Toutes les commandes
Règles
Spécifications
I'm getting nowhere programming it myself, I am more a mathematician than a programmer, it gives compile errors for all sorts of things that shouldn't cause errors, I can't even get off the ground and get past the most basic roadblocks, and the forum posters are a bunch of trolls who scoff at you for not knowing implicitly things that they're not going to even say and probably they themselves don't actually know. Basically "ha ha ha you don't know the answer but I'm not going to tell you, I'll just spend the time needed to tell you I think you're stupid without actually saying anything relevant and therefore imply it would be child's play for me to do even though I actually have no idea what the problem is". As you can see here https://www.mql5.com/en/forum/37207 I have a low threshold of tolerance for that.
But here's the deal. You get either 1000 dollars - possibly more - I want it done right, not necessarily by the lowest bidder - but you get this if you agree to an exclusiveness contract that you don't use the method yourself and certainly above all else don't sell/give it to anyone else. You can certainly use what you learned to some extent in your future projects. And you WILL learn some amazing things, count on it. Or you agree to get nothing but you have full rights to use the program yourself for your own trading but not to sell it to anyone else. Naturally there would be virtually no way I could enforce this, you could take the 1000 dollars and use it yourself and not tell me about it, but that's a chance I apparently have to take. And then it gets murkier if you use code snippets you have already developed within, I ought not insist that's part of the program and that you can't use it anywhere else. I dunno, I'll show fair discretion. I'm not sure, tell me your own terms if this sounds no good and I'll think about it. 1000 dollars doesn't buy a quick and dirty job though. That is a lot of money to me, just so you know. But it's that much because I know this is going to take time. Time for me to make sure you understand what needs to be done, and we can discuss it and your input will be valued as well, if you know expedient/better ways to do certain things I want done. I know how to do it in theory but the particulars of the system may make certain shortcuts possible.
I will not publicly post the trading method. I do not want to post it because the last thing I want is others doing it, because then it wouldn't work. The method will come if you agree to take the job, and I will ask you to sign a non-disclosure agreement (you can still back out of taking the job after learning more about it, but the non-disclosure agreement is still binding) before I even tell you anything about it except some things that are necessary for you to know just what you're getting into so you know if you're biting off more than you can chew:
I will say it is not an evolutionary algorithm (which is why it's annoying to me that the acronym EA is thrown around here willynilly, it's not an expert "adviser" anyway, it's an expert doer) but a deterministic but adaptive one. It is to involve trading a great many different entities all together, regardless of what chart it is applied to - for instance, if it is applied to a 15 minute EURUSD chart, it should completely neglect that chart and instead access the 1, 5, 15, 60 and 240 minute charts (let's restrict ourselves to that set) for not just EURUSD but any set of symbols that can be defined at the top of the mq4 file. In other words, using the iBars and iTime and iClose functions I am assuming (that's what is in my version). I also didn't just store iClose output data but MathLog(iClose ... because what we're interested in is ratios, not absolute changes, and a decrease of 1.000 on a natural log scale is a decrease by a factor of e, which is the same factor for anything as anything else, it's what allows you to compare apples to oranges in a sense. If you don't know what a natural log is, we've got a problem. I want someone who is not merely mql4 literate but math literate, and so would appreciate someone who has had coursework in (or simply knows about) calculus, differential equations and linear algebra, bare minimum. But back to the point, we're not talking about just trading the item the chart is about, we're talking about trading a set of things that is specified at the top of the program, and not just for the timescale of the chart either. So for instance, in the abortive program I tried to make, I had this before the start of the init function:
#define numberofentities 10 //Number of different things that are traded or examined
string nameofentities[numberofentities]={"EURUSD", "USDJPY", "GBPUSD", "USDCAD", "AUDUSD", "USDCHF", "NZDUSD", "USDMXN", "XAUUSD", "XAGUSD"};
Get the idea? In other words, it will collect data not from the particular chart it is applied to but from many different timescales on whatever set of symbols. I would ostensibly also increase numberofentities and put in futures contracts in there too, the only reason I don't have that there now is because you have to specify a certain month and I didn't want to hardcode something into my program that would cease to exist before I was done writing the program.
This program should also build up a database of price histories in the computer's local memory - stored as natural logs of the prices. This is because the computational time in taking a bunch of logs is too much. So the first thing it will do on starting, in the init function, will be to store that for all the defined symbols for all scales of time for a certain amount of allocated space for each one. Then it should figure out the starting time and ending time of each week - I don't want it to hold any positions over the weekend except for a minimal amount (more to be explained on that if you agree to the job) - nor should it jump in at the start of each week - it should take a little time to collect data at first at the start of the week and wind down the positions in the final hours, so there should be a maximum total leverage which is a function of time, which also means it must figure out from the data when the market will next be closed.
This program will then produce multiple covariance matrices - which it will invert - if you need an efficient method of inverting a matrix with minimal mean-square numerical error, I will provide this to you. It needs to determine the volatility on various scales of time (the mean square change in price over various displacements in time, and build up a weekly schedule of volatility) and determine a distribution of times taken for trades to be reversed (the time after buying a position after which it is closed - so for instance if you have .14 positions of EURUSD which have all been held for 2 hours and you buy .03 more, and then 30 minutes later you buy .04 more and are now up to .21, and then 30 minutes after that, you sell off .05 and then 30 minutes after that, you sell off .06, you would have sold off .04 after 30 minutes, .01 after 1 hour, .02 after 1:30 and .02 after 3:30 - naturally the bullish positions bought when it was already leveraged with a bullish bias will be the first to go, so the time for trade reversal will be not just a probability distribution but one which depends on the current amount of leverage and gets shorter the more heavily it is leveraged in the same direction - and can be quite a long time if in the opposite direction - namely if you're short .14 positions of EURUSD and you start covering it - buying it back - the amount of time until you're back to -.14 again could be a long time, whereas the time from when you're +.14 and buy more and back down to .14 will be much less time) and produce a mean and variance of the price of each of the symbols being traded in that time into the future and trade accordingly, which means possibly opening new positions and closing old ones, though probably in a small quantity unless something big happens. I will not tell you the optimal method of calculating this time or this probability distribution of times here, that's part of the method.
It needs to also separate all sorts of weekly schedules - a weekly schedule of mean square values of the spread (difference between bid and ask), and just in case it does exponentially grow to outrageous proportions, it should determine the bid and ask quantity and not put in a market order greater than the appropriate values, it should also not do ALL the calculations EVERY tic, maybe there's a way to make it ignore tics that happen while it's still processing the last one?
Naturally it needs to know about the broker's limits on leverage. Like forexbrokerinc has this fruity deal where you're allowed to leverage 200 to 1 until your account value gets to a certain level and then it becomes 100 to 1 and then 50 to 1. I don't know if there's some discontinuity that occurs at some specific time right in the middle of trading, but preferably it needs to not go haywire if there's some sudden transition like that. A few other things to be aware of.
If any of this sounds over your head, I don't want to hear from you. This is hard. Damn, damn hard. Hard enough merely to understand, hell I barely understand it, and actually writing code for it, that will be a monumental achievement. That's why even the 1000 dollar pricetag is negotiable. I can even sweeten the deal conditional upon my success in actually using it. If it performs as well as I think it will if implemented well, I'll share some of the money with you once I have it. I know the meaning of gratitude. I just have next to nothing now. Right now, 1000 dollars is a significant fraction of my total bankroll but I am offering it because I know how hard it will be to write this program. But it's for it done RIGHT. Not some halfassed thing. And these are words coming from someone who has written rather serious programs before, so you know it's no joke when I am giving up trying to get this rotten programming language to do anything useful - I wrote the microcontroller programs that run these clocks http://medexamtools.com/oscommerce1/advanced_search_result.php?keywords=clock and it's going to be hard to even explain it, I know it will. It's like nothing you have ever seen before, I guarantee it, so I'm not going to be concerned about your experience in the past, only if you think can do this. Though if you have anything worth bragging about, i.e. awards and accolades, trophies on the wall for programming or math competitions, tell me about it. It took me years and years and to figure it out, years of agony and adversity and I'm tired of it, but I know one thing, you'd better be at least as abnormal as me (more, since I apparently can't do it) if you even have a hope of doing this thing. I want someone meticulous, perfectionistic, obsessive-compulsive, outrageously intelligent and of course can actually get a frigging program working in this silly crippled C programming language variant (maybe I should call it C-- as opposed to C++ maybe it's only the metaeditor compiler that's crippled). Only freaks need apply.
Write to me first at my account deuteriumathotmailcom, that's the isotope of hydrogen at hotmail dot com. I would prefer to get a little dialog between us so I may decide if you're likely to be the right one for the job.