Big Sandwich System

The thread and system are still active, however, the system does not generate signals during trending markets which is currently the case.

SliverB, I am assuming that your question stems from a fear that this may be an optimised system, and hence it may crash and burn.

To make sure that the value is not optimised I look for, amongst other things, a bell curve when the parameter is varied
http://img526.imageshack.us/img526/747/bellcurveup8.jpg

A bell curve gives you some confidence that you have not optimised or curve fitted the system to suit the data, or a particular event.

If you now feel confident that the system won't crash and burn, then don't! All systems run the risk of that; it could happen tomorrow, or it could happen years from now, you just need to keep an eye on the systems performance using specific measures.

Bought 1 contract of GBPUSD @ 20291

Nice 680 profit per contract so far. It really doesn't get much easier.

http://img294.imageshack.us/img294/5...d1aug20zd4.jpg

Closed out at the upper band for a $1057USD profit

http://img182.imageshack.us/img182/5...dresultan3.jpg

Thanks for your answer. If your happy to discuss it, I am curious to what 'specific measures' you've found useful. Most of the advise I've seen around the forums is non quantitative when discussing the issue of when to terminate a system.
I've been thinking a lot about this issue lately. What is the exit strategy for a system? How do we stop systems from 'crashing and burning' and rather give them soft landings?

Sold GBPUSD @ 20441


The measurements of the performance of a trading system could be as complicated and vast as you desire. Just head over to Wilmott.com and ask the question, but make sure you have an asprin handy.

I like to keep it simple.

*Flat time * 1.5
WHY: If the flattime is too long then you are spending too much time in drawdown. It would be like owning a business that doesn't make money, you would be better off working at McDonalds i.e. (putting money into interest bearing account)

If the longest real time flat time is longer than 1.5 x longest backtested flattime then I stop trading the system and start investigating. Sometimes it is just that the volatility of the market has changed, or a there has been a fundamental change such as when the SPI went electronic; you need to find out what changed and fix it, or start developing a new system.

(You can also monitor annual profit factor to the same end. If adjusted PF < 1 then there may have a problem).

NOTE: You would have to investigate trade frequency to determine if the system is failing, or just trading less.

http://img69.imageshack.us/img69/3757/flattimexa8.jpg

*Consecutive loosers
A bit of probability comes in handy for this measure. Lets say that I backtest the system over the past 5 years and the maximum consecutive losers is 10. We then do the following calculation in Excel for a sytem with a backtested 50% probability of a win.

=ROUNDUP(LOG(0.0001,(1-0.5)),0)

the answer is 14 which means that we have a 1 in 10,000 chance of 14 consecutive losers. Since 14 (calculated) is higher than 10 (backtested) then I use that number to measure the performance of the system. If it goes one trade over, then I stop trading the system and give it an overhaul.

This has happened to me in markets that have quickly increased.

For example, if I traded the GBPUSD from 1993 till 2001 with a stop of 50pips, that would be .32% of the price range (14000 - 17000).

If I kept the same 50 point stop from 2005 - 2006, that would be only .26% of the price range (17000 - 20000), hence I would get stopped out more.

There are other reasons for consecutive losers. This is also a real life example. If you trade large volume on non-fx markets, and someone finds out about your system, they can step in front of your orders. Over time, their accounts get bigger, and your account starts to get smaller.

*Average Win:Loss
Most systems will revert to a 50% win 50% loss period at sometime over their lifecycle. For this reason, a win:loss of >1 is preferred. If the system drops below 1, then review it.

*Win%
If the win% is more than 2 standard deviations from the mean, then I stop trading the system and investigate changing parameters etc.

http://img120.imageshack.us/img120/9...iationshc1.jpg


MDD
Take your average backtested annual gain and divide it by 2. If my maximum drawdown ever exceeds that number, then I stop trading and review the system.

When designing a system I make sure that the maximum drawdown is one third of the annual gain. For example, in the chart below you can see that the maximum drawdown is 5.5%. 5.5 x 3 is 16.5% which is my minimum annual return to make the system tradable.

http://img129.imageshack.us/img129/2...wichmddvq3.jpg


In general, if the Big Sandwich system makes 60% p.a. then

* my maximum CALCULATED drawdown is 1/3 or 20%
* my trigger drawdown which is use to measure the system is 1/2 or 30% upon which I stop trading it



http://img113.imageshack.us/img113/8...dsheet6lp2.jpg

Out for a $1180USD profit. Reversed positions now and buying GBPUSD.

http://img300.imageshack.us/img300/5...dtrade7px2.jpg

Mongoose,
thanks a lot for this. I really appreciate you giving specifics. What you wrote triggered some of my own ideas which I need to think through.

This is really valuable.

I found this today in the Trading Game by Ryan Jones
"I do not like to use statistics to gauge how much money I am going to make by trading any particular system. I use statistics to gauge when I should say “uncle” and when I should stick things out."


Thanks.

I have discontinued this journal as I have other more pressing time commitments. Please feel free to keep papertesting the system.

In a year from now, I will backtest the system and compare it to my real results, but so far it is on track.

Where's Mongoose?

Hey Mongoose,
would like to see results if available?

regards
Kiran

He heasn't been on the forum for 8 months don't hold your breath for a reply!

Andy

Just an update on the Big Sandwich, it has currently averge me 15-30% per month on average since inception.

There have been a number of people modifying the system, and some even added to the rules to create their own system. One notable person has entered their modified system into a trading championship, and currently ranks #36 with a gain of 140% in the last 2 months alone.

See "Dazler74".

http://championship.mql4.com/2008/users/index/3

He is obviously a good trader in his own right and I wish him all the best with his modified Big Sandwich.

Of course, I do reccomend that when you PAPER TRADE, you do so conservativley as if it were real money.

This is not a reccomendation to trade any instrucment at all. There is substantial risk involved in trading.

What a pity, this thread has been discontinued. And even more that the posted charts are not avaiable anymore.

Anyway - thanks, Mongoose.

Regards