We launched because there is no product in the financial investment software marketplace, at the time of writing in Q4 2011, which provides a comprehensive, cointegration-based arbitrage tool covering all the ground from testing to tradability. There are a couple of sites that provide what is described as the probability that a given pair is cointegrated (this may help explain why that is far from being cointegration). But we have yet to see one of these document how the number was calculated or to what degree of confidence. Did it include the right variables in the right order? What are the mean-reversion characteristics – if it has any – of the pair? Was it calibrated accurately in terms of lags? And so on.
Judge for yourself. Our current documentation can be downloaded here and will give potential clients a good indication of how the product works and if it is suitable for them.
There is also a 14 day trial available in order to road test the software before making a decision.
Much of the statistics inside the ArbMaker engine is likely to be familiar to professional traders.
For most retail investors the complexity of the stat may have put reproduction of the procedures themselves – using the likes of Matlab, R or even Excel – out of reach. However, the principles of going long/short are familiar to many. Perhaps a parallel is driving a car: we don’t need to know how to make one in order to operate one.
We aim to reply to all email requesting support or information within 48 hours and most in less than 24 hours. But there is a small likelihood that replies may sometimes take longer. It is due to this constraint that we do not try to sell 24/7 ‘Maintenance Plans’ but include support and upgrades in the subscription plan prices.
Yes - we have this Quick Guide: Principles & Practicalities which lays out the key idea and pre-requisites behind trading cointegration. It includes a few examples of what to look for and what to think twice about.
This depends what is being tested. For small samples and multivariate tests (ie anything more than two variables) the Johansen method is better. But for bivariate testing of typical runs of financial price data the Engle-Granger method has certain advantages. For example, by using a criterion of minimum variance (as opposed to the Johansen criterion of maximum stationarity) the method lends itself far more to risk/portfolio management applications. For a fuller discussion go here to read Professor Carol Alexander’s Optimal Hedging Using Cointegration paper.
It may also be worth pointing out that the Johansen method has its own drawbacks: sensitivity to lag selection, how to select the best cointegrating vector, ambiguity in the presence of conflicting t-values and (according to the Michael Wickens critique) a tendency to signal cointegration where none exists.
When you place an order you will receive two email notices in return. One with a download link and activation key for ArbMaker; and the other with links to your invoice and order details. The link to the order details leads to a page with a cancellation option you can click if that is what you decide to do.
You will also be sent an email several days before the end of the 14-day free trial to remind you that the trial is ending. This is designed to avoid any surprise charges being made.
Finally, if you have any problems trying to cancel you may also ask us to do it for you.
These errors indicate that for the date range chosen the instruments do not have full data. For example, if the scan start date was 2000 a "Bad Data" error means the symbol in question started trading after 2000.
There may also be symbols that have stopped trading due to acquisition or delisting during the data range requested.
Finally, there are cases where although a symbol has traded the full range at some point there was a corporate event that fundamentally changed its capital structure - a merger of different classes of shares for example. When this happens it will generally not be reflected in a free feed like Yahoo! or Google. But IQ will usually carry its price only from the date of the change in structure.
Appendix I in our documentation covers the different types of error ArbMaker will flag.
The minimum free RAM required at all times for ArbMaker 3.0 is just under 3Gb. ArbMaker was coded for speed and avoids writing unnecessary raw data to its database for this reason. In beta testing a 6GB RAM Windows pc with a i7 processor will download price data from Yahoo! and compute them at a rate greater than 25 cointegration calculations per second over 3 year date ranges. Speed is also a function of the type and quality of the web connection and it is here bottlenecks tend to form.
There have always been additions in our planning that come to fruition in due course; and we continue to work in this way (ie there is always an evolving to-do list. Please contact us if there is a particular feature you would like to see.
Subscribers currently have a choice of
- Yahoo! free historical and delayed data
- Google Finance free historical and intra-day data (real time and delayed)
- IQFeed historical and real-time data
We are very interested in adding a feed to cover European markets in the same way that IQ covers North American bourses (and at a similar price). If you have leads that will help our research please drop us a line!
We implemented a trading bridge that connects ArbMakerPlus and Pro versions to the Interactive Brokers (IB) TWS platform in January 2013. This permits semi automatic and fully automatic trading of signals generated by our software via TWS.
However, we do not use Interactive Brokers data to run our cointegration algorithms for two reasons:
- IB's daily history generally only extends 1 year back
- IB 'chokes' repeated data requests thus preventing high speed calculations (IB refers to these as 'pacing violations')
So although IB have great global coverage they are not good providers of historic data for our software.
ArbMaker is not a black-box program doing the same thing for everyone. It is a tool-box traders will configure to their separate methods, markets and preferences. Some, for example, may only trade pairs cointegrated at 95% confidence while others prefer pairs well inside the 99% interval. Or perhaps some stick entirely to US markets while others mix geographies. Or maybe some trade only large-caps and other do not. And so on.