Search
× Search
Tuesday, November 19, 2024

Archived Discussions

Recent member discussions

The Algorithmic Traders' Association prides itself on providing a forum for the publication and dissemination of its members' white papers, research, reflections, works in progress, and other contributions. Please Note that archive searches and some of our members' publications are reserved for members only, so please log in or sign up to gain the most from our members' contributions.

Testing for Market Efficiency at Different Frequencies

photo

 Sercan K., Software Systems Engineer at Foreks

 Wednesday, September 24, 2014

Hi, Currently we have machine learning supported trend following system works excellent at the times when turnovers are more smooth, but in the times like this, it is frustrated to see how we miss opportunities just because we respond late because we use 15 minutes data. I would like to try higher frequencies on like tick data or 1 min data on our model. But first thing I am trying to see is; To show that those frequencies are more predictable. So I did Augumented Dickey Fulley Test on log returns of different frequencies of futures we trade and observed that as frequency goes to tick data, augumented dickey fulley test results become more negative like 5 times or higher. Do you think it shows that tick data is more predictable&profitable with a good model . What methods would you recommend for testing efficiency of market?


Print

5 comments on article "Testing for Market Efficiency at Different Frequencies"

photo

 Matthias W., Software Designer at Carmeq GmbH

 Thursday, September 25, 2014



Adam, indeed, all these bars are simplifications by reducing available information. Many variants of such bars would even change their shape when initialized at a different time.


photo

 Adam Cox, FFIN, MFTA, CFIP, Proprietory FX Trader

 Thursday, September 25, 2014



I agree Matthias. You need to be careful to ensure ergodicity of your data.


photo

 Sercan K., Software Systems Engineer at Foreks

 Thursday, September 25, 2014



Hey thank you both for your great answers.

Yes I am testing log returns of monotonic tick by tick data vs regulary spaced tick data.

2 -) Right like you say volatility changes over time and test shows that it has a lot of noise. I guess I cant use tick by tick data.

Matthias thank you for this great idea I definitely try it :)

By the way in our model, we regularly train including transaction volume of 15 minutes


photo

 Simon Thornington, Quantitative Developer at FINCAD

 Monday, September 29, 2014



http://www.amazon.ca/gp/aw/d/0122796713# has a lot of information about how to efficiently work with inhomogeneous high frequency tick data, including its scaling properties, the bias introduced by bid-ask bounce, intraday seasonalities and so forth.


photo

 Sercan K., Software Systems Engineer at Foreks

 Monday, September 29, 2014



This is what I've been looking for for a while thanks a lot Simon

Please login or register to post comments.

TRADING FUTURES AND OPTIONS INVOLVES SUBSTANTIAL RISK OF LOSS AND IS NOT SUITABLE FOR ALL INVESTORS
Terms Of UsePrivacy StatementCopyright 2018 Algorithmic Traders Association