Interest rates and historical testing

Discussions about the testing and simulation of mechanical trading systems using historical data and other methods. Trading Blox Customers should post Trading Blox specific questions in the Customer Support forum.
Post Reply
budonk
Roundtable Fellow
Roundtable Fellow
Posts: 62
Joined: Tue Dec 30, 2003 7:44 pm
Location: Lake Tahoe, NV

Interest rates and historical testing

Post by budonk » Fri Jan 20, 2006 3:02 pm

is that margin/interest rates are not fixed hisotrically. if you look at the mt. lucas index (and a number of other CTA studies) you will find that the tbill return is a MAJOR contributor to managed futures returns (in many cases, about half).

so, "fixing" the margin and t-bill rates doesnt make sense to me. rather, trading blox should allow for rates that are/were consistent with the year/month of the backtest.

while commercial paper rates are about 4.5% right now....they were in the mid teens in the early 80's . ..

Paul King
Roundtable Knight
Roundtable Knight
Posts: 207
Joined: Mon Feb 23, 2004 9:13 am
Location: Vermont, USA
Contact:

Other caveats of backtesting

Post by Paul King » Fri Jan 20, 2006 3:13 pm

Budonk,

You make an excellent point, and I believe that backtesting is useful in general, but does have some drawbacks that should be taken into consideration when estimating 'real' trading performance.

These include (and are not specifically applicable to TradingBlox - I have not used this software - only read the user guide):
  • How far back tick data is available
    The quality of the historical data (compared to what actually happened)
    How contract rollovers are handled
    How slippage is estimated
    How margin utilization (including margin debits and credits) are handled
    Survivorship bias
    Corporate actions (splits, dividends, mergers, acquisitions)
    Earning restatements or other 'after the event' corrections
    Look-forward errors in system programming
All we can do is use backtesting as a useful estimation of how a system may have performed, and use it to test the relative merits of various instances of the trading system. Real money trading is a whole different game, and one should always proceed with caution (i.e. trade very small at first) no matter what kind of fantastic results one gets from backtesting (with whatever software).

Paul

Paul King
Roundtable Knight
Roundtable Knight
Posts: 207
Joined: Mon Feb 23, 2004 9:13 am
Location: Vermont, USA
Contact:

Post by Paul King » Fri Jan 20, 2006 3:31 pm

p.s. if you email me I will send you a copy of an article I wrote called 'The Main Caveats of Backtesting' which goes into more detail (but BEWARE, by revealing your email address to me you are leaving the door wide open for possible occasional trading-related emails from PMKing Trading containing special offers due to my business relationships with other trading-related firms, with the option to unsubscribe and never hear from me again of course :-)

Paul

durabo
Contributor
Contributor
Posts: 2
Joined: Wed Jan 05, 2005 10:46 am
Location: Fort Worth TX

margin int issue

Post by durabo » Tue Feb 28, 2006 8:25 pm

hey gang -- any progress on adding the interest rate/margin rate doohickey? :lol:

thanks - durabo

Angelo
Roundtable Fellow
Roundtable Fellow
Posts: 90
Joined: Fri Apr 29, 2005 4:31 am
Location: Italy

Post by Angelo » Wed Mar 01, 2006 9:38 am

Hello everybody,

About the margin amount, yes, that's a problem: just compare today's crude oil margin with those of 10 years ago.
But the solution is pretty simple ( even if it requires some programming effort): as some other testing softwares, just allow the user to input margins both in absolute value or in percent terms.
As a sort of "wish list", I propose this change for future TB updates.

As for interest income, until it will be possible to input the historical series of T-Bill rates, I suppose that is better to test without considering this income (e.g. as if interest income were zero or near zero), unless we are testing very conservative position sizing.
In this latter case, I see no alternative than the input of the average TBill rate in the period considered.
This is an approximation, but - no matter how precise we try to be - hypothetical testing will always be an approximation.

Jimbo
Senior Member
Senior Member
Posts: 32
Joined: Fri Aug 05, 2005 9:57 am
Location: Boston, Ma

Post by Jimbo » Tue Mar 07, 2006 8:54 pm

How about incorporating some of the SPAN algorithims into the margin calculations? Would this be at all feasible or appropriate?

Post Reply