The work of robots on the real. Why people are disappointed - page 6

 
nowi:
Maybe you can share with us how to deal with such pseudo information?

Standard. They get datafeed\tradefeed from the broker on the real stream. They take a library and write a software (terminal). They develop different mathematical models in it and make a simulation block for real trading. They test it and try it in real trading.

As for how to analyze the order flow there is enough literature (though in most cases it is of no value) both in English and in Russian, all in the open sources.

 
Programmer96:
= how to trade based only on a stream of quotes without considering candlesticks at all =
I don't know if you mean the same thing as I do, but I'll take my chances...
I 'draw' candlesticks myself. Virtual ones, of course. No one would argue that if you trade on a day, the risk is minimal and the profit is maximal. But this is when trading manually.
If we use МАшку or other Inductors on the days and with decent averaging, then we obtain only a few deals per month. However, as a rule, all of them are profitable.
I've worked out another method (not for МА, I use my own practice) - the same daily (or hourly or four-hourly or weekly - it's not important) I build on minutes, shifting the starting point every minute, i.e. I have NO jumps on opening a new candle, so the reaction to price changes is faster and more accurate.
it's much simpler for me, very primitive
i have several trades a month, all of them profitable... this system suits me fine! i have not yet reached such heights of mastery
 
Going_Crazy:

Standard. They get datafeed\tradefeed from the broker on the real stream. They take a library and write a software (terminal). They develop different mathematical models in it and make a simulation block for real trading. They test it and try it in real trading.

As for how to analyze the order flow there is enough literature (though in most cases it is of no value) both in English and in Russian, all in the open sources.

As for the order flow in the forex market, as for me it is a fake.
 
Programmer96:
That one is disappointed is a fact.
I have some observations and conclusions, but I will post mine later, now I would like to hear what anyone thinks on the subject.

It strongly depends on how high-frequency the robot is. If it makes 1-2 trades a day, then the difference in quotes between the real brokerage company and those generated by the tester can be easily ignored. If it opens at the beginning of TF there will be almost no difference. But my multicurrency scalper works on tick data and difference between tick data in the tester and between different brokerage companies is huge. Therefore, I take real tick data, use it to run the algorithm in Matlab, improve the algorithm and set parameters. And only then I transfer it into my robot. Here, I gave an example of a jump at USDCAD news with markers for clarity; look how the price plunges downward in 30 seconds. The tester will generate a smooth smoothed line within M1.

USDCAD 4 April 2014

 
nowi:
Once again: I do not take the flow of orders on the forex market seriously. for me, it is a fake. a flow of rubbish that is just taken from a background and there is nothing to analyse.
And what is a "stream of forex orders"? Is it a typo and do you mean a stream of quotes?
 
VDev:

It strongly depends on how high-frequency the robot is. If it makes 1-2 trades a day, then the difference in quotes between the real brokerage company and those generated by the tester can be easily ignored. If it opens at the beginning of TF there will be almost no difference. But my multicurrency scalper works on tick data and difference between tick data in the tester and between different brokerage companies is huge. Therefore, I take real tick data, use it to run the algorithm in Matlab, improve the algorithm and set parameters. And only then I transfer it into my robot. Here, I gave an example of a jump at USDCAD news with markers for clarity; look how the price plunges downward in 30 seconds. Tester will generate a smooth smoothed line within M1.

Analyzed what was going on that day. )

On 4 April 2014 at 2.30pm news came out in Canada with positive Employment Change and Unemployment Rate.

At the same time in the US news came out with negative figures inNon-Farm Employment Change,Non-Farm Employment Change,Unemployment Rate andAverage Hourly Earnings.

Since the news is released with a slight delay and there can be strong slippages in such strong movements, at best one could take at least half of this movement.

At the same time, the analysis of the relative strength of the USD on that day shows that you could have easily entered the SELL on a tidy pullback up and tried to take everything. But since you don't know in advance what the indicators will be released, you can get into a more complicated situation. After all, the indicators may be contradictory and it is difficult to predict whether the price will go in one direction or it will be a chaotic disturbance. ))

For those who stick to conservative methods, during the release of several important news from different countries, it is better to refrain from transactions at all, waiting for a pullback (if such a unidirectional movement took place). If you look at what happened next, you can see that at the opening of the American session, this rollback was about a third of that movement. One news release from Canada( Institute of Statistics CanadaPurchasing Managers Index) at 16.00, which was negative, was a reference point. First of all, we should wait for the data of the news and how the price reacted to it. Because the price reacted with a pullback, and the totality of all the indicators, released earlier from Canada and USA, are negative for the American dollar, and besides, as it was said before, the relative dollar's strength is negative too, we could confidently open the SELL trade.

Let everyone determine the exit for themselves. This is homework. ;)

 
tol64:

Analyzed what was going on that day. )

I remember there was an analyst at the late Broko in St. Petersburg who gave live lectures on how to analyse the market, unfortunately I don't remember his name. He explained everything like clockwork! Once on the forum they took him for a challenge: "Open an account and show me how to trade))). He opened his account for $300 on CFD. He held it for a month or two, +/- going back and forth, then quietly closed it with deficit.

I'm not making fun of you, it's just that everything is perfectly explained in hindsight. Great example - Elliott waves ))))

And to the point. If my article contains the article as a base for calculation, I would not expect it to be true. He asked to develop a robot that would use his forecasts for news and their time of release. He was even going to buy a subscription to the Dow Jones's flow, which costs from $6000 per month. I persuaded him not to get too excited and to try forexfactoring for a start.

The robot has been showing some gains as a result, but only after I added my own tools to the robot for on-the-fly analysis, as it was losing on pure forecasts with stops and takes. It is like the weather in St. Petersburg - sun is shining, birds are singing and in half an hour there is a storm with thunderstorms )) Scalping and DSP analysis are more profitable.

 
VDev:

But my multicurrency scalper works on tick data and the difference in tick data in the tester and between different DCs is huge.

Here is a jump on USDCAD news as an example, I put markers there for illustration, look how the price goes down in 30 seconds.

This is my personal viewpoint, based on practical experience:

(I'm going to omit words about latency, because it's clear anyway)

MT4 does not show bid>ask arbitrage within the symbol.

If you watch Level 2 (without delays in updates), the arbitrage display is a natural phenomenon there.

It happens that not just Best Bid > Best Ask, but full order book Bid > full order book Ask, nothing is seen in MT4.

At full order book Bid > full order book Ask (where parts of the book do not even intersect), how will you be executed in the sell, for example? Question!

Late indicative quotes are the scourge of any history (and real), they should be taken into account during modelling (testing).

There is also no point in testing at the best prices, there are other effective methods that ensure understanding that the test is close to what it would show in real conditions.

Regarding April 4 and the USDCAD, that is not what the chart looks like. There was a bid>ask present in those 30 sec, one can only predict what the execution would have been with a very large assumption (if there are no accumulated statistics on such cases). Bid prices and volumes were just outdated broadcasts (and the quote archive was also written to the server).

Generally truthfulness of brokers' history depends on the quality of LP (if they are technically backward, which introduce artefacts), where the servers are located, where you keep your algorithm in relation to trading servers, on what hardware and even on the quality of the code, library, optimization (how much is lost within the program).

The more trades the trading system generates per day, the more difficulties and nuances are waiting for the developer.

 
VDev:

I remember there was an analyst at the late Broko in St. Petersburg who gave live lectures on how to analyse the market, unfortunately I don't remember his name. He explained everything like clockwork! Once on the forum they took him for a challenge: "Open an account and show me how to trade))). He opened his account for $300 on CFD. He held it for a month or two, +/- going back and forth, then quietly closed it with deficit.

I'm not making fun of you, it's just that everything is perfectly explained in hindsight. Great example - Elliott Waves ))))

And on the point. Not so long ago, I was contacted by a man who is a fan of trading on the news, who really knows how to trade and successfully trades manually. He asked for a robot that would use his forecasts for the news and their time of release. He was even going to buy a subscription to the Dow Jones's flow, which costs from $6000 per month. I persuaded him not to get too excited and to try forexfactoring for a start.

The robot has been showing some gains as a result, but only after I added my own tools to the robot for on-the-fly analysis, as it was losing on pure forecasts with stops and takes. It is like the weather in St. Petersburg - sun is shining, birds are singing and in half an hour there is a storm with thunderstorms )) Scalping and DSP analysis are more profitable.

I can't even imagine how to write a robot based on the news. Every time there is a unique case. Of course it is good that they tried to implement it, but personally I would have said in the beginning that it was a crapshoot. ))

I think that fundamental and economic analysis is a good addition to trading strategy if trading manually. But of course the news alone will not get you far. It is only an addition to everything else.

Nothing prevents making money by scalping also with fundamental analysis. ;)

 
Going_Crazy:
...

Thank you. The secret is accepted for consideration. ;)

P.S. Noticed that you deleted your post to which I was responding, so I deleted the quote.