My system for the past year is unusual and complicated. It is a big spreadsheet with columns deriving from various aspects of the equity price histories, arbitrary weights as to the importance of those columns, and quantified entry/exit “instructions”. I make decisions manually by reading from the spreadsheet and inserting judgment. I now want to automate in order to back-test and optimize, but doing so is beyond me. For one thing, two of the columns measure constructs which I draw on the price charts based on visual observation. They can probably be reduced to algorithm but I don’t know how. Also, my position sizing is by feel and I don’t know how to automate that. Should I hire someone?
Need system back-tester?
- Thread starter Steve Ladd
- Start date
Yes you should. But be prepared to have a high budget, nobody serious works for less than 250$ an hour in this industry and if you have to model visual stuff you can easily count 30-40 hours of work (5-10 if it's only indicator based and no need for visualisation).
Alternatively you can learn how to program.
Platforms like TradeStation will be of little help for what you need.
My wife is an unemployed code developer, busy with our kids but willing to help. Unfortunately she knows nothing about trading. She may be part of the solution.
My platform is ThinkOrSwim.
My platform is ThinkOrSwim.
https://www.quora.com/Algorithmic-trading-How-to-get-started-building-an-algorithmic-trading-system
Google will get you more results.
If your wife is a developer/coder.. have a look at Amibroker. Pretty robust backteting platform and can do some pattern recognition. There are lots of stuff on the 'net specially from Indian programmers/traders where Amibroker seems to have a wider following. I have seen code for descending triangles,pennants,etc. I specialize in AB programming but mostly in the option space so I could consult to a limited degree after market hours. That is system compares to others like Multi, Ninja, Sierra, Linnsoft which i've all tried in my past lives BUT it gets some separation from everybody with the custom backtester interface.. where you can introduce position sizing, virtual tickers, foreign tickers,etc. oh and wicked fast.
MattZ
Sponsor
Let me give you practical advice: Coders can make a mess out of a system that does not have simple rules.
Before you get to automation, try and trade your system on a discretionary basis and you may discover that many of the things that you thought are crucial, may not add as much as value as you thought.
Your systems may require low latency execution and discretion may not be appropriate, nevertheless, you will see of if you are on the right track.
MattZ nailed it .. extra filters invite curve fitting and overoptimization.. how do i know? years of making errors- my story!... make sure u walk forward at the least
This.
Very much this.
The fact that you instinctively describe your system as "unusual and complicated" makes it pretty questionable whether it can really be worth spending much money on it. Apologies for sounding negative and saying something you didn't want to hear, but that's the reality. The more unusual and complicated, the lower the chances of success and the higher (usually exponentially!) the amount of time, effort, energy and money testing it methodically involves, so the potential "worthwhileless" of what you're asking about really decreases exponentially in proportion to its "complicatedness". Ideas that are really worth backtesting methodically and thoroughly tend to be very simple and very straightforward ones, and there are very real and valid reasons for that being so.
Last edited:
MattZ
Sponsor
Simplicity is hard because each variable added bares the question of "is it really necessary"?
This is what most traders/programmers have a difficulty addressing and answering.
The inexperienced think that each variable in trading further verifies a signal when in fact all it does is add a layer of complexity that breaks the system/method.
This is what most traders/programmers have a difficulty addressing and answering.
The inexperienced think that each variable in trading further verifies a signal when in fact all it does is add a layer of complexity that breaks the system/method.
like i have trigger x > 100 and the backtest is 45%?win w avg pnl at $10 then i flip it x < 100 and across wak fowards and it jumps fr 45-58% that is big . i am like what if i add filter y and it jumps fr 58- 60! now i am in overoptimization zone!