Here's another one

FetteredChinos

Veteren member
Messages
3,897
Likes
40
ok chaps, lets try a trend following strategy...


been doing some serious work evasion this morning guys, and found a slight improvement to the SIBKISS method over on moneytec.

sod putting order levels in, this literally should only take 5 minutes per week, and if you can then switch the pooter off and collect your profits on a friday, then we could be in business.

firstly here are the stats.. (no stops at the moment.. just wanted to share the idea first.)


since jan 03 trading the GBP/USD fx cross.

Points 6,064
Trades 89
Wins 52
Win% 58.4%
Ave Win 266.63
Ave Loss -210.84
Profit Factor 1.78
Drawdown 798
pts per trade 68.13



a bit better than the standard sibkiss method which i think had about 45-50pts per trade over a similar period, and the equity curve was a bit on the spiky side.

all in all, i think this a lot more robust than earlier methods. will attempt to fwd test it with fresh data.

could someone please verify my results?

the rules are.....

take the price of cable at Midday GMT on a tuesday (my spreadsheet is 3 hours ahead, so 1400 is actually 11am. using the close of this bar gives midday)

if the price at midday is greater than the price at 8am (1000 on my data) then go long, else go short.

close positions out on the 2nd friday. eg 8 trading days later.


looks to be a winner to me. and stops will doubtless improve the drawdown and profit stats, since for once we are trading with the trend.


comments, as usual, will be ignored lol :cheesy:

FC
 

Attachments

  • 2 week Cable.zip
    1.4 MB · Views: 417
incidentally, the EUR/USD with the same parameters returns -2200 points or so over a similar period..

suggests that it might be a useful hedge in order to remain relatively market neutral.

hmm.

interesting...
 
Very interesting as always, FC. You presumably have results (and presumably they're not so good?) for closing out on the same Friday? What made you look specifically at the second Friday?
 
well the results are still good closing out on the same friday, but in terms of profit per week, then holding a bit longer is preferable. adds 15-20% to the gross pips..

just tried it on EURGBP with the same parameters - 62% win rate, smooth curve,15 pips per week (not great, i know... but it shows that it has potential)

more currencies to come..

FC
 
FetteredChinos said:
not great, i know... but it shows that it has potentialFC
Thanks. It does indeed. Very interesting. Maybe we'll make a trend-follower out of you yet. :)
 
hmm, testing on USD/JPY yields approx 6,600 points since jan 02 with only marginally different entry criteria...(shifting the time periods back 3 hours)

might be onto something here..
 
balls, 2002 only yielded 500 points on GBP/USD before costs..

10 pips per week is crud...


but i have a cunning plan, sir....
 
FetteredChinos said:
10 pips per week is crud
It's not your best, certainly. But 40+ pips per week on the USD/JPY over a 3-year period is looking pretty impressive (unless there's some huge drawdown/hole in it, of course ...).

FetteredChinos said:
but i have a cunning plan, sir....
Rarely, if ever, has my breath been so bated ...
 
right here we go again... lets get some stability into this system...


attached is a test from jan 2002 to late 2004... averaging back to 40pts per week, but each year looks decently profitable, and of course 2004 is sh1t hot because of the bigger weekly moves.

good to see 2002 and 2003 helping out though..


slight adjustment to the rules...

take the price on tuesday at 2pm GMT (1600 on my sheets). if higher than the 8am price go long, else go short.

close out on the friday of that week..


will triple check the USD/JPY stats..

FC
 

Attachments

  • 3 years cable test.zip
    2 MB · Views: 289
and here are the results of the USD/JPY jury (jpy?) ....


6610 points at 45pts per trade...

rules are take price at 3pm, if higher than the price at 11am, then go long,else short

close on the FRIDAY AFTER.. (ie a 2 week hold)

a 1 week hold seems to return about half...

FC

curve-fitting anyone?
 

Attachments

  • 3 year USD JPY.zip
    1.3 MB · Views: 175
FetteredChinos said:
curve-fitting anyone?
I was too polite to comment. Maybe, yes; but still pretty impressive when tested over 3 years, n'est-ce pas?
 
just tested the cable version as per the revised rules for 2001:-


505 points over 41 trades, 23 wins

not great, but still profitable...plus volatility was pants back then..

i wont post the sheet as the files are getting a bit on the large size (20,000 rows of data now)

FC
 
thats what im thinking roberto...

i have tried most combinations of the first few hours movement on a tuesday from the UK open.. they all are profitable, just waiting til after any economic data on that day seems to be more prudent.. hence the 2pm crucial time..

FC
 
FetteredChinos said:
i have tried most combinations of the first few hours movement on a tuesday from the UK open.. they all are profitable, just waiting til after any economic data on that day seems to be more prudent.. hence the 2pm crucial time.
Hmmm ... thanks ... you're getting very good at anticipating my questions and pre-empting them!
 
The next question is whether improvements can be produced with a stop-loss (presumably? ... one for opening the trades and maybe a trailing one as well, but very difficult to verify the latter?) and/or a take-profit limit (presumably not?).
 
hmm, indeed, but i havent had time to code those in at the moment..

it is probably easier for someone with tradestation or the like to do that, as frankly with excel we could be here all night..

unless someone has already had a stab at these??


its on my to-do list, but then so are leaving drinks and footie tonight :)

fc
 
FC, interesting idea. If its any use as a pointer I've played around with different SIBKIS settings for an initial stop, trailing stop and a take profit level.

The optimum values I found to be;

Initial stop 100 pips (of course this value is effectively preset by the strategy so I havn't varied it)

A trailing stop of 175 pips to be applied once a profit level of 75 has been reached.

The best take profit level is anything from 250 to 1000. (makes very little difference)

Using pretty loose stop and TP levels gives the system the 'breathing space' it needs to work its way up the trend.

As the SIBKIS timescales are very similar to those of the strageies you have presented here I hope that you may find these values useful . If nothing else you could use them as starting levels for your tests.


BTW I,ve found that most of the SIBKIS variables can be amended and the system still makes a decent profit. Using the Cable value at 7 or 8 am tuesday to set the trigger values is about as good as using the 1am value as specified in the original SIBKIS

Also closing at 1900 on wednesday evening instead of friday gives virtually the same returns but with the obvious benefit of being exposed to the market for less time.

Im currently working a variation of the system where I close one position at 1900 on tuesday, a second at 1900 on wednesday and leave the last to run to friday 1900 (unless stopped out of course)



Neil
 
Top