Ward 6 - page 68

New comment
 
DmitriyN:
Michael, what is your estimated share of profitable trades in Matcad? Does it make sense to use different TP and SL values?
In the sense of the theoretically expected value? Without spread - asymptotically close to 2/3 exactly when increasing the number of trades to infinity. With the spread taken into account, of course, it is noticeably lower. About 60 percent at the most comes out. The model has some assumptions the role of which is insignificant, but their sign is such that less should be expected in reality. Well, that's about as it should be. Applying filtering of inputs almost "by eye" selecting most effective ones can be increased up to 90 %, but it is already shamanism and therefore there is no desire to do it. The point of changing the TP=SL comes down to reducing the relative importance of the spread, nothing more. If at TP=50 pips and 3 pips spread, we have 47 pips for the SL and 53 pips for TP, i.e. the probability of reaching 53 and 47%, i.e. a negative 3% margin, then in case TP=10 pips and 3 pips spread we have initial negative 15% margin in favor of the SL, i.e. 65% probability of the SL and 35% probability of TP, and the system cannot show profit. It will hold on to the last of its strength at zero.
 
Dr.Brain:
What makes you think that? You can take the state and highlight trades on any one pair by hand. And at the same time, post the results of your "research" here. And I will laugh. You can also show several curves in one picture, corresponding to the results of trading on several separate pairs. By the way, I would even be interested to look at it.

So you laughed once already.

This is EURUSD

 
Dr.Brain:
The filter must "depend on the row". That is, its shape must be similar to that of the original graph. Otherwise it is not a filter. If the distribution of the first differences in the series you planted is white (particularly Gaussian) noise, it will show a probability of TP=50%. There are no miracles.

(You're excused) . Let me ask you another way. If you give you a series in which the distribution of first differences will not be white noise, the trick remains?
 
YOUNGA:

So you laughed once already. it's EURUSD

I don't understand what your picture shows at all. I see some kind of rectangular momentum. So, one trade closed on SL, then the next one on TP. So? Where to laugh or cry?
 
Nikitoss:

Let me ask you another way. If you give you a series in which the distribution of the first differences is not white noise, will the trick remain?
Depending on the nature of the series. Imagine that I would "weight the tails" of the distribution so that the probabilities of outliers (purely random) of 200-300 pips would be noticeable, that almost every tenth bar 200-300 pips. Of course there will be a purely random result in the end. In my opinion, it is obvious that what is in the input allows or does not allow for tricks.
 
Let's have a bet - the forum people below will write that without my prompting
 
YOUNGA:
Let's have a bet - the forum people below will write that without my prompting
Maybe. I'm not good at 'standard indicators'. I'm just not interested in this nonsense. And I advise you to throw it all away.
 
Dr.Brain:
Growing up slowly I see :-) Ootie, ootie, ootie :-)


Where to look ?

Where does it grow ?

72 pages of clinical fetishism

or is there an investment or monitoring ?

 
YOUNGA:
Let's have a bet - the forum people below will write that without my prompting
The clue on the chart is the name of the indy.
 
DmitriyN:
Mikhail, let's assume that you are working on the EURUSD currency pair. The information for the filter for this pair you take from GBPUSD and EURGBP, for example. Is it enough?
Or do you also need information of the EURUSD pair itself? How many pairs do you need at least for the filter to work?

That's enough. We've discussed it. There are two independent variables. If there is GBPUSD and EURGBP, then EURUSD is "unnecessary", in the sense that it is a consequence of the first two and can be calculated, the in situ measurements are simply redundant here. It's just unnecessary noise at the spread level or less. We said there are two tasks:

1. three separate graphs A, B, C=f(A,B). we need to build a non-lag filter for each of the three graphs separately. For A, using only graph a. For B, using only graph B, and for C, using only graph C.

2. the three graphs considered together A, B, C=f(A,B). you have to construct a non-delayed filter for each of the three graphs simultaneously. For A using the whole data set, for B using the whole data set, and for C using the whole data set.

I argue that these tasks are fundamentally different. Problem 1 is undefined. It has no solution. Problem 2 is well-defined. It has a solution. For the non-delayed filters under construction are linked by a coupling equation - a triangular recalculation similar to the original price graphs.

Is that what you want to understand? It's obvious. If the question is "how to do it?" then there are no comments:-)

1...616263646566676869707172737475...78
New comment