviation of measure/mean of measure. This ratio can be calculated without restriction since it's just a formula. But it doesn't mean than you can make statistical inference upon it and it is not used as for the real purpose of risk control in quality field where it is not the CV that allows to control the process but the limit controls of mean and variance so I don't attribute much utility to the CV and so to the sharpe ratio. Also I'm sorry to tell you that if there is something that is important in spc is the normality assumption because it is statistical nonsense or at least it is dangerous to make inference when it is not the case.
As said here the method requires that the candidate process fulfills some requirements:
http://www.1stnclass.com/spc_tutorial.htm
"Tutorial Statistical Process Control for the Non-Statistician:
CANDIDATE PROCESSES
This method will only be useful for high-volume (>60 pcs / hr) manufacturing processes. The process must be automatic, or at least semi-automatic, such that the human component has little or no effect on the outcome we are going to address. In other words, machinery/tooling needs to produce the actual characteristics we wish to control, not the way an operator loads a fixture, for example."
So that's what I am aiming to do now: fully automate the trading process because without that it is not reliable to claim any high statistics based on personal choice. So when it will be done I would make claim about sharpe ratio (although I won't be willing to trumpet it to too much people, today I can claim I know you don't believe it so it's not a problem
a process especially high skewed one. My interest is rather to predict and garantee statistically speaking a performance and a risk in the future and not in the past: who is interested by the past ? And to garantee future means that you must be able to make prediction or you are not able to prove rigourously if it is by chance that you make such or such performance. Let's take Larry William who made 10000% you would pretend that it is due to chance then it is funny that for 1000% and of course much lower people just don't apply the same doubt. So if I claim high performance like xxx% it will be in the case of a fully automated system where there is no human interaction. The automation has now been engaged under the form of decision tree tool it is not finished yet but it is partioned in modules so statistics should be collected by module.
