again on the tester - page 9
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
I can't ask intimate questions about your TS, but at least give me a hint - at what sane TS you need such accuracy in ticks, not to mention the fact that the historical ticks would be different in different dts,
and the difference may be greater than with simulated.
it's not the trading system, it's the quality of the prediction. Look very carefully and read here 'Tick collectors. Optimization. DDE in VB (VBA)' (Prival 30.12.2007 00:44 ). Take a pencil and a ruler and draw a forecast. Everything will become clear and understandable. If the current measurement is +- lapot, the forecast is +- 100 lapot to the right of the sun. These are the basics of forecasting methods and there is no getting around them.
it's not about the trading system, it's about the quality of the prediction. Take a very close look and read here 'Tick collectors. Optimization. DDE in VB (VBA)' (Prival 30.12.2007 00:44 ). Take a pencil and a ruler and draw a forecast. Everything will become clear and understandable. If the current measurement is +- lapot, the forecast is +- 100 lapot to the right of the sun. These are the basics of forecasting methods and there's no getting around it.
No, you can't do that.
This approach is used to calculate the angle of the slingshot when shooting at sparrows.
And again - you will get different history ticks from different DTs, more different than the difference with a generator in the tester.
Being in a society of technicians, play by the technicians' rules. Be kind enough to prove your "ticks cannot be trusted" claim publicly with detailed facts (screenshots and exact tables of real and simulated prices that will show the discrepancy). Make it simple and technical by publishing a table of bid-ask prices in an hour of collected real ticks and simulated by a tester based on minutes. With all descriptions so that any user can replicate your experiment.
I'm sure you'll be quick to retract your words when you try to find evidence.
Done. Now you refute just as clearly beautifully and correctly. Or retract your words.
No, you can't do that.
This approach is used to calculate the angle of a slingshot when shooting at sparrows.
And again - from different DTs you will get different historical ticks, more different than the difference with the generator in the tester
is the prediction of a straight line, an ordinary straight line. If the model is more complex (not straight) then the forecast is different, less accurate (there is also a concept of model error). There are errors here and there, and the result -> no possibility to forecast for quite a long period of time.
And the most surprising is that they think it is important for pipsing. This is nonsense if the forecast horizon is 24 hours, i.e. the goal is to obtain a forecast in 24 hours with accuracy +- 10 points, let them call it whatever they want. We do not need anything else for the trading system.
And if a simple straight line can be predicted with a minimum error of +-1000 points for the next day because of the inaccuracy of the current estimate. There is no sense to talk about more complex models.
This is the negligence of ticks, 1 point here and 1 point there, but the result of this negligence is the impossibility of long-term forecasting.
is the prediction of a straight line, an ordinary straight line. If the model is more complex (not straight line) then the prediction is different, less accurate (there is also the concept of model error). There are errors, there are errors, and the result -> no possibility to forecast for quite a long period of time.
And the most surprising is that they think it is important for pipsing. This is nonsense if the forecast horizon is 24 hours, i.e. the goal is to obtain a forecast in 24 hours with accuracy +- 10 points, let them call it whatever they want. We do not need anything else for the trading system.
And if a simple straight line can be predicted with a minimum error of +-1000 points for the next day because of the inaccuracy of the current estimate. There is no sense to talk about more complex models.
Here it is - the neglect of ticks, 1 point here and 1 point there, and the result of this neglect is the impossibility of long-term forecasting.
From the "primary source" to the figures in the Market Watch a darn lot of filters with unknown for us filtering algorithms (distortions) which bring a lot of clutches into your minute values.
and it's not just the minutes that vary from dtz to dtz, the oscillator in the tester is an innocent child in the background of the filter chain.
There is absolutely no information on teak history on the above link. The link to the api is not "publicly available tick history".
I have no doubt that you haven't even used this api in practice.
You can use the API, open a demo account in a few seconds, and use that API to obtain the publicly available tick history in whichever form you want. The platform implements a multi-currency tester on real ticks. A more accurate tester is hard to imagine. API is much more complex than MQL4 (not because it's Java), because multithreading is implemented, you can send as many trade orders at once, without waiting in queue for the response, like in MetaTrader. Plus we have a much more complex scheme of order execution (not the API, this is the laws of the real market). For example, partial execution (not for the whole requested volume). Plus there is the concept of liquidity. And there is full access to the history and current values of the market depths. Plus all the power (capability, not speed) of Java.
And, as I mentioned before, all these laws of the real market can be tried through the MetaTrader4 platform...
Again, it will be much more difficult to create a system on this API and in the near future on MetaTrader4, under the laws of the real market, than it is now under RC even with floating spreads.
About use in practice: my system works through this API. And you only have to write a few lines of code to get the tick history.
Yes, there is from April 2007. But that's disastrously low for tests and few people can apply it.
And we have a really free, available in a couple of clicks minute history from 1999. And our tester perfectly simulates the development of bars on this minute history.
The ticks are not insufficient for tests, but more than enough.
There's a myth that the tick history is only necessary for pipsing and arbitrage. This is not true. For example, the tick history can be useful in developing multicurrency systems.
The history may be from 1999, or from 1970 - it's a good marketing move. Applicable for theorists. In practice, it is not necessary at all. While your history is still completely missing data on spreads and liquidity.
Your tester simulates ticks perfectly for tests + remarkable speed at the tester as a universal solution on the trading conditions of most DCs. There was stringo's claim that the simulation is SUPER (not a quote, by implication). And a suggestion on his part to refute it. A simple example of a possible nuance of imperfect tick modelling was given above.
As a practitioner, for most tasks, especially at entry and mid-level, the MetaTrader4 platform is excellent.
From the "source" of quotes to the figures in your market overview window, a shitload of filters with filtering (distortion) algorithms unknown to you and me, which introduce a whole bunch of noodles into your minutes, and the same minutes
and not only do different DTs have different minutes, compared to a chain of filters, the generator in the tester is an innocent child.
That's why I wrote that I don't care how it's generated. I understand a priori that no algorithm will ever restore temporal structure if history storing format is like now. A way out suggested. Let the developers decide if they need it or not.
There is a way out of the situation you describe too. I am talking about the "primary source". I have already written somewhere in my MetaTrader 5 wishes that it is necessary to give me possibility to connect to different sources of quotes simultaneously. If I want to use them I will connect eSignal (Bloomberg) and receive current estimation from these quotes. But I think developers will not do it, because brokerage companies will be against it.
From the "source" of quotes to the figures in your market overview window, a shitload of filters with filtering (distortion) algorithms unknown to you and me, which introduce a whole bunch of noodles into your minutes, and the same minutes
>> Not only do different DTs have different minutes, but a generator in the tester is an innocent child.
i don't have anything to do with the filters... why are you so hung up on them...
understand, i work with a specific brokerage company and it gives me a quote, not forex... it doesn't matter if the dc uses filters... "price takes everything into account"... including filters...
let's say I have a neural network working - it's looking for patterns in my dealer's price behaviour with all its filters taken into account...
I just can't understand why Prival needs such accuracy... Why can't you use for the same neural network, say, some kind of weighted average price of a minute bar...
Why can't the same neural network be used with some kind of average price for a minute bar... at the moment I am inclined to think that it is quite suitable...
One is enough. Recorded tick history, as I wrote above, for two years (since January 2007). A very convenient way (as one would like) to retrieve them via API. Publicly available.
About brokers with traders. This broker will be available on MetaTrader4. And it will not be a DC, but precisely a full-fledged broker on your own platform with available initial and minimum deposit.... wait for the new year.
Do you have the broker's name?