Half-Baked Potato

[FetteredChinos]

Right guys, i have been tinkering in excel for the past few days, hence my absence from the boards for a while, and have been testing the system put forth over on MoneyTec for trading GBP/USD


http://www.moneytec.com/forums/_showthread/_threadid-9388/_s-588b9cbe4d50548f7d493bedc81d90a4/_s-


bascially the rules are:-

take the price @ midnight GMT on monday

then put limit orders in to buy at this price + 50 , and to sell at -50

these are stop and reverse orders. so if the buy level is hit first, then the price reverses to hit the -50 level, then you stop out the BUY for -100, and activate a sell.

the stop for the reversed trade is the original BUY level (ie -100)

there are only 2 trades per week, one in either direction. therefore maximum weekly loss is -200.

all trades close at 1800 GMT on friday.


i have tried my best to backtest this, and i attach the results. for 2001- early 2003, it was just about profitable, but then as we move into late 2003/2004 it seems to have gone ballistic, as you can see from the equity curve.

i have taken my data from http://www.fin-rus.com/analysis/export/_eng_/default.asp

and it (the data) looks a bit ropey (and im sure ive made some errors in my coding), but 140-odd trades should give a fair indication of the merits of the system.

drawdown is manageable, if not too decent, and the win rate is definitely useful (if a little distorted by the 2004 performance)


Total Points 6041
Total Trades 142
Total Wins 83
Win % 58.45%
Max Win 426
Max Loss -200
Pts per Trade 42.54225352
Average Win 153.3
Average Loss -113.2
Pts per Month 137.2
Max Drawdown 660
Profit Factor 1.90



i will attach the sheet in due course, to enable you to tinker with the entry and exit variables, and of course , i would appreciate it if someone could check the code (if they have time)

possibly the reason for the underperformance is that volatility in 01-03 wasnt as high, and using the same entry triggers as for 2004 isnt such a good idea. i will look into coding a more dynamic entry method in due course.

additionally, im also fairly sure that returns could be improved by implementing a trailing stop, rather than the fixed -100, but sadly i dont think my excel skills are quite up to that yet.

usual caveats apply (dodgy data, dodgy coding, do your own research, obey the green cross code etc) , but im throwing this one to the hoards to see if it can be made more robust..


answers to the usual address,

Chinos

 

[FetteredChinos]

ok, and here is the spreadsheet i have been working off.

its a bit of a whopper, so may take a while to download.

you will need to hit F9 to recalculate after making adjustments.


the yellow cells are ones you can alter, and SHOULD alter the results.

please enter them in the formats that they are already in. eg type the day in full, and the times are 500, 1800, 2300 etc.

please also note, that i think the times on my data are GMT + 5 for some reason ( could someone verify this please??) so 500 is actually 0000 GMT , 2300 is actually 1800 GMT etc.


hope this helps,

FC

 

[TWI]

I had a look at this too and it seemed to work OK although recently there has been a historically bad drawdown.
When I get chance I will post equity curve.

 

[FetteredChinos]

ok, here is my version of the equity curve. pretty rough and ready though.. but it gives you the idea.

FC

 

[Goober]

Step back SIBKIS system

FC

What a coincidence! I've just been reading the very same thread. Like all B/O systems, the SIBKIS system works well in trending markets but is cr*p in ranging markets.

My thinking is to eyeball Bollinger Bands on a 60 minute chart to identify trending or ranging markets, then trade the SIBKIS system in trending markets and a range-trading system in ranging markets. Obviously, this will require some screen watching, but we all have to work for our money.

As far as the SIBKIS system is concerned, you may also want to consider a strategy to minimise false B/Os.

Goober

Right guys, i have been tinkering in excel for the past few days, hence my absence from the boards for a while, and have been testing the system put forth over on MoneyTec for trading GBP/USD


http://www.moneytec.com/forums/_showthread/_threadid-9388/_s-588b9cbe4d50548f7d493bedc81d90a4/_s-

 

[FetteredChinos]

agreed Goobs . if been thinking of introducing a daily filter into things to try to ensure a correct trigger.

at the moment, im thinking of (but havent coded yet) this kinda set up for long entries

main entry is a breakout of mondays trading range, providing that mondays low > friday's low. ie taking the breakout as a continuation of the move.

this is gonna require some serious tinkering, but it might be worth it.

what im puzzled by is the crap (well relatively) performance in 2003. the market trended pretty strongly (on the daily chart) , and yet this system seems to have fallen over..

odd..


very odd.

still, onwards and upwards..

 

[Goober]

FC

I haven't looked at 2003 charts. However, I wonder what the Friday afternoon retracements were like. Normally, in a strongly trending market, punters will take profits on Friday afternoon. Be interesting to see ...

agreed Goobs . if been thinking of introducing a daily filter into things to try to ensure a correct trigger.

at the moment, im thinking of (but havent coded yet) this kinda set up for long entries

main entry is a breakout of mondays trading range, providing that mondays low > friday's low. ie taking the breakout as a continuation of the move.

this is gonna require some serious tinkering, but it might be worth it.

what im puzzled by is the crap (well relatively) performance in 2003. the market trended pretty strongly (on the daily chart) , and yet this system seems to have fallen over..

odd..


very odd.

still, onwards and upwards..

 

[FetteredChinos]

right here goes, another modification, with a slightly more dynamic entry....


this is tested using daily data, so results arent going to be as accurate as testing on intraday data. again, perhaps someone with Tradestation could do the honours......


new entry criteria - trade the breakout of the monday high/low and hold until friday.

eg, note mondays high and mondays low.

on tuesday/wednesday, set up orders to buy at mondays high and sell at mondays low.

each order acts as a stop for the other.

close open trades on friday each week.


these "appear" to be the results since Feb 01


Total Points 8671
Total Trades 176
Total Wins 106
Win % 60.23%
Max Win 457
Max Loss -315
Pts per Trade 49.26704545
Average Win 139
Average Loss -87
Pts per Month 193
Max Drawdown 386
Profit Factor 2.43



im attaching the new sheet for you lot to tinker with. its a lot smaller than the original sheets..on this one, you can only change entry and exit days...(again could someone check the code? i think there may be something wrong with friday's when there is a big reversal, causing the losses to be greater than they should be)

i need to double check this new method, as i only just knocked it up at lunchtime, but the equity curve seems a lot smoother, and therefore more robust.

FC

 

[FetteredChinos]

just noticed..

changing the exit to thursday boosts the profit factor to 3, reduces drawdown by 10% and increases the win rate to 64%.


either there is something wrong with my friday coding (as suspected) or exiting on a thursday before the major economic data is released on a friday is more prudent..


right, back to me sarnies, as im suffering from having a flu jab this morning, and need all the strength i can get..

FC

 

[FetteredChinos]

ok, had a check of the data, and it seems as though the feed is 9pm-9pm GMT.

this means that for the daily breakout method, we have to take the high/low from sunday 9pm (ie weeks open) to monday 9pm, and proceed from there, and the trade can be triggered anytime from monday 9pm to thursday.

trades close at 9pm thursday.

looks pretty useful at 40-50pts per week over nearly 4 years. i've had a quick eyeball of a few months worth of daily charts, and it seems to verify my coding, so im reasonably confident.


havent looked into stop and reverse potential just yet, but it might be worth it, based on previous findings, as may trailing a stop...

toodle-oo

FC

 

[ChowClown]

Well done FC. Thanks for sharing your hard work Looks like another low maintenance retirement fund builder. Nice one.

 

[Roberto]

Very interesting work, and observations!

I've also been playing around with this (from the same thread) for some time, and in my own notes have called it a "midweek system" as I was already regarding it as an open-on-Tuesday, close-on-Thursday system.

This year's results certainly seem to be better than last year's.

My reservation about it is that (like so many systems) it makes money in markets with a tendency to be trending, which isn't really when we need it!

 

[FetteredChinos]

no worries chaps...

that thread got me thinking about things, and it kinda followed from there.

basically it acts like an Opening Range Breakout strategy (which we know are profitable) , but instead of an intraday basis, it is on a daily basis.

looks good, and im pleased that after spending hours coding up this fella that it does look ok.

tonights task, apart from cooking a slap up stir-fry, is seeing if i can code another idea or two.

ah the delights of abusing a company laptop lol.


toodle-oo chaps

FC

 

[Roberto]

basically it acts like an Opening Range Breakout strategy (which we know are profitable) ...

Yes indeed. Always nice to see someone saying that openly and casually. I agree with you. I've just ordered Mark B. Fisher's "Logical Trading", following a couple of recommendations, which I understand revolves around various methods of "ORB" strategies on various time-scales. Anyone read it?

 

[alliance]

The original strat would have turned in a nice lil 300+ this week...Tues - Thurs

 

[tradermaji]

Chinos,

Thank you for your hard work. I have a few comments, maybe you can incorporate it in your spreadsheet. However, it will take good amount of effort, but maybe worth it.

1. Make the breakout levels dependent on volatility. Calculate the Average Daily Range for a period of say 10 days, and then use a multiple of it, instead of using an absolute 50 pips. This way, the system will become more dynamic and perhaps will avoid some whipsaws. Say, place the breakout level at 0.5 times the range.

2. Close out the trade at the end of every day, and recalculate the entry points and re-enter the trading points. That way, the intraweek volatility can be better captured. In this case, a maximum of two trades per day can be allowed to prevent whipsawing to death on days with too much intraday volatility.

3. To complicate matters, intraday data should be used for testing. Perhaps 60 min data should be the data of choice. It will avoid overcrowding the spreadsheet, but will give a fairly decent idea of intraday swings. If you need 60 min data, email me at tradermaji at cox dot net and I will send it to you. I think I have data till last year, and it is US New York Time data. You may have to massage it to suit your time zone.

Thanks again for your work.

 

[TWI]

tradermaji,
I implemented your points 2 and 3 when I originally started playing around with this system and found that they did improve the curve smoothing but did not improve net Profit.
As for 1 I did not use a volatility dependent breakout. I have not been able to find an effective method which significantly improves returns of this model. Still fiddling around with it however so will let you know if I do.

This week has shown a pretty good return currently 268 points/contract on the original but it needed it as recent performance was not looking great. September was a particularly bad month losing close on 500pts net.

 

[TWI]

Amended system incidentally showed 247 tick profit in Sep but only has 180pts in bag this week.

 

[FetteredChinos]

thanks tw for testing the ideas..

tradermaji, isnt your point 2, just trading the daily breakouts and closing each day? part of the reason why i didnt want to do that is transaction costs mount up, especially with a spreadbetting account..


also, i had a play last night with more than one entry per week.


entering on friday on a break out thursday's high/low and closing on the monday the week after has the following results..


Total Points 3894
Total Trades 80
Total Wins 59
Win % 73.75%
Max Win 281
Max Loss -181
Pts per Trade 48.675
Average Win 84
Average Loss -50
Pts per Month 87
Max Drawdown 181
Profit Factor 4.68

if you run this parallel to the tuesday-thursday method, you are looking at 12,500 points over 45 months, less transaction costs, (probably 1500 points including spreads and slippage)

which is 250 points per month over 4 years, which is a pretty sizeable backtest, and pretty useful , IMHO.

only concern is that trading the breakout on friday could be prone to some serious slippage upon economic data release.



btw, in case any of you are trading direct access with minimal spreads, pretty much any combination of trading a daily breakout and closing 2 days later works.



anyone got any other ideas?

FC

 

[TWI]

As well as trading a breakout, it can be interesting to trade a range. I remember seeing something about a daily FX range some years ago. I played about with this and ended up with something that sold within a number of ticks of multi hour high, bought within number ticks of multi hour low starting at 01:00GMT. Combined this with a breakout system and ended up with something that functions consistently profitably across GBP/EUR/CHF.
It i not often you just change the underlying instrument and the system continues to produce a nice curve. CHF leaves something to be desired as a little volatile but the 3 together create a smooth profile.
I will not share exact details of this system as I actually trade it and it took a lot of work and maybe I am just being a little selfish, sorry.
I trade it on 60min charts. Main problem is lack of automation for execution as most of the day I concentrate on interest rate spreads so miss some orders and so suffer excessive slippage or just do not take some signals. This is very bad so I have been looking at lots of auto execution possibilities. TS seems to be best bet at moment. Just do not much like their rates.