The Science behind AME

Top  Previous  Next



The challenge in making an Expert Advisor like AME lies in developing its ability to:


Detect price formations,

Track patterns of behavior as the market changes.



Let see how AME does it:



The Science behing AME

Step1: Price Patterns

The first thing AME does is parse price data into patterns of various time-scales. During this early stage of the process AME does not know how to trade yet. So it increases its chances by looking at many different options of what the patterns could be.


Click to expand/collapseMultiple Time-Scales

Price patterns can be intricate to the point they lead to conflicting interpretations. Fortunately price can be broken at different time-scales into constituent parts that can be better handled separately. At this point, a Causal Singular Spectrum Analysis (CSSA) is performed. CSSA recognizes whether price should be decomposed and determines how best to break it down.




Step2: Trading Policies

Of course it’s not enough for AME to just identify patterns. It has to know how to exploit them. So it solicits the help of Expert Algorithms with different areas of expertise on trading policies. Hundreds of algorithms work in parallel to test each and every one of the patterns identified. At this point quantity trumps accuracy. It is more important for AME to choose from a large number of policies and narrow down from there.


Click to expand/collapseAdaptive

The Expert Algorithms are not told what they should do in any particular situation. So they must learn by themselves how the market works. Reinforcement Learning is used as the learning method; it is a technique that borrows from biology, where the behavior of an agent adapts on the fly from the amount of positive stimuli, or reinforcement, it receives.


Click to expand/collapseRisk Adjusted Returns

The change in Sharpe ratio constitutes the reinforcement signal for the Expert Algorithms; it is a measure of performance versus risk:


Sp = change of ( Rp / σ )



Rp = return

σ = standard deviation of return also commonly called "volatility risk"



A low Sharpe ratio indicates either that a trading policy is not getting returns in proportion to the risk involved, or that the risk involved is excessive by any reasonable standard. It is important for the Expert Algorithms to make that distinction so that they properly compensate for the amount of risk undertaken.




Step3: The No-Regret Games

Different trading policies are better than others at different times. AME has to know how to quickly shift emphasis from one policy to another. So just like a person learns from practice, AME puts all policies to work through practice games between the experts and the market. These games are then combined together in such a way that "regret" is minimized to produce the final ratings for the current trade. The term “regret” refers to how sorry we are in retrospect, for not having followed the best system available to us. This way we do well no matter what the market does.





© <2012> <Noxa Analytics, Inc.>. All rights reserved.

Noxa-AME Home Page ι Buy License ι Welcome to our Blog ι Ask your question