ROFL ok but expect this thread to explode into 1000's of posts where I changed mind but instead of editing one post it will go on for 2 or 3 pages:
(Gnat if you edit real quick it won't show as edited)
Just had an idea.
Talking to myself here all alone tends to do that hehehehhe.
Now, what if, instead of backtesting the last 8 months, I backtested the last 8 years pouahahahahha. I've got data for most (not all tho grrrrrrr) pairs that go back to 1999 (=8 years, yeah right my nomination for math nobel is on the right path)
Guess what I'll be doing tomorrow morning. Also, I need to analyse why my forward tests for the past 2 weeks (real trading) outdid my last 2 week backtests. Results are very encouraging, but I need to understand why.
Ahhhhhh magnifique sunday morning ahead for me
And (nb) although the "economic calendar" here at FF does NOT mention it, monday IS a holiday in Canada (thanksgiving). So madam will be available heh on monday as well.
I will look a bit more into your Excel mamouth project work - fortify seom ideas.
Testing the really long timeframes may actually bind your computer for hours, but if we come up with good results through years... 4 may actually be enough..., then it all stands on solid ground.
It seems that many in the forum are still clinging to shorter timeframes and would like to have settings for that. Do you happen to have any good setting for 15M or 30M?
GMatt
I'm quoting DaveFloyd here. There's a discussion back at the Trading Room thread on automated trading:
"Regarding automated systems, I actually do not know anyone personally who does it but do know that the big Managed Futures firm Winton Capital in London is automated 100% and their track record is spectacular. I met with some of the staff a few years ago and their roster was mainly stat and heavy math folks - not a trader among any of them."
For me personally I would not run the supernova sysem on short term timeframes until you can code discretion
:I'm testing away right now
The results I got for shorter tfs were really bad. I'm reposting here so you can see what I mean. I even went so far as to think that if we could reverse something to get these results in the positive, it would be unbelievable (do you follow me?)
In a nutshell what's your approach to this
But we really do not want this. It is uncontrollable in risk.
Good point nevertheless! +9899% is a good figure!!
Gnattel
What a nutshell!!!!
Gnat, the nutcase
Thanks for that nutshell. I think I'm starting to get interested in them newevents/reports. Do you estimate their relative importance to be the same as the color codes in the calendar or not?
So basically, if I understand correctly, you go for about 15 pips for a given newtrade (?) times a month, and these trades last for a few minutes only?
Am I correcto maestro?
This time I’m citing Mr Trend (I put the stuff in red):
-----------------------------
Originally Posted by philmcgrew
I don't think automated trading itself is the culprit for failure. But when it gets in the wrong hands it is dangerous. If the user doesn't understand optimization, walk forward analysis, sample ratios, testing methodologies, and establish some kind of maintenance schedule then they are doomed.
Automation doesn't last forever and I think where most get into trouble is having an objective method for taking the system offline before their account is wiped out.
You really hit it on the head, Phil. All systems deterioate over time, and you need to have a good system building methodology as well...
----------------------------------
My paleontholithic tests (back from 1999) are over. And the results (specially the graphs) kind of reflect the above. Conditions change over time and the bots need to adapt, I suppose. Let me know if you're interested in the reports (graphs) and I'll email them to you.
http://i24.tinypic.com/5f2a20.jpg
Although you are right by definition... you are especially right on that.
I trade a few amber reports like UK GBP and I leave quite a lot of red ones by the roadside. Most of the reports are red. And the relative importance is given by the value of such a report at the given time for instance for the FED or the credit market or growth or whatever is on the agenda of the big players. Some reports traditionally move the markets, others like US ISM Manufacturing Index are losing a bit of their importance.
:
