Data set up for backtesting

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
chrism
Contributor
Contributor
Posts: 1
Joined: Fri Oct 15, 2004 4:38 pm
Location: Memphis, TN

Data set up for backtesting

Post by chrism » Mon Nov 01, 2004 5:51 pm

After checking around, I am told that the so called "industry standard" for roll dates is as listed on this web site: http://www.pinnacledata.com/clc.html#details

Does anyone have any comments on these dates?

Roscoe
Roundtable Knight
Roundtable Knight
Posts: 250
Joined: Sat Jan 24, 2004 2:06 am
Location: Houston TX

Post by Roscoe » Mon Nov 01, 2004 5:57 pm

I use them and like them, but then again I am not smart enough to come up with anything better on my own.

Jake Carriker
Site Admin
Site Admin
Posts: 1493
Joined: Fri Sep 12, 2003 10:32 am
Location: Austin, Texas

Post by Jake Carriker » Mon Nov 01, 2004 6:48 pm

If you have backadjusting software like CSI's Unfair Advantage, you can put contracts together in a variety of different ways in order to come up with something to your liking. Looking at charts with volume and open interest plotted on them while you change rolling parameters can be enlightening.

One thing that I do is take the Pinnacle published roll dates and set up UA to roll on those dates. Now I look at each market and adjust the roll date (keeping in mind 1st notice dates, etc.) to try to get a nice smooth "liquidity curve". I also like to have somewhat consistent roll dates (i.e. roll TY and FV on the same date rather than 2 days apart if practical). It is messy and overly complex if you trade 50 markets and each one has a different roll date.

You might also experiment with other ways to trigger the rolling of positions rather than rolling on a fixed date. These may smooth the liquidity curve even more. They may also be harder to implement in live trading.

Also keep in mind that different rolling methods may significantly impact trading performance. You might find that a method that does not produce a very smooth liquidity curve produces superior backtested results. If so, you have the opportunity to ponder some questions about the importance of liquidity to your methods and the repeatability of your backtested results in live trading.

Hope this helps.

Best,
Jake

TC
Roundtable Knight
Roundtable Knight
Posts: 101
Joined: Thu Jul 08, 2004 2:12 pm
Location: Toronto, Canada

Post by TC » Tue Nov 02, 2004 8:43 am

Hi Jake

Re your earlier comments;
Jake Carriker wrote:......I look at each market and adjust the roll date (keeping in mind 1st notice dates, etc.) to try to get a nice smooth "liquidity curve". I also like to have somewhat consistent roll dates (i.e. roll TY and FV on the same date rather than 2 days apart if practical).
I assume you want consistent roll dates to simplify trade execution and a smooth liquidity curve to minimize slippage but how coincident are these two objectives and what takes precedence when they dont ?
Jake also wrote:......
You might also experiment with other ways to trigger the rolling of positions rather than rolling on a fixed date. These may smooth the liquidity curve even more. They may also be harder to implement in live trading.
What are your views on the following methodologies:

1) Roll on Volume
2) Roll on Open Interest (OI)
3) Roll on Volume & OI

Thanks

Tom

Jake Carriker
Site Admin
Site Admin
Posts: 1493
Joined: Fri Sep 12, 2003 10:32 am
Location: Austin, Texas

Post by Jake Carriker » Mon Nov 08, 2004 10:39 am

Hi TC, my views on the rolling methods you mention are that I like to test them all and see how I feel about the results. If a trader wants to pay the narrowest spreads and incur the least slippage while rolling, he might look for the point where liquidity shifts from one contract to another (i.e. the time when everyone else is rolling). If a trader has some kind of a system that gains an edge by not rolling when everyone else does, another method may be more appropriate.

Mostly, I keep in mind that results are determined more by whether or not I stick to my system than by the details of my rolling algorithm. Therefore, when the "optimal liquidity" roll point conflicts with the "keep it simple by rolling on a common date" roll point, I take inventory of my feelings and let them help me decide which one I can consistently implement in my trading.

Best,
Jake

Kvadrik
Full Member
Full Member
Posts: 15
Joined: Sat Jun 19, 2004 3:08 am
Location: USA - Israel
Contact:

Backtesting and simulation daytrading strategies

Post by Kvadrik » Fri Nov 26, 2004 9:19 pm

The problem is not only in back testing for optimization your strategy, but also in the construction of the model and in the realistic simulation of the order execution. You need to include into the model the order queue management on the Exchange, slippage during market order execution and some other conditions. I analyze strategies using Strategy Tuner (product of Strategy Runner company). It allows simulating these conditions, which makes the results more reliable. :idea:

sluggo
Roundtable Knight
Roundtable Knight
Posts: 2986
Joined: Fri Jun 11, 2004 2:50 pm

Post by sluggo » Fri Nov 26, 2004 9:38 pm

I take a look at different rollover "triggers" and inevitably they produce different backtest results. In some cases/markets the differences are small and it's easy to conclude that it doesn't matter what rollover trigger you choose, they're all pretty much the same.

Other times the differences are large (I say again: large) and then you are faced with that horrible question: to think, or not to think.

If you choose not to think, which is your privilege and right, you just go with whatever rollover trigger gave the most pleasing backtest result and you sleep soundly.

On the other hand, if you choose to think, you have the joy of creating hypothetical scenarios in your mind that potentially answer the question "WHY did rollover trigger A significantly outperform rollover trigger B, and is this blind luck / random datamining? Or is it a structural "fact" of the market I'm trading?" Then you get to test those hypotheses and think some more.

I don't claim to have figured it all out, but for now I'm placing my bets and trading my money on the Chuck Branscomb hypothesis that some markets "need to be traded" in the further-out contracts. Just as an earlier message on this site discussed John W. Henry's decision to trade the Crude Oil & Energy markets using further-out contracts, I do the same thing in (other non-energy markets). I could be wrong. But I've got my bets down and I sleep soundly.

TC
Roundtable Knight
Roundtable Knight
Posts: 101
Joined: Thu Jul 08, 2004 2:12 pm
Location: Toronto, Canada

Post by TC » Sat Nov 27, 2004 11:50 am

sluggo

Assuming you have lower liquidity and therefore potentially higher slippage relative to trading the front month, what additional benefits specifically accrue from trading non-front contracts ?

If trading long-term do you benefit from backwardation and lower trade transaction costs (comm. + slipp. x # contracts in trade) or perhaps lower volatility allows for tighter stops ?

How easy is it to backtest ie ease of constructing continuous contracts ?

One aspect of robustness, that helps me avoid curvefitting, is the application of consistent rules across all markets. My natural inclination therefore is to apply this rule to contract rolling as well.

Could custom contract rolling for each market be considered a form of curve fitting ? (A question and not a criticism from someone still wrestling with this issue :D )

Tom

ksberg
Roundtable Knight
Roundtable Knight
Posts: 208
Joined: Fri Jan 23, 2004 1:39 am
Location: San Diego

Post by ksberg » Sat Nov 27, 2004 1:05 pm

TC wrote:Assuming you have lower liquidity and therefore potentially higher slippage relative to trading the front month, what additional benefits specifically accrue from trading non-front contracts ?

[...]

One aspect of robustness, that helps me avoid curvefitting, is the application of consistent rules across all markets. My natural inclination therefore is to apply this rule to contract rolling as well.
Some markets naturally have contract months that trade more than others. In these markets it can make sense not to trade the front contract month and eliminate unnecessary roll overs. December Eurodollars is a good example with plenty of liquidity. I don't think of this as curve-fitting so much as a trader's common sense. On the other hand, it would be interesting to see what difference it makes.

Cheers,

Kevin

leonardo
Roundtable Knight
Roundtable Knight
Posts: 100
Joined: Wed Apr 30, 2003 2:27 pm
Location: Minneapolis, MN
Contact:

Post by leonardo » Sat Nov 27, 2004 2:20 pm

Roll Dates

People engineer trading systems so as to make money most efficiently. Unfortunately, some researchers have made roll-overs part of their curve fitting, usually accidently. I think it has become less of an issue as data streams used for testing have become more standardized.

If it can be statistically proven that increased profits are probable by extending the roll date for a current trade approaching the switch/delivery period, you now have another edge or even another potential system. For example, I've found that holding past the "typical" roll-over time on largely profitable physical commodity positions can result in substantially increased profits with little additional risk. {READ: statistically significant)

As futures prices become the cash in the delivery month; if there has been a big spread between the two you can receive outsized additional gains into the delivery period. If you are short, you do not have to cover until the expiration of the contract (not a very good idea to wait too long because of very low liquidity issues) while a long position risks getting delivered unless there is a fierce demand for the commodity.

Systems also exist for trading the spreads of nearby vs. deferred contracts. I monitor them also to indicate these squeeze or anti-squeeze situations, as do the market makers at the exchanges.

Leonardo---

Post Reply