dont know where this thread will go but here's a start

[Adamus]

OK, Mr Acronym,

here's a thread that's not called "Traders from Australia?" for chat about whatever.

I put it in this section because I'm a system trader, my own systems developed on TradeStation - although I could easily have put started it in Futures, or Indices or Commodities or Technical Analysis.

You know what though? After all, I don't actually have any specific questions for you, although you said you traded the kinds of markets in the same way I intend to (end-of-day), so I figured you're bound to say something that's useful to me.

The Holy Grail that I'm looking for, which I won't find and which will mean I'll probably end up trading my system without being 100% happy with the risk, is something that I can trade in all markets - and I look at 24 markets with the stuff I'm working on now, and I've only ever once got it to the point where a system is profitable in all markets - and by profitable, I don't mean I've factored in slippage and commission

So what sort of trading did you do?

 

[acronym]

Hello, just got here, doesn't sound like your that new if you can code As stated elsewhere, the stress of trading (ordering, &risk to be precise) gets to me, fatal personality flaw i guess, trading wise.
Ive followed the markets for 15 years or so though, and use what looks to most people to be a hodge podge of indecipherable technicals-a 3xmacd method with bollinger bands/ma's, to be precise.

Entirely discretionary, unfortunately i cant code. And if i could, i wouldnt rely on it 100%, I don't think.
That's not to say there has never been money on the table, a few years ago now, there was an opportunity to put my skills to work, but the account holder, after a family loss/breakdown choked, and coudn't place trades either.
So, I ran my method from the start, came up with potential trades, hoping something would happen-It didn't-within 6 months starting with a theoretical $15.000, (there was actually $15k, but never even touched the market) my "paper" was up to $150k, a year out, quarter mill, year and a half pushing half a mill, then i lost count and gave up, basically.

You might think $15k isnt much for commodities/fut's, it isn't, but i headhunted /cherry picked comparitively "cheap" contracts like corn, euribor, london coffee etc. It seemed to work-in one three month stretch, there was a winning run of 17 or so winners from twenty, helped get off the ground-the virtual ground that is.

So, first lesson for a newb? Never rely on anyone else, It has occurred to me that might have been my one shot, and i had to rely on someone else, and it cost a LOT. Money not made, it seems to me, is the same as money lost, that's the potential of the market, something you seem to have figured out already by grail seeking perfection.

You may be surprised to know, i did this with only one contract in a given market, never more than about 5 open positions or orders, and critically, all new trades (except for one or two down months, which basically broke even) were from closed out profit.
I factored in slippage and commission, by taking the worst possible entry/exit for a given month, and just doubling losses every other month to see if it was still in the black.
Bizarre, im not sure how i did it either. Wait, yes i do-diversification, and hard work.

Risky? Hell yes. Not as risky, as NOT having the positions in though.

More comments to follow, i figure maybe you have thought of some questions at this point maybe.
That's why i say, Im a charting guru-you can guess how much time i wasted, and I don't like to see newbs wasting there's, if I can do anything about it.
As stated, plenty will disagree, question is if you get some ideas or not. good ones, that is.

 

[acronym]

Quick follow on, while my brain is still working correctly, you will never find a system that is profitable in all markets, and isn't worth the bother looking for it. Trust me, even if you did, it would make so little ROI , in the long run,it would be absurd.

You say you have a "universe" of 24 markets, that's not dissimilar to my discretionary approach-but only a madman would try and trade all simultaneously, even assuming you had the capital. Big ask.
Were I to try and code a system to work well "most" of the time, I would use a number of filters, and roughly three seperate systems, with a randomized factor which may constitute either the third system, or a purely discretionary aspect. One thing i would not do, is place any mechanical exit rules, or profit targets-IMO the money is made on exits, not entries.

 

[acronym]

Another idea, maybe looking at a fisher transform for coding purposes.
Beats the heck out of MA crossovers, thats for sure.
Normalised data, well, isn't that a coders dream?

 

[Adamus]

... use what looks to most people to be a hodge podge of indecipherable technicals-a 3xmacd method with bollinger bands/ma's, to be precise....

OK I get it. That would be easy to code at first glance. But maybe you do alot of discretionary stuff around the edges that couldn't be reflected by a mechanical system those rules.

I'm actually coding EL in TradeStation4, using end-of-day data, so I'm not much of a rocket scientist when it comes to trading - although I've worked with Java in IBs building trader portfolio management tools, but that's all about interfaces and where the trader wants to click, much more than to do with trading.

Actually I strongly believe I will find something that I can trade in all markets - not only that but also using end-of-day data, and without curve-fitting over-optimisation. But it takes such a long time. I've been working full time on this since Easter.

My knowledge of the markets is quite in-depth now in terms of the mental picture I have build up about what futures time series look like. The average true range and how everything else relates to it.

I don't really like BB's, I think they just lag and aren't so useful. In fact I don't like moving averages on the whole at all.

I don't really know what MACD is (having said it would be easy to code!) - I'll have to check it out. I'll get back to this later, have to dash now.

 

[Adamus]

I looked up MACD and see it's heavily trending-market-oriented, so that's not something that I need at this point with my system.

Sounds like you found something you were happy with though - happy at least with in theory, if not in real trading.

In your situation, using discretionary processes to generate and end trades, I would probably be too nervous to trade as well. At least with an all-mechanical system and lots of historical data, I can see exactly what would have happened in the past and I can build up stats showing the probabilities of various results and the chances of huge losses etc.

I'm working on a break-out system and I am using moving averages, but I'm smoothing them so that the latest price is always the most influential.

The thing about simple moving averages that stops me using them is the effect of big values dropping off the end, say a 21 day MA moves forward one bar, and the 21st value happened to be huge - it will cause the MA to rocket or dive, even though it was actually 21 days ago. See what I mean?

 

[Adamus]

Another idea, maybe looking at a fisher transform for coding purposes.
Beats the heck out of MA crossovers, thats for sure.
Normalised data, well, isn't that a coders dream?

Oops, missed that post. What's a fisher transform then?

And normalised data?

There's no smilie for 'staring blankly'.

 

[acronym]

Oops, missed that post. What's a fisher transform then?

And normalised data?

There's no smilie for 'staring blankly'.

Fisher transform-
QuoteTracker Technical Indicators

Might clean up the code a bit, if nothing else. Might be worth playing around with?I dont know why on earth they invert it, but as far as indicators go its very much price based the way i figure, sort of a smoothed function, with reasonably sharp turns-wont lag as badly as an ma possibly.

Macd's are good for momentum, though I mainly use them for diveregence-using three time frames, for example, i need divergence on two time frames , plus a pattern to be to excited about a reversal, or depending on how it looks, switch to a stop and reverse/band method instead of treating it as a trend zone.
Complicated, yep, but theres a lot to be said for picking (guessing? Is it really random?) the right strategy for a given chart.

 

[acronym]

Here's a simple system to work on, the EasyLanguage code is around somewhere, cant find it straight away
Trader.Online.pl® - MetaStock™ Zone - Linear Regression Reversal System by Barbara Star

You mention your "looking" at 24 markets, but backtesting them, is that right? Is that over random data samples, testing profitability as a singular instrument, or as a whole?
Reason i ask, is i tested a lot (manually) on a lot of years data, just trying to get the hang of charting and price behaviour-what i found, was some instruments, no matter what i did, will be net losers for a given time frame; confidence crushing, for sure.

It wasn't until i started walk forward testing, and being able to look at recent correlations, recent news, (not so important, but nice to know after the fact) and most importantly, culling losing instruments (arbitrarily, 3 consecutive losses and i dumped it for a while) on an ongoing basis that everything clicked, legitimately treating it as a portfolio. The fact that there was/should have been orders going in probably tightened up the style to.
Uhg, one trade, i was dead set sure of it, so the guy put an order on-i didn't hear it-come tuesday (time zone thing, us market) i thought id nailed it, $4k in the bank, thanks very much-he'd forgotten to say "on stop". It had gapped, no position.
You've been going since easter, i wouldn't be to bothered by any results so far, i called the coffee bull in the mid 90's, tried to buy gold at $430 odd, (missed my limit-if you want a position, you GET a position, no matter what-cost me 5-10k easy money) recommended oil at $38 years ago, you name it, most stuff in between to, despte being a truly awful trend-um-chartist i guess.
Hope you don't mind the rambling, but there it is, other people's mistakes should be invaluable to somebody out there.

 

[Adamus]

Don't want to sound like a ponce, but most systems I see out there just don't stack up over multiple years and multiple markets.

I couldn't see the nitty gritty on the Barbara Star link so I can't say anything there.

I am backtesting 10 years in 24 markets. As a whole - what do you mean by 'random data samples'? Sounds a bit like Pardo's optimize/walk-forward windows. 18 months optimisation, 3 months walk-forward, shift the whole timeframe 3 months forward, repeat.

I tried that a while back but found it was too much work - produces huge amounts of stats and I didn't have the ability to analyze it effectively.

So why do you hang around here? Do you still have a sneaking suspicion you could do it right if you got back into it? Or do you just regret those missed opportunities? I have to say, I'm doing it now because I don't want to regret not doing it - I think that would hurt most. In a year's time, I'll be happy I've done it, whether I've been successful or whether the money's all gone. I can't not do it, if you see what I mean.

 

[acronym]

Don't want to sound like a ponce, but most systems I see out there just don't stack up over multiple years and multiple markets.

I couldn't see the nitty gritty on the Barbara Star link so I can't say anything there.

I am backtesting 10 years in 24 markets. As a whole - what do you mean by 'random data samples'? Sounds a bit like Pardo's optimize/walk-forward windows. 18 months optimisation, 3 months walk-forward, shift the whole timeframe 3 months forward, repeat.

I tried that a while back but found it was too much work - produces huge amounts of stats and I didn't have the ability to analyze it effectively.

So why do you hang around here? Do you still have a sneaking suspicion you could do it right if you got back into it? Or do you just regret those missed opportunities? I have to say, I'm doing it now because I don't want to regret not doing it - I think that would hurt most. In a year's time, I'll be happy I've done it, whether I've been successful or whether the money's all gone. I can't not do it, if you see what I mean.

Fantastic reply, and oddly you nailed it, which is what i was trying to get across.

Most systems wont work, year in ,year out, i mentioned a system, even a discretionary one, has to adapt to its market.
Be it time frames, be it higher/lower correlation, the need to change strategy is always there.
The barbara starr method, is so bog basic its ridiculous, BUT i mentioned it to try and illustrate the importance of exits, even in a basic approach, purely for testing purposes.Also, because my manual testing suggests even the most simple method, can work well in the right circumstance, regardless off any supposedly good discretionary ability.




As stated, im not a coder or backtester, and certainly was not reffering to any variety of optimisation, at all, what i mean by random data samples is exactly that, clearly you arent just picking and choosing a code that worked, at some point-you seem to have a "robust" system, so go with it! Just examine your markets, and step back a little.

As stated, you hit the nail on the head, to a degree-i chart, because i love it, i love being right; I love it, because i set a target, originally, -to prove it WASN'T possible to make money in the markets. I proved myself wrong, , due to factors explained here in this thread.

Reliving past glories? Sure, why not. If you DONT want to do that, via the potential results of your system, then you could well end up like me; that, WAS my point, overall.

i chart, because i love it, and like to know why things cost what they do. I chart, because it allows me to see the future., as an economist and a speculator.
That's all.