Profile Picture


Posted By hmccoy99 3 Years Ago
Posted Tuesday April 28 2015
I understand it is difficult to quantify performance but some rules of thumb are?
poor   to   outstanding
                                              poor                   good             outstanding                               
apr                                         < 5 %                    >12  %               >30%
% winners                                 30%                    50%                   >75
drawdown                                  50%                    <30%                 <10%
any comments or feed back  here would be helpful....
I have developed models that seem to perform excellent in most metrics
apr  70%  
% winners 88%
max drawdown  7%
avg bars (daily) 18
avg monthy gain 5%
this was done with selected NASDAQ stocks (approx 80 of the 100)
over a 15yr period of backtesting  with a max 20% of equity 
per trade.
my goal is to automate the system can i expect the same performace with a live system?
the paper trades indicate good trades, I have taken some of the trades indicated thru a 
broker with sucess.
my questions are?
1) woud this performance scale up  without issues?  if not what problems would i have?
2) how well is the trade management handled with the API with interactive brokers?  slipage?  missed trades? out of sync problems? etc,
3) does a cloud server make sense hosting the system for issues of reliablity ?  if so any recomendations?
4) has anyone buitlt a AI rule based modeling system interfaced to right edge platform? if not anyone interested ? (I have a background in AI research and have built real time AI systems)
Posted Monday May 04 2015
I'm working on integrating a BRMS in C# with RE : WWW.OBJECTCONNECTIONS.COM.
It's done in a pragmatic way. I extract all info from RE into dictionaries and pass it into this application.(list of securities, current positions, openpositions, working orders).
The trading rules are defined in Object connections.
Every minute a photo is taken and passed into Objectconnnections. The rules decide on buy, sell  or do nothing. If a decision is made this is communicated to RE , again using a dictionary object.
The rest is handled in RE. All entries in the dictionaries are translated into orders and executed by RE.
I've only started with a simple POC    
Posted Thursday May 07 2015
sounds like a great approach,  I developed a system similar to that some years ago. using a distributed system.  of a data base, numerical analysis system, and a rule based inference engine. a snapshot was taken of approximately 100
quantitative data parameters every second  fed to decision model of approximate 400 rules.  the model provided entry , exits, trade management, feed back to the operator when and why any actions was taken.  the design approach is still a good one but  todays software systems provides a better environment for  integration.  


Similar Topics

Reading This Topic

2005-2018 © RightEdge Systems