Which days of the week are good for trading - page 12
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
all knowledge is derived from the collection of statistics. Even reflexes in animals)))
There are lies, there are big lies and there are statistics.
But seriously, back in 2009 I posted a script in the codebase that counts all these days of the week. Larry Williams wrote about it in his book Long-Term Successes in Short-Term Trading. Anyone interested should go there.
In 2009 I didn't know much about statistics, so I would have done the script quite differently. In particular, if you really want to measure the effects of the day of the week, and not just know the average temperature in the room, then you need firstly to compare not the prices themselves, but their returns, and secondly you need to consider weekly trends. That is, the price fluctuations must be taken from the approximating linear function calculated for five days of the week. If regularly, let's say Friday's closing result will be lower than the trend line of the linear function, it means we can really talk about the presence of the weekday effects. In the same way we can calculate the readings for the day of the month and day of the year for example.
There are lies, there are big lies and there are statistics.
But seriously, back in 2009 I posted a script in the codebase that counts all these days of the week. Larry Williams wrote about it in his book Long-Term Successes in Short-Term Trading. Anyone interested should go there.
In 2009 I didn't know much about statistics, so I would have done the script quite differently. In particular, if you really want to measure the effects of the day of the week, and not just know the average temperature in the room, then you need firstly to compare not the prices themselves, but their returns, and secondly you need to consider weekly trends. That is, the price fluctuations must be taken from the approximating linear function calculated for five days of the week. If regularly, let's say Friday's closing result will be lower than the trend line of the linear function, it means we can really talk about the presence of the weekday effects. In the same way we can calculate the readings for the day of the month and day of the year for example.
The method isn't fancy either. The trend - as an approximating linear function of the 4 days of the week? And if there was a drop on Mon-Fri and then a rise on Thursday, what does this approximation show?
Friday's profit taking starts in the middle or end of the European session - there is a V-shaped price structure during the day. What does the approximation show?
Trend and flat is an abstraction. Just like GRAAL - everyone has their own idea
The method isn't great either. Trend - as an approximating linear function of the 4 days of the week data? And if there was a fall on Mon-Sat and then a rise on Thursday, what does this approximation show?
Friday's profit taking starts in the middle or end of the European session - during the day a V-shaped price structure. What will the approximation show?
Trend and flat is an abstraction. Just like GRAAL - everyone has their own idea
You spell it wrong. It should be like this:
Friday profit taking starts in the middle or end of the European session, just don't dare check!
Because you do not know how to check. And your methods suck! I've seen it once! See the picture in the post!
The method is not so good either. The trend - as an approximating linear function of the data of 4 days of the week? And if on Mon-Fri there was a fall, and then on Thursday there was a growth, what will this approximation show? The profit taking starts on Friday in the middle or at the end of the European session - there is a V-shaped price structure during the day. What will the approximation show?
Trend and flat is an abstraction. Just like a GRAAL - everyone has a different view
The trend is an approximation function of the 5 days of the week: Analyzed day minus 2 days, analyzed day plus 2 days. For example, analyse Wednesday:
The environment was well below the main trend (red line). Let's record the environment as "minus". If there are a sufficient number of such low environments, their resultant point will be much lower than the resultant line of the general trend. If, on the contrary, the environments will generally be above the average trend line, their midpoint will be above the average trend line. And only if half of the environments are above and the other half below the mean trend line, then their total resultant point will be zero or so, i.e. their total result will be close to the general trend line:
As for the intraday Friday formations, of course there is no way to approximate them, but even in this case the problem is solved.
The trend is an approximating function of the data of 5 days of the week: the analyzed day minus 2 days, the analyzed day plus 2 days. For example, analyse Wednesday:
The environment was well below the main trend (red line). Record the environment as "minus". If there is a sufficient number of such low environments, their resultant point will be well below the resultant general trend line. If, on the contrary, the environments will generally be above the average trend line, their midpoint will be above the average trend line. And only if half of the Wednesdays are above and the other half below the mean trend line, then their total resultant point will be zero or so, i.e. their total result will be close to the general trend line:
As for the intraday Friday formations, of course there is no way to approximate them, but even in this case the problem is solved.
All problems are solvable, the only question is how to solve them.
For example, your first picture shows a descending trend in Mon-Fri, in Thursday - fixation of profit. On Friday - continuation of the trend, as the profit has already been fixed. Although, formally, the 4th day is down trend and therefore there is a high probability of profit taking on Friday.
Globally, it doesn't make sense to determine the trend this way. Why? On the first picture - the first three days are trending. Thursday is the change of the trend. And the model will still show its continuation.
The trend is what? Is it defined on the hourly, on the minute data? On the daily, monthly data? If the trend is an approximating line, then a flat does not exist? It is very difficult to find the interval, where the approximating function is strictly horizontal.
The point is that we don't accidentally take an approximation with a look ahead 2 days. That is, we already know the trend that will be tomorrow and after tomorrow and it is already taken into account in our today's approximation function, which means that it is also taken into account in today's movement.
We should also distinguish the difference between the tasks: to find the effects of profit taking and to find the effect of the day of the week. These are different tasks and both are solved differently, I only suggest a solution for the second of them (to determine the probability and strength of the day's close in one direction or the other).
As for the first figure, you cannot tell anything about any day of the week from it. Perhaps it is just a combination of random numbers. However, for example, if at least one of the ten Thursdays tends to be higher, it does not matter if it closes above or below the approaching function, its result will still raise the average result of all Thursdays above zero (the average minus will be lower, and the average plus will be higher). Indeed, if out of a thousand coin tosses, 50 are not random and are always positive, then 25 will turn positive, which will cause the result to be 475 versus 525.
When all values are counted, the approximating line is zero or neutral (that's why it's parallel to the floor in the second figure). If any point is higher or lower than it, this point will be exactly what we are looking for - the effect of Thursday or Friday, because it only possible in case of non-accidental closure of any day of the week (in fact the average result will always differ from the zero line, but the limiting deviation, when we can talk about its non-accidental can be calculated).
It is also important to consider the possibility of alternation of days of the week: Wednesday - up, then Thursday - up, etc. If this alternation takes place, and most likely it is, then simply collecting all days in one pile will not give anything - we just get 50/50. But this alternation can be accounted for by applying a sliding calculation window.