Machine Learning has incessantly changed markets

[Turveyd]

Turveyd Yes it is possible that your 10 year old computer could process few indices at once on Mt4.

However the only thing we use MT4, for is trade execution. Our Algo's run in different environments and processes a basket of instrument per tick against each timeframe, it does this to decipher trends in real time as they develop. We are currently using 5 Gig processors and this is still not enough for our operations.

While many think that ML/AI is all bull ask yourself this question, doesn't it seem that your Gmail, Google or Bing search engine is getting smarter and more accurate with word prediction? Why is that? I can assure you that its not because they are using 10 year old processors and think ML/AI is fake.

More processing power and better code to take advantage off, not ML or AI!!

 

[Pips2morePips]

More processing power and better code to take advantage off, not ML or AI!!

Turveyd Maybe I am missing something here, but is On Chip learning somehow different than ML???

 

[Turveyd]

Turveyd Maybe I am missing something here, but is On Chip learning somehow different than ML???

Chips can't Learn, Machines can't Learn, it's just marketing hype, it's merely fast enough that they claim it can do it, but in reality a slower processor can do exactly the same, just it'll take longer.

 

[Pips2morePips]

Chips can't Learn, Machines can't Learn, it's just marketing hype, it's merely fast enough that they claim it can do it, but in reality a slower processor can do exactly the same, just it'll take longer.

Turveyd So why doesn't someone charge the guys over at www.nopilot.ai for false advertising they are pioneers in methods of deep learning.. I guess we data scientists must be wrong and you right.

 

[GRULSTMRNN]

Both, faster processors or more threads (gpu) only speed up computations, nothing else (OK, they are less power consuming, too and come with sometimes useful features, AVX being one of those). What really pushed productivity in industrial applications but also daily life to new levels was the broad adoption of deep learning networks. They are not intelligent in themselves but they can be very good (when properly structured and trained) at learning complex relationships that might not be apparent to a human being. Autonomous driving, spam filter, network intrusion detection, IoT application, financial compliance and regulation, medical diagnostics and imaging, financial risk management and trading, among many others...

More processing power and better code to take advantage off, not ML or AI!!

 

[SteveH]

OP's equity curve looks like he got lucky half-way and it skewed the avg win / loss ratio.

Post the individual trades and you'll see the skull-fk-ery. It smells of Martingale.

 

[Pips2morePips]

Both, faster processors or more threads (gpu) only speed up computations, nothing else (OK, they are less power consuming, too and come with sometimes useful features, AVX being one of those). What really pushed productivity in industrial applications but also daily life to new levels was the broad adoption of deep learning networks. They are not intelligent in themselves but they can be very good (when properly structured and trained) at learning complex relationships that might not be apparent to a human being. Autonomous driving, spam filter, network intrusion detection, IoT application, financial compliance and regulation, medical diagnostics and imaging, financial risk management and trading, among many others...

GRULSTMRNN Well said.

 

[Pips2morePips]

OP's equity curve looks like he got lucky half-way and it skewed the avg win / loss ratio.

Post the individual trades and you'll see the skull-fk-ery. It smells of Martingale.

SteveH Sorry, not interested in divulging our system strategy in public forum, I take it that's why you want me to post trades..

We consider Martigale to be suicide, or a strategy for gamblers.

 

[Pips2morePips]

There will always be exploitable niches, whether from market inefficiencies or taking advantage of recurring activity patterns.

ML/AI makes some type of trades unprofitable for a human, but creates new opportunities if the human can learn to take advantage of the algo behaviour. Later new algos will replace these humans at taking advantage of the previous algos and the human traders will have to evolve again. The old algos will have to change to become profitable again.

It's a never ending cycle.

xavik Its a never ending cycle indeed. We try to keep up with new tech but we don't rebuild algos with good results, but instead keep them as is and try to improve on the newer codes. Certainly it doesn't work that way all the time. Therefore we use Java for a part of the program and R and Python for other parts. In total we have over 40k lines of code in our program. Some might say that's crazy, but our results doesn't.

 

[Van_der_Voort_4]

OP's equity curve looks like he got lucky half-way and it skewed the avg win / loss ratio.

Post the individual trades and you'll see the skull-fk-ery. It smells of Martingale.

Few who know what is part of successful strategy will not get into discussion with someone who throws topics like headlines from pages of a tabloid newspapers.

In these posted stats there is no robustness,nada,zip and OP knows it.The real theme is "i have something you don't and i may or may not let you in on it",i am on a quest to get rich and here i am sharpening my skills.