Xmas Project

[FetteredChinos]

Ok chaps, how about a wee project for xmas?

and no, im not taking the pi$$.

boom, and errr, boom.


In this supposed time of giving and sharing and wearing silly jumpers, why dont we attempt to build a basic system between us. It might get the old brain juices flowing, and help the newer chaps out...

in the interests of getting the ball rolling, may i present a raw method as perhaps a starting point.

if a market is up over the past X days, the go short near the close, and close the trade Y days later. Vice versa for longs..

it seems to work with the same criteria for all of the indices with similar parameters(which is a good indication) and having coded in basics stops (not included on the attached sheets) the equity curves almost look tradeable lol.. (Far-east markets and the dow look a bit hairy, IMHO)


so, who has any improvements then? trailing stops? extra filters etc? using in conjunction with breakouts etc?


usual caveats apply:- this is based on cash data, not futures (shouldnt matter when trading close-close) , data could be wonky (always likely), Chinos is an idiot (this is for certain) etc...


FC

 

[FetteredChinos]

bah humbug the lot of you!!

not even a sniff of a response....

 

[trendie]

It is a good idea, but will get bogged down in pedantics !!

I tried something similar on the Fib thread. I wanted traders to apply their advice, and different takes on a single style, Fibonacci, and to arrive at a consensus trade.
( My idea was that some traders may be very good at entry, but have loose exit strategy, and make mediocre money. Then some are not good at entry, but seem to have a good exit strat. By combining the consensus good entries with the consensus exits, we could arrive at a system that is the amalgam of everyones good ideas. It trailed off at some point, never to be heard of again ).

If someone starts the ball rolling, I hope I can add valuable views.
But, you need to identify the key parameters, such as whether it is intra-day, swing, etc.
Also, pre-define risk-reward profile, otherwise you will be inundated with advice which contradict each other becuase they have different perspectives.

So, FC, what are the ground rules ?

 

[FetteredChinos]

ok, the ground rules are simple...

ideally, a counter-trend system is the plan. these have a higher win-rate and are therefore easier to trade, as the time between equity highs is less than those of trend-following methods.

secondly, the method can be intra-day, or EOD , or even on weekly charts. Doesnt matter which really.

thirdly, the method should probably be indicator-free, or at least simple enough to deduce from a quick glance of the charts. keeping it to price only would help.

fourthly, it shouldnt be instrument specific. ie the basic method can be translated accross several instruments, preferably uncorrelated. (incidentally the method i posted above works on FX to a similar degree)

those should be loose enough constraints to begin with?



hence my simple buy low, sell high method. piece of cake to operate.

if you can tolerate the wonky equity curve, and the high transaction costs relative to profit, then it works. as is, but surely can be improved upon..

 

[Roberto]

Far-east markets and the dow look a bit hairy, IMHO

They always were, old man: don't let it inhibit you.

 

[pkfryer]

Alright Fettered my old mucker...

How about bunging: only take orders in the direction of the next time frame up. i.e. if the weekly's are trending down, after the daily chart has been trending up sell. and opporsite for buys.
Using orders at the low of the current day for sells and high of the current day for buys instead of jumping straight in at the close just in case there is still momentum left in the up trend or you catch a buy frenzie or a selling climax and get scuppered.

 

[FetteredChinos]

interesting thoughts PK. i agree with the fractal aspect of the system. ie using alternate timeframes..

regarding your use of selling lows after an up day for example, that is a standard candlestick reversal pattern (cant remember the name, my japanese extends as far as Wasabi). might be called dark cloud cover. might not be.


will look into it and report back my findings..

 

[pkfryer]

What are you using for your testing?

 

[FetteredChinos]

finger in the wind.


nah, excel all the way. im most comfortable with it, and i like the flexibility it offers. you can change parameters in a flashand see the effect.

admittedly it can cause some comedy errors by accidentally using future data (85% systems anyone?) , but if you triple check results that look startling, then you should be ok.


only thing i havent figured out is how to trail a stop. am working on that though...

FC

 

[TheBramble]

finger in the wind.

Elton John wasn't it?

 

[FetteredChinos]

Elton John wasn't it?

Sorry seems to be the hardest word, I wont let you make such poor gags again..

 

[ottos]

ok, the ground rules are simple...


thirdly, the method should probably be indicator-free, or at least simple enough to deduce from a quick glance of the charts. keeping it to price only would help.

fourthly, it shouldnt be instrument specific. ie the basic method can be translated accross several instruments, preferably uncorrelated. (incidentally the method i posted above works on FX to a similar degree)

.

IMO there has to be an indicator which will prevent you from buying ( selling ) too early .

I think that system should be modeled for specific market first like QQQ , made it work there and then applied to other markets (instruments). Universal tools usually suck .

 

[FetteredChinos]

indicator? whats one of those?

what do you suggest then to filter the trades out then? surely if we get in late on a counter-trend system then most of the move will have gone?


incidentally, i have modified the existing method, employing the same method accross the indices.

looks to have boosted the results by about 25%.

will post sheet in a bit.

FC

 

[FetteredChinos]

ok, who said pivot points were crap?

ok hands up, twas I, amongst others.


anyway, been doing some tinkering...


again this method is counter trend, selling into strength and buying into weakness.

so when do we go short? when the close is greater than the average pivot for the last X number of days (usually 5 or 6).

when do we go long? when the close is less than the average pivot for the last X number of days.


when do we exit? after Y number of days (usually 6).

does it work? possibly. can it be improved? of course.

is 15pts a day on the FTSE acceptable? i would say so!

 

[jmreeve]

Biggest problem with all these counter trend systems is knowing when to get out if a big trend is developing.

Otherwise some very nast drawdowns result.
Perhaps adding a trend detection exit would help?

 

[FetteredChinos]

hmm could do.

i have coded some stops into these fellas. and they seem to improve profit/drawdown considerably.

15-20 pts works well on the FTSE for example.


what triggered this whole idea was that i read somewhere about a guy who's sole method is attempting to pick tops and bottom, so he will sell repeatedly with tight stops until he eventually does pick the top for example.

kinda makes sense, IMHO



but any ideas on a trend detection exit? falling momentum perhaps? hmm, i'll have a further play. but may have to wait til later in the week as im on the lash tonight, then playing footie.

FC

 

[Tuffty]

Oh no. You've blown the system for me already.

I have a rule that I can't trade any system that's mentioned on a public board. Why? The thing is how do I know its not over fitted. I noticed the data you're using is up to date so there is no possiblility of doing a walk forward. I'll have to wait till Xmas 05 now to test it without 'hind sight'. What can I have for Xmas this year Santa!!

 

[FetteredChinos]

ok, if you want, i will do an out of sample test from say 1990-1997 on the FTSE and Dax.

that should be enough data??

methods like these shouldnt fall over too much.

in theory.

famous

last

words...

 

[FetteredChinos]

ah knackers.. mixed bag. with some interesting results.

quite puzzling in fact..

rerun a test with the same parameters on 1998-2001 data.


FTSE performed even better than the 2002-2004 period., the CAC quite a lot worse, but still positive.

the DAX on the other hand has done quite the reverse. Reversing the rules originally postulated looks as though the ideal strategy.

so what changed between 2001 and 2002? apart from the bear market? and why does the FTSE and CAC still work.

odd. very odd.

might offer the option of hedging the FTSE trade with a Dax trade. another one to investigate..

harrumph

FC

 

[pkfryer]

Fettered old bean, how do you define a pivot point?