[Trader's Handbook] Draft articles, "out of pocket" discussions - page 25

 

A few words about Forex trading through a bank.

Legal entities can have different statuses, one of which is called "Bank". To become a bank is to obtain an appropriate banking licence from a particular regulator (the Central Bank, for example).

It is naive to think that the presence or absence of any regulation (not necessarily banking regulation) seriously affects any kind of reliable and profitable trading conditions. The owner of a legal entity (a bank in particular) determines a lot.

You can see how the biggest banks are fined for fraud, how many go bankrupt, and how those very licences are revoked. Moreover, money is even taken away from depositors (Renat, as a resident of Cyprus, I think he could tell you a lot about this topic).

Let us leave out the slippery subject of taxes when trading through a bank. Let us look directly at the trading conditions. There is only one way to measure them: how much potential profit you can make compared to other trading venues.

This depends directly on the perfection of the algorithmic and technical trading infrastructures. Obviously, the presence/absence of a banking license has no bearing on this infrastructure.

Forum on trading, automated trading systems and testing trading strategies

Trader's Guide: orders, prices, money, funds, currency

hrenfx, 2013.06.17 12:07

It is therefore wise to choose a place to trade through the most advanced aggregator, which is also neighbouring with a strong legal department.

That is, in addition to the fact that the trader will be anonymous in the general flow of the aggregator. He will also get free professional legal assistance in dealing with cancelled trades. Free of charge, because the legal department is not defending the trader's rights, but the face of a much larger aggregator owner, on behalf of whom, among other things, this very trader has made deals. For this, of course, there has to be a principled position on the part of the owner.

Don't forget also that absolutely any regulation is essentially a restriction on freedom of action. And this applies not only to restrictions on machinations, but also on points that are beneficial to the trader. Sometimes these restrictions are, frankly speaking, of the dumbest nature, in terms of running this business successfully on a transparent and mutually beneficial scheme. That is why not every legal entity, even if it has all the possibilities to obtain a banking license, has the desire to become a bank.

At the moment it is quite simple to choose the best trading conditions - it comes down to choosing from less than ten brokers. And few of them have a powerful algorithmic and technical trading infrastructure in development. That is, the step of choosing a broker has become the simplest. The only thing to think about is creating your own profitable TS.

 
Trading at kitchen DCs is idiotic, which unfortunately almost everyone suffers from. Even trading at big STP brokers is sometimes stupid. Again, the choice of places with the best trading conditions is minimal. Some of them even have serious regulations from the common man's point of view (e.g. FSA (FCA)). In my opinion, the question of choosing a broker has long been irrelevant to algotraders. All of them have made up their minds and only deal with their TS.
 

Better here:

Forum on trading, automated trading systems and trading strategy testing

Market Laws

hrenfx, 2013.07.04 22:14

Every few months I go there and quickly IMMEDIATELY browse for patterns I don't know much about yet. Filter very quickly (the right items):

  • Good broker.
  • A lot of transactions.
  • Availability of commission is desirable.
  • Each pair is considered individually.
  • Presence of conspicuous trade time limits (days, hours).
  • A lot of trades with small targets.
  • If the profit in any column is positive, the number of pips should also be corresponding (without negative values).
  • At least one of the lines on many charts has a stable view (or such sections).

In general, there must be a clear statistical significance and at least something that immediately catches your eye.

Then a more detailed analysis of the remaining ones (~ 10):

  • Analysis by different periods of history.
  • MAE/MFE.
  • Classification of TS: Breakout or rebound. A breakdown is unlikely to reach this point.
  • Probably something else.

It all takes about an hour. You just have to understand well the meaning of each figure and the capabilities of the service. Almost always the transaction history is closed, even the profit and commission data is closed. But I look at the open stats last. As a rule, they have enough simple research tricks to form at least an opinion.

I do not use other services. The best one for detailed analysis is MT4i. But I have an excellent feel for what and where to dig. With all those features the explorer interface is much more complex there.

If the quote can be obtained, it does not matter where from, a parser is created and attached to the history in the terminal. Then we have our own methods. To be honest, I have never gotten to this stage. Because I haven't encountered anything so far, that would make me want to overcome my laziness. But there are people who are relatively good at solving TS.

Types of TCs are few (not classified). Usually they are some very simple channelers (adaptive channel to bounce) with a time limit. The trick is not in writing the channeler, but in tuning it intelligently. Various optimization criteria, trying all possible symbols (synthetics are also possible), enabling various filters. In general, the right market is searched for under the TC that has already been found.

For this reason a different research approach is dictated to a large extent - from the opposite ( I mentioned it a long time ago). In general, in order to make money, the most difficult thing is to overcome laziness. And in passing to turn the brain in a different way.

I have never met anyone with complex profitable TS. Almost none of the owners of these systems do not use their testers, they just use MT4-optimizer on M1 with genetics on.

And also one and the same TS at different brokers (even STP) can show opposite results. Here we need to choose the broker with the best trading conditions. Fortunately, nowadays it is elementary to do so. And if you can write your own tester, which is much more accurate in its readings, then you can squeeze out a lot, which others on dumb testers are not able to do and do not even think about.

As for the tester, it takes one or two days to write. You just need not set the task to write a universal framework, but something simple and fast. This approach makes you realize that you've been an idiot to write this task, thinking it impossible to solve. Generally speaking, a tester is the easiest thing in the world. The main thing in this business is not to stumble. You don't have to write a visualisation, for example. You just input your calculations into some matrix, and the visualization is ready to use. You come up with your own optimization criteria, try your own experiments. Nothing abstruse, in general.

That is how it is in general terms. Further, all the technical details are coming - ways to reduce commission, improve execution, increase the liquidity ceiling, take into account positive slippage and re-jacks. In short - to more or less coincide the tester results with real money. But we need to get to this point. And at the initial stage the above paragraphs are more than enough.

P.S. I realize that it is unlikely anyone will teach me something new, but it would be so nice to see the presence of even once passed, but sensible reasoning. Algotraders are assholes: coddling.


 
hrenfx: Rough algorithm for getting T&S data from Level2.

Assume that we have a certain sequence of Level2 values - large vectors where each price level corresponds to a certain volume (gangs). Let us number elements of this sequence from zero (current) to the past.

ECN (exchanges).
We compare Level2[0] and Level2[1] with each other. If Bid[0] >= Ask[1], then all gangs from Level2[1]_Ask, which are not above Bid[0], get into T&S. Similarly for the situation Ask[0] <= Bid[1]. In other cases, nothing gets into the simulated T&S.

For ECN the main idea seems to be clear - a large market order is filled by Limit orders, consequently eating opposing gangs until it reaches acceptable slippage level, the unfulfilled part (if there is any) is pending order.


hrenfx:

STP (preferably a lot of LPs).
Vector_Ask[0] = Level2[0]_Ask - Level2[1]_Ask. This vector sums all negative values from the best (by price) band until the first non-negative band is encountered. This sum is written into T&S[0]. With Bid, it is the same.

But the T&S estimation for STP is a problem, I did not get it right. I don't get the main idea (I may have misunderstood the calculation algorithm). If you understand it, please explain it by example.

 

A rough algorithm for getting T&S data from Level2.

Let us imagine that we have a certain sequence of Level2 values - large vectors where each price level corresponds to its volume (bands). Let us number elements of this sequence from zero (current) to the past.

ECN (exchanges).
We compare Level2[0] and Level2[1] with each other. If Bid[0] >= Ask[1], then all gangs from Level2[1]_Ask, which are not above Bid[0], get into T&S. Similarly for the situation Ask[0] <= Bid[1]. In other cases nothing gets into the simulated T&S.

You can make a big mistake here, as the price can change without actually executing the trade, just rearranging the limiters. I would suggest the variant with the last chart and Level 2 volumes. Sometimes, in the absence of last price, the algorithm automatically sets the bid or ask price in order to remove gaps and make the chart easier for analysis.

 
over2u: ... You can make a big mistake here, because the price may change without the actual execution of a deal, just rearrange the limiters.

Since Level2[0] and Level2[1] are separated by a single tick, placing/removing a Limit that does not cause counter-execution causes a change in only one gang. A particular case of such a change can be a change to one (and only one) of the bestbands (Ask or Bid). Externally, this appears as a widening/decreasing of the spread to one side. The case"Bid[0] >= Ask[1]" means the change of both bestbands in one tick. I gave my explanation of this case above (large market order execution). But this is for consecutive execution.


over2u : ... Rough algorithm for getting T&S data from Level2... ECN (exchanges) ... I would suggest the variant with chart last and Level 2 volumes.
from here

If a trade is executed, its price and volume is called Last. And this information is also broadcast by the exchange. The flow of Last data is called T&S

 
GaryKa:

Since Level2[0] and Level2[1] are separated by a single tick, putting/removing a Limit that does not cause counter-execution causes a change in only one gang. A particular case of such a change can be a change to one (and only one) of the bestbands (Ask or Bid). Externally, this appears as a widening/decreasing of the spread to one side. The case"Bid[0] >= Ask[1]" means the change of both bestbands in one tick. I gave my explanation of this case above (large market order execution). But it is in the case of consecutive execution.


from here

You may have a situation when a brokerage firm rearranges its orders according to its MM strategy according to the prices from the CME. So we will have exactly the movement you are talking about without actually hitting the Limiter's market order. This is possible when you have your own ECN system in which there is little trading, but to fill the liquidity to which either other ECNs or LPs are connected. There will be a seemingly paradoxical situation here - a change in price on the FI without actual trading.

I just want to emphasize that it is very difficult to collect quality data on trading volumes without a centralized platform and seemingly harmless assumptions can distort the final result.

 
hrenfx:

This is just a motivation project (launched), no additional information. I've already set out practically all the background information here.

People just do not want to know (learn) anything, unfortunately. So if there is any interest to be seen, it is as a sociological study on the subject of inquisitiveness.

In principle, anyone who cares about the general level of knowledge on the near-trading topic can simply link to the educational material wherever they see fit.

One can not only be a writer or reader, but also contribute to the dissemination of knowledge.

P.S. As for my writing, I disclaim any copyright - bullshit. The main thing is to get the point across.

As a rule, people want to study and people want to know, your intelligence helps people to tear up their "stereotypes" and "old templates", and somebody after reading your "writing" finds a missing puzzle in the overall picture of trading, lots of good information which shows the other side of trading

PS. Thanks a lot for organizing this thread and the "trader's guide" -sergeev, big thanks tohrenfx.

 
over2u: ... This is possible when you have your own ECN system in which little trading takes place, but to fill the liquidity that other ECNs or LPs are connected to.
You are already talking about ECN/STP, and for that you need to first understand how to evaluate T&S in STP
papaklass:

I honestly don't understand what you are disagreeing with in my post (if you are opposing it).


hrenfx, sergeev
It would be a good idea to add a post about ticks and the main types of trade matching algorithms
 
hrenfx:


How to read a libretto.


I moved this post to the front page