Where is the line between fitting and actual patterns? - page 43

 
A good system will remain profitable both with a SL/TP exit and with a fixed position life. And with a combination of both.
 
IgorM:


...yat!

with all these searches you can't remember what you've found and why else you need to look - sorry, but then for the point: the forum is evil!

)))

SZS: Artem, thanks for some intellectual stimulus, now I have no time to sing your praises (at home early traditional vanity))) ), but in order not to forget, I've drawn your halo on your avatar with a marker! )))))))))))))))))))))))

Igor, stop it... It looks like a good regularity or a trading method ( ???) has been found, that allows to draw the lines of a chart stretching upwards in the past 10 years. А?

ZS. Wipe all the squiggles off the monitor... :) I had nothing to do with it. She's just a thinker... :) Are you a master of Photoshop? :)))))))

 
It is better to divide systems by methods: trend following, pattern following, swing trading, spread trading, etc. The necessary limitations on the time in a position, position management, ways to formalize entry and exit are determined by the type of system, and in fact by the type of the underlying regularity. So what makes sense to adjust/optimize for the appropriate method and what doesn't - is also characteristic of each method.
 
Avals:
It is better to divide systems by methods: trend following, pattern following, swing trading, spread trading, etc. The necessary time limits on positions, position holding, and methods of entering and exiting are determined by the type of system, and in fact by the type of underlying regularity. So what makes sense to adjust/optimize for the appropriate method and what doesn't - is also characteristic of each method.

Can you make a comparative table of TCs according to your typing to make it clear which are susceptible to fitting and which are not? It is possible without explanations.

The same question is addressed to MetaDriver.

 
joo:

Can you make a comparative table of TCs according to your typing to make it clear which are susceptible to fitting and which are not? It is possible without explanations.

The same question is addressed to MetaDriver.


they can all be fitted. For each type of system, there are different dangers of fitting and what can be optimised relatively harmlessly. Writing a lot is lazy and not enough will not do :) Each type of system and discussion of what it should contain, what and how to look for/optimize is the subject of a decent branch. Take for example the "Let's talk about patterns" thread on the spider.
 
There must be no dedicated parameters in the system for there to be no fitting. Any system which does (say, a waving with periods of 8 and 33: why 8, why 33?!) is subject to fitting.
 
Mathemat:
There must be no dedicated parameters in the system for there to be no fitting. Any system which does (say, a waving with periods of 8 and 33: why 8, why 33?!) is subject to fitting.

for H1 at round-the-clock quotes are understandable numbers
 
artmedia70:

Igor, stop it... Really, it seems that we have found a good regularity or a trading method ( ???) that allows us to draw evenly upwards the lines of the chart in the past 10 years. А?

I have found the pattern and it is not going anywhere - it was a cross between))) trend advisor and a locker based on Avalanche, I upgraded Avalanche a bit, but it did not lose and in fact the code in the tester always worked in profit due to the fact that I optimized risks - I limited the output of the locker by the number of flips, i.e. In other words, during the third real turn the target was not breakeven, and on reaching a small loss I closed all the orders - both the first ones from the trend part of the code and from Avalanche itself. The system of flips itself was sketched in general terms in topic Avalanche and I left the research because the main trend strategy proved to be completely inefficient even on a demo account - in a week of trading what had been earned would have been lost without the locker. Now the problem is to find effective basic strategy, as discussed in this thread - without parameters that need optimization, but when developing your strategy first need to test it by hand trading, so that then be able to formalize the problem for programming. I think that the problem with many people is that "playing" with the strategy tester takes all the time and there is no time to trade or even to watch the charts online.

 
Mathemat:
There must be no dedicated parameters in the system for there to be no fitting. Any system which does (say, a waving with periods of 8 and 33: why 8, why 33?!) is subject to fitting.

For example, a 120 hour waveform. By the number of hours in a week.

And why would the presence of this wristwatch suddenly mean that it is subject to fitting? In a week, do the hours change from time to time?

 
paukas:

In a week, do the hours change at times?

but the starting point of MA120 will change every hour, to be logical, draw a trend line, first point 0 on Mon, second point 0 on Sat - imho this line will give you much more information to think about than a moving average;)
Reason: