A great exercise for someone learning genetic algorithms. Your right with the assertion that you need to expand the available information to the patterns.
Really cool The diminishing performance year by year is quite intriguing indeed, makes you imagine how much stuff has already been tested by big institutions...
It could be relaly interesting if you try another video but this time wirh some specific candle pattern (doji, morning star, hammer, etc...) and show what's their probabilities. It could be compmeted with some other parameters (resitance / support zone, etc...)
Can you release a course for all of these methods? Like building strategy, properly backtesting it and finding if it works datascience way? From 0 to trading bot
Hey man, I can't believe we made the exact same experiment, I have run the exact same setup using G.A. 4 years ago to mine candle patterns! What are the odds!
They did * Fixed genome size * Equity curve correlation * Walk forward optimization They could have * Varied genome size * Showed best score per population graph * Tried different trading strategies (limit/stop exits, trailing stops) * Not ignored fees * Tried hyperparameter optimization (with various walk forward long and short periods, different candle sizes) * Compared with the performance of the underlying instrument there is so much more *you* can do with your brain than what you see in this video.
I'm curios what's the reason you didn't share the code like other videos, is it because sharing these information publicly is not ideal which results in disorder and turbulence in markets or the reason is something else?
I implemented this with C++ and its around 1500 lines of messy code. I didn't want to explain it all, or receive tons of questions about how to use it.
Great work on using a genetic algorithm to find optimal candlestick patterns for Bitcoin! Keep up the amazing work!
How’s your game coming along brother
@@MidnightV6 paused for now im afraid 😧 other stuff came in the way.
A great exercise for someone learning genetic algorithms. Your right with the assertion that you need to expand the available information to the patterns.
this is gold! keep pumping these amazing videos!
Really cool
The diminishing performance year by year is quite intriguing indeed, makes you imagine how much stuff has already been tested by big institutions...
Wow, brilliant approach.
Thanks for sharing!
great explanation, thanks bro
It could be relaly interesting if you try another video but this time wirh some specific candle pattern (doji, morning star, hammer, etc...) and show what's their probabilities.
It could be compmeted with some other parameters (resitance / support zone, etc...)
Can you release a course for all of these methods? Like building strategy, properly backtesting it and finding if it works datascience way? From 0 to trading bot
keep going
Hey man, I can't believe we made the exact same experiment, I have run the exact same setup using G.A. 4 years ago to mine candle patterns! What are the odds!
why not doing exhaustive optimization?
What did you study to understand all these
Concho al menos comparte el codigo para analizarlo y ver si podemos usarlo hacerle mejroas o tomar porciones para otros ejemplos, super bueno.
How stable in time are these rules?
such kind of expsriments shoul be proved with cross validation
He just did cross validation by testing walk forward test
Would appreciate if you share your code regardless of messiness
please share the code, i want to explore it more
Super
Patterns getting arbed out of the market is speculation I subscribe to. Chasing good fits did well for me god it was fun
My only question is, what did you study to get this good? How can you make these out of your own brain?
Read books and practice
They did
* Fixed genome size
* Equity curve correlation
* Walk forward optimization
They could have
* Varied genome size
* Showed best score per population graph
* Tried different trading strategies (limit/stop exits, trailing stops)
* Not ignored fees
* Tried hyperparameter optimization (with various walk forward long and short periods, different candle sizes)
* Compared with the performance of the underlying instrument
there is so much more *you* can do with your brain than what you see in this video.
I'm curios what's the reason you didn't share the code like other videos, is it because sharing these information publicly is not ideal which results in disorder and turbulence in markets or the reason is something else?
I implemented this with C++ and its around 1500 lines of messy code. I didn't want to explain it all, or receive tons of questions about how to use it.
@@neurotrader888 ok thanks for the reply
@@neurotrader888 😂 great answer
Good
I think you need to run a simulation with an entry price, stop, and target. Simply saying it went up or down doesn’t pass any kind of real test.
bro ....
whats the point of posting video and you dont want with the source code