Interesting topic for many: what's new in MetaTrader 4 and MQL4 - big changes on the way - page 29
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
This is where the "almost" comes in, strange as it may sound. You can't just act statistically here. If 99% will coincide and 1% will not coincide, but at the same time 1% are local extrema (tops of ZigZag). That's it, the accuracy of the tester results is dead. And this is how it turns out in practice that the most important 1% carries these most serious distortions.
I do not even know how to explain the obvious things to the practitioner, who is far from it.
What is the spread of a bar in the MT5 history? I will write a light MQL4 script that will show the divergence. Then I will be able to draw a picture based on the results. However I doubt that there will be as many questions.
The most difficult thing is to explain the obvious things.
The spread of a minute bar (and MT5 history is based on minutes) is equal to the difference between Ask and Bid at bar opening.
You argued that [Low_Bid+Spread != Low_Ask] I have shown you the contrary (even though I have tons of work myself), you say that it is critical for most algorithms, the example shows that it is not so, and the information is recoverable in most cases.
You state that 1% of extrema but that is again an unfounded assumption, I was watching all day today (normal trading day with volatile sessions etc), just the extrema fit well into the MQ model, but the inter-bars (small shifts hidden in the shadow of extrema when slicing) have non critical discrepancies. For example, the HighBids of two neighbours are actually equal and when restoring them from the history, they diverge by 1 point, or vice versa the HighBids differ by 1 point and when restoring them they are equal.
I am not against introducing HighAsk & LowAsk + HighBid & LowBid of history (it will increase the accuracy of pipsarian testing), but to say that it is critical for most algorithms (imho) is not making sense.
Sorry, but I've ineffectively spent a day of my life testing your thrown-away theory, so you're on my blacklist (whatever you say == grandma said two words).
SZS here's a look at how the minute bar is dynamically built, the indicator only builds on OHLC & Spread data. I do not see any difference with HighAsk & LowAsk + HighBid & LowBid model.
In the indicator High displays HighAsk, Open --> HighBid, Low --> LowBid, Close --> LowAsk. All these HighAsk & LowAsk + HighBid & LowBid models are recoverable from the MQ model
I don't even know how to explain obvious things to someone who is far from it.
Code and pictures, code and pictures.
Maybe you should stop calling everyone stupid.
It is not dullards who are here, but those who view everything only under concrete implementations and code.
Give me algorithms, not empty words!
The spread of the minute bar (and MT5 history is based on minutes) is equal to the difference between Ask and Bid at the opening of the bar.
Great! I.e., let's clarify the MT5 history model: OHLCV_Bid + Open_Spread.
You have stated that [Low_Bid+Spread != Low_Ask] I have shown you the contrary (although I myself have a lot of work), you say that it is critical for most of the algorithms, the example shows that it is not so, and the information is recoverable in most cases.
You apparently see what you want to see. So, you claim that this equality is true: Low_Ask - Low_Bid == Open_Spread == Open_Ask - Open_Bid. Sounds about as plausible as "the spread inside the bar is fixed".
Call for help:
Forum on trading, automated trading systems and trading strategy testing
Interesting topic for many: What's new in MetaTrader 4 and MQL4 - big changes on the way
MetaDriver, 2013.08.08 00:00
Nicholas let's split the question.
1. whether they are equal or not?
2) Why?
The first one you can verify. You can check it.
The second is largely irrelevant. Nevertheless, you can easily understand the mechanism of this difference, even without an explanation.
I think you have a problem with the first point. It's like, "I don't believe it because I don't understand it."
However, facts do not depend on whether we understand them or not. They don't care. Limiting our perception to understandable facts is a very popular way of maintaining our self-esteem and keeping our beliefs chaste. However, pops don't rule in forex. Here, pops are fucked, wholesale and retail.
MetaDriver, I know I'm being cheeky, but I have to.
You claim 1% of extrema but that's again an unfounded assumption, I've been watching all day today (normal trading day with volatile sessions etc.), just the extrema fit well into the MQ model, but the inter-bars (small shifts hidden in the shadow of extrema when slicing) have non-critical divergences. For example, the HighBids of two neighbours are actually equal and when restoring them from the history, they diverge by 1 point, or vice versa the HighBids are 1 point apart and when restoring them they are equal.
Are you sure you checked there (LowAsk)? The HighBid of the MT5 history will always be exactly the same as the real HighBid you had. The reason above isthe MT5 model: OHLCV_Bid + Open_Spread.
I am not against introducing HighAsk & LowAsk + HighBid & LowBid history (it will increase the accuracy of pips testing), but to say that it is critical for most algorithms (imho) is bullshit.
Make up your mind whether it will increase or not. The secret is that it will increase the accuracy for absolutely all TS. And it will be especially critical for TCs that have small MO. And don't say that small MO is a pipsqueak. Because that, too, is knowingly wrong. Even strategies holding open positions for hours and taking dozens of points of profit may have small IR. Because MO = Profit / AmountPositions.
Sorry, but I have ineffectively spent a day of my life testing your thrown-away theory, so you're on my blacklist (all that you say == grandma's two words).
I didn't think this crap from you old timers would take so much time. I'm sorry to have wasted your time.
ZS here's a look at the minute bar in the dynamics, the indicator is built using only OHLC & Spread data. I do not see any difference with HighAsk & LowAsk + HighBid & LowBid.
In the indicator High displays HighAsk, Open --> HighBid, Low --> LowBid, Close --> LowAsk. All these HighAsk & LowAsk + HighBid & LowBid patterns are recoverable from the MQ pattern
Can't you see the loss of information? You can just as easily use the symmetric model: OHLCV_Avg+OpenSpread. Avg = (Ask + Bid) / 2 - that's why it's called symmetric with respect to Bid and Ask, i.e. it has no discriminative (errors and inaccuracies are the same) to any of the two prices.
I need to think of some logic that could dissect your understanding.
Code and pictures, code and pictures.
Maybe stop calling everyone dumbasses.
It is not dumbness that sits here, but those who view everything only under specific implementations and code.
Give me algorithms, not empty words!
In my opinion, literalism is 100 times harder to understand and comprehend. There's not much more to it than that. Damn, it turns out that few people understood what he's been racing in the tester for years.
It seems that the rules of building TS, research, money management, writing a fighting robot, etc. need a detailed explanation. I can't pull it off.
I feel like an idiot because I certainly don't have any outstanding abilities.
To everyone:
Why do you test on a shitty brokerage company (even though it has a name), when there are objectively better ECN/STP-platforms for trading conditions! You can even look it up on the so beloved spread indicator. It is a fact that there is no one spread that is better right now. For example, the average EURUSD spread is ~0.2 pips (negative quite often).
You test your TS for example and see an MO of ~ 2 pips on some GBPNZD. Well, that's pretty cool. Go to ECN/STP and see that your same TS on GBPNZD will have MOs ten times higher.
I don't know how you test and optimise. But it all seems to be very illiterate. You are cutting your own potential profits.
Do you want me to prove the need for an asc story? Gather a quorum, like you did with the libs. I don't fucking need it. Even disappointed in my participation with the liquorbase. But if I have to, I'll scrape together some really good examples. Really, I don't believe they'll understand.
And you are trying to conduct a discussion on HFT and Level2. No way, if the simplest things cause such a stupor of understanding.
I don't think anyone is stupid, self-criticism is over the top. But there's something wrong with your understanding.
You apparently see what you want to see. So, you claim that this equation is true: Low_Ask - Low_Bid == Open_Spread == Open_Ask - Open_Bid. Sounds about as plausible as "the spread inside the bar is fixed".
The divergences are exactly within the floating spread. Those points where Ask==Bid are definitely divergent, but a check showed that most of these divergences are in the intra-bar space and far from extremums.
Again the divergences (the vast majority 1 point). So bootstrapping the model (giving huge traffic savings within even a single dilling) is justified.
I didn't think this shit would take so much time from you old timers. I'm really sorry to have wasted your time.
I wrote the indicator in 10 minutes, debugged it for about 20 minutes. I spent the whole day testing your hypothesis on real ticks.
Can't you see you are losing information? With the same ease we can use a symmetric model: OHLCV_Avg+OpenSpread. Avg = (Ask + Bid) / 2 - that's why it's called symmetric with respect to Bid and Ask, i.e. it has no discriminative (errors and inaccuracies are the same) to any of the two prices.
I need to think of some logic that would be able to dissect your understanding.
I can see the loss of information, moreover I've seen it since the days when Prival was busting his spears for introducing tick history. I don't need half-measures, the information is lost when passing from ticks to bars, if the bar model remains, it makes no difference whether you write it as it is or in some more complex, for most EAs it is not critical, for pips machines the information from the bar about tick's behaviour pattern is lost forever. So what's the moaning about?
Get out on the straight and narrow already and convince me that ticks are important and necessary.
Make up your mind whether she will promote or not. The secret is that it will increase accuracy for absolutely all TCs. And will be especially critical for TCs with small MO. And don't say that small MO is a pipsqueak. Because that, too, is knowingly wrong. Even strategies holding open positions for hours and taking dozens of points of profit may have small IR. Because MO = Profit / AmountPositions.
Yes, it will increase it a little, for all TC, but it will allow to lose more confidently :)
Algorithm taking 2-3 pips when you change feeds in the DC will immediately start to drain, those as soon as you give the DC toxic, they will change the filter and recover the losses at your expense.
So don't delude yourself that increasing accuracy by 1-3 pips will immediately make a lot of grails.
ZS Orevoir, that's it for today.
I need to think of a logic that would be able to unlock your insight.
I understand what you're talking about for a dozen pages. But it's the same as you would come into the 80's and start telling people that they have to rebuild their whole lives and be ready for the 90's.
The main reason for this is that the majority of the people in the world are not interested in the real world, but in the real world. The real roar and wearing pampers is impossible to do without, not to mention the number of lost deposits, which is needed until the epiphany.
I cannot do without real earning history and open TS (even if it is dead, but it still has profitable history).
I can't understand, why do you need it? Why bother (to explain the basics, to waste time, etc.) when you have almost 100% of pseudo traders who are not ready to get the information?
I understand what you are talking about for a dozen pages. But it's the same as you would come into the 80s and start telling people they have to rebuild their lives and be ready for the 90s.
The main reason for this is that the majority of the people in the world are not interested in the real world, but in the real world. The real roar and wearing pampers is impossible to do without, not to mention the number of lost deposits, which is needed until the epiphany comes.
We cannot do without real money history and open TS (even if it has died, but it still has a profitable history).
I can't understand, why do you need it? Why bother (to explain the basics, to waste time, etc.) when you have almost 100% of pseudo traders who are not ready to get the information?
So explain to us unlearned people, what do you understand?
For example I see that the effectiveness of the proposal is minimal and the cost of its implementation is huge (and it's not just the cost of MQ), while in MT4 and in MT5 there are a lot of postponed to better times deficiencies, for example I have two years hanging application to SR for valk-forward testing (very good, karacho, but I have too much to do).
Do you write anything at all?
I can't understand why you need it. Why go to so much trouble (describing the basics, wasting time, etc.) when the number of pseudo-traders who are not ready to absorb the information is almost 100%!
Everyone needs an accurate tester in MT(they don't realise it yet), except me and a few other algotraders with their research infrastructure. So I'm not trying for myself - an ideological message. I didn't need a litmus test either.
1. Needed.
2. Understood.
3. this is all very fine and wonderful, but:
->
Code and pictures, code and pictures.
.......but how else? I don't know about others, I speak for myself: all these "lowbid" and "hiasc" are very badly absorbed when reading them, not to mention that a person starts thinking "why do I need it the way it is and why do I need it the way that one suggests?
It's not about the quality of quotes or even about providing an ascetic history. We're talking about the fact that the current metatestors are not suitable for normal algotrading. And what is the simplest thing you can do to make at least an MT4 tester usable.
My God!
You recently understood the principle of price movement, today you realise that the tester is useless except for checking the correctness of the algorithm, so what next? )
The question is: What's the point of all this?
God, he looks like Don Quixote with his windmills.