Page 1 of 502 1 2 LastLast
Results 1 to 10 of 12

Thread: A challenger appears

  1. #1

    A challenger appears


    I hadn't ever heard of ForEx until two weeks ago. I've now set up a couple of trading stations (including apparently 3 copies of MT4. Can't have too much of a good thing, I guess ) and I have some insane demo account money to burn.
    I've also bought currency trading for dummies along with technical analysis to dummies. Those names are a tight match for me personally.

    So I've been spending a few evenings so far attempting to compose an EA that may always survive a egy tester encounter across different timelines and pairs, and even failing, unsurprisingly.

    Well, no hurry. I will not play with real money until I have generated enough reproducible information to convince myself that I might not be wasting my time here.

    Basically, I refuse to think in voodoo, therefore any indior I don't know, I will not use. My current EA functions purely with averages, and tries to discover localized trends as they begin and piggyback on the ride. I know why, although it really does a job so far.

    Reading currency trading for dummies so far made it clear the market could possibly be 24/5, but not long are equivalent, therefore an EA worth a darn will most likely have to watch the clock and understand the anticipated volatility effect for a given pair, at least. Too bad I have no idea what to do with volatility changes that are predictable .

    At this point, I am still not convinced this is not a giant sucker play, but the 2 books I've bought so far have cost me just as much as Civilization 4 so as a worst case, I can write it off as playing a fun make-believe game.

    If that is okay, I'll keep posting updates on my ever-so-slightly decreasing ignorance amounts within this thread. A chart might even throw in every now and then to keep folks interested

  2. #2
    Alright dear diary, it is time for another update.

    I haven't made as much progress as hoped, in part because I purchased a home (yes I know OMG my timing sucks), but this Foreign Exchange thing provided a welcome diversion from reading large quantities of legalese.

    My todo list was:Catch a complete bunch of tick-level data, using Find out a reasonably speedy format to store them in a DB Pick a programming environment to do this things in. It must run on Linux, and with a lot of math/financial libraries are a plus. Design an bridge, to expose the time tick data and the order putting stuff, at the least. I've got the whole tick-level DB of dukascopy today, updated. It is large, but the whole thing is automated, so it doesn't really hurt.
    Given the magnitude of this thing, I am not going to trouble stashing it in a SQL DB (or worse, convert it to a CSV format or something equally size-exploding.)
    Rather, I am keeping it in the current dukascopy format for today (that is dead easy and fairly streamlined: each tick takes 40 bytes to keep 5 64-bit fields: timestamp, ask, bid, askvolume, bidvolume. That's it.)
    I'm probably better off adding caching layers on top of raw file entry layers (memcache, most likely) than attempting to divine a more effective format to keep this.
    The programming environment issue isn't entirely clear-cut. The glue is going to be Python because I didn't know the language and I wanted an excuse to learn it. There is going to be other things orbiting about it though, including most likely R. I am also considering things such as quantlib, which is a C library, and Weka, that is written in Java. The list of items that orbit the core code is going to evolve as time goes by.
    I am diminishing the MT4 bridge thought, mainly because I don't need to bother. The historic data will probably be enough to keep me occupied for a while, and I'll only plug into a FIX-enabled demo account when I feel ready (likely using QuickFix to do this component )

    In retrospect, I guess that is all my todo list. So the Upcoming things to strike are:Finish writing a backtester that is reasonable. I execute some TA indiors so I will throw a few agents together by ripping some simple MT4 EAs and confirm that they act roughly similarly. Keep working through a neat financial math book ( Introduction to Quantitative Finance by Robert R. Reitano ) - Because I don't have any formal financial training but I was able to know some rudiments of mathematics, that will serve as a preliminary stop gap for now. Get to know R . In particular, I have no great charting solution so R might be convenient there. Item: Pick up some background information supporting Weka's filters and classifiers. This goes back to me refusing to use yet needing to use Weka. Also on the horizon, I am pretty sure I will buy some new hardware. My current linux box is. Its principal objective is to be quiet and to stream things to the screen, and it is most likely not going to be enough to store large quantities of things or crunch a lot of numbers.

  3. #3
    Quote Originally Posted by ;
    Anyhow, I'm not submitting to piss folks off
    Neither am I. Thank you for the answer...

  4. #4
    Thanks Brian for observations and your principles for me to bear in mind. And thanks to Krakoukas for pointing out the somewhat ugly potential for (unacceptable) losses which appear quite clearly on the transactions which stayed open (my prep review of the trades), albeit so far closure in profit or tiny losses. What I totally missed was that the losses weren't singled out to be shut by a algorithm - but rather because they belonged to a bunch of small-lot transactions - all being closed together when there was sufficient profit in the other transactions! It appears so obvious today...

  5. #5
    Ryan from FXCM called me. He asked to talk to Metal, that's the first name I put in my demo account information. So that was fun.

    He was fairly nice and offered to help, though he seemed bummed I did not seem anywhere near prepared to throw some real money at the system.

    I'm going through a bit of a trading funk I think, in part because my day job has been keeping me busy, and in part because I don't really know where to go next.

    It seems I have to either stick to TA approaches using a bit of fundamentals and stay at it until I begin to acquire some instinct for it, or beef up onto some significant mathematics and compose a lot of code using it, some of which can eventually give me a small advantage for a limited period of time.

    The later is a lot more likely to occur for someone similar to me, but it demands a lot of work.

    So I'm reading books. It kinda stinks that data were never my strong math area. Oh well, gotta ch up a while.

  6. #6
    Alright, after a few months of giving up with this mess, I'm stepping back into the arena.

    Within my off period, I have come to a few realizations:
    MT5 and mT4 suck. Using them is a little like shooting your own foot, except the hospital bill could be smaller. Crappy APIs, language that is crappy, there is no shortage of crap to move around. JForex is vastly stronger, and yet it has sucky facets, not the least of it's the system that is lock-in using a lone Swiss entity that merely has a 14-day trial to play their demo, along with the documentation. Any system that has a shot at not sucking is going to need to exist as far away from all of that crap. The information has to come from somewhere, and the orders need to be put so some kind of bridge is very likely to be needed, but if the code isn't just a toy, it needs to be platform- and - broker-independent.
    So that's the story. My next few steps:
    Grab a complete bunch of tick-level information, using Figure a reasonably speedy format to store them in a DB Pick a programming environment to do that stuff in. It has to run on Linux, and with a lot of libraries available would be a plus. Design an MT4 bridge, to expose the real time tick information and the order setting stuff.

  7. #7
    I'm surprised that you find MT4 useless. Lots of us use it to trade profitably, even though some have a preference for other platforms.

    Upsetting one of our brightest and most generous abilities - by the name of PeterFM - doesn't cut it with me, either. Maybe you discover the grind of learning gloomy. Some people find it thrilling.

    You don't need tick level information. Any EA work can be made by me. Tell me I will tell you when to change on it and what it's searching for. I say once - I don't mean every Monday at 8am. I mean once I visit price action at support and resistance levels. But I might want to switch it off until it f**** up the trade greater than I could. Quite simply - learn to trade. Can you smoke that?

  8. #8
    Quote Originally Posted by ;
    I'm surprised that you find MT4 so unworthy. Lots of us use it to trade profitably, though some have a preference for other platforms.
    I was specifically speaking about the programming aspect of MT4, which limits you into a intentionally sub-par and under-featured speech for no fantastic reason.

    I am not arguing that plenty of folks may well be getting loaded with it, only that I don't mean to cover using that tool.

    Quote Originally Posted by ;
    Upsetting among the brightest and most generous talents - by the title of PeterFM - doesn't cut it with me, either. Maybe you find the grind of learning depressing. A number people find it thrilling.
    Well, hello abandoned area! PeterFM (and now you) misread one of my entrances. I answered to him right away to explain the misunderstanding, but it was too late. The master had abandoned my thread thread.
    Thank you for bringing this up though, I get to feel bad all over again about it.

    Quote Originally Posted by ;
    You do not need tick level data. I can create any EA work. Tell me I will tell you when to change it on and what it's searching for. I say once - I do not mean at 8am. I mean once I see price action at support and resistance levels. But I may want to switch it off before it f**** up the transaction greater than I could. In other words - learn to exchange. Would you smoke that?
    No. Because you are awesome, * You do not need tick level data, but *I* do.
    To pick one example, back-testing a scalping EA without historical tick data is a big exercise in wishful thinking. Further down the line, historical data is helpful for a lot more than back-testing, and finer resolution is better .

    Besides that, my final goal is to get a mostly fully automated trading system. An automated trading egy that works well when I maintain its hands isn't really much of an automated trading egy in any respect.

    Anyway, I'm not posting to knock off people (though that does seem to be an occasional side effect), but to keep track of where I am through the procedure, and invite the occasional constructive opinions.

  9. #9
    Dear journal,

    I hope this letter finds you in great spirits.
    Has it been three years now since I last wrote ? Time flies in which you're having fun and are not doing much of anything forex-related.
    Things are going great on my end. I moved into new lodgings and acquired $300K of debt in the procedure. I wasted two years of my life. As such, I have generally been having a fantastic time.
    Unfortunately I have fallen prey to the Foreign Exchange insect. This has to be a strain, and also a virulent one at that, although I believed my earlier episode could have inoculated me out of further relapse.

    And so here I am again, writing to you. . The feelings of despair that once seized me have long dulled, and I've come to believe once again than the golden ticket is just out of reach. All I want is just one ridiculously elaborate self-adapting EA that could bring MT4 to the knees were it to be written in mql4 (Would you believe that language/platform/abomination is still about? Whatever happened to MT5 and its beautiful promises to molest C syntaxes the way MT4 abused C syntax?)

    I understand, dear journal, as I write it down sounds absurd, and perhaps it's. But that truly is the egy, and I shall pursue it at least until some other glistening and object of desire distracted me.

    The egy, in real terms:
    - mt4 zmq-based bridging to node.js ( odd choice maybe, but I console myself with the possibility of using rio to perform R things out of it)
    - a few DB backend appropriate for this. ( could plausibly be SQL-based. . I've beefed up my hardware sube. Yay for startup leftovers)
    - compose some controller logic which manages lots of largely idiotic EAs, assesses their performance and chooses you to control another trade's entry and exit (not actually my idea. Bonus points if you realize the newspaper this is pulled out of.)
    - compose a number of largely idiotic EAs.

    Second week is shot, but then I expect to be able to put enough chunks with each other to begin getting some terrible backtests from that architecture in a couple of nights.

  10. #10
    A progress update.

    Node.js is not a fantastic match, as it happens. Too many things.
    So C is your clear option. Simple tasks like reading a CSV file are still CPU bound, but it is nonetheless a wonderful addition to have the ability to parse a fluffy CSV file in 4 minutes as it had been taking 60 hours in node.js.
    The mt4/zeroMQ bridge works as expected, so no big surprises there. It will need to get fleshed out, but that busy work. I have not tried stressing zmq and it is unclear that I'll ever want to.

    No final decision made on a storage DB, as I want to understand my usage patterns better.
    For the time being, I'm converting tick files into a very easy direct access binary format, which should buy me some time.

    It's been about a decade since I did any kind of acute C , so I'm becoming re-acclimated slowly.
    As a side effect, that means whatever code that I write now and for the next few weeks is likely to be thrown away and rewritten soon after.

    Alright that's it. I want to remember to be funny in my next upgrade. This felt entirely dry.

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
This website uses cookies
We use cookies to store session information to facilitate remembering your login information, to allow you to save website preferences, to personalise content and ads, to provide social media features and to analyse our traffic. We also share information about your use of our site with our social media, advertising and analytics partners. Read more.