Machine learning in trading: theory, models, practice and algo-trading - page 3740

New comment
[Deleted]  

I made ML work on low TFs (here 5-min EURUSD), forward from 2025. It took me a long time because of low learning speeds. I found a workaround :)

I still need to switch to ticks somehow, as Saber willed. The puzzle is serious.

File who wants to test. I have already started holidays, not before :) For the same reason I may have made a mistake somewhere.

Statistics of errors by hours, in the bot the best model (hour 0)

Iteration: 0, Cluster: 0
R2: 0.9678626340659742
 Iteration: 0, Cluster: 1
R2: -0.5447192261946558
Iteration: 0, Cluster: 2
R2: -0.7413142233465053
Iteration: 0, Cluster: 3
R2: -0.7660350578598544
Iteration: 0, Cluster: 4
R2: -0.7393956301369173
Iteration: 0, Cluster: 5
R2: -0.81466783528003
Iteration: 0, Cluster: 6
R2: -0.5328809526022955
Iteration: 0, Cluster: 7
R2: -0.4819371579645384
Iteration: 0, Cluster: 8
R2: -0.7194667685923652
Iteration: 0, Cluster: 9
R2: 0.036910760744729876
Iteration: 0, Cluster: 10
R2: -0.7410474652199799
Iteration: 0, Cluster: 11
R2: -0.23225385299559764
Iteration: 0, Cluster: 12
R2: -0.057876213851149605
Iteration: 0, Cluster: 13
R2: -0.11052210424152764
Iteration: 0, Cluster: 14
R2: 0.6999765381754814
Iteration: 0, Cluster: 15
R2: -0.041169637019330185
Iteration: 0, Cluster: 16
R2: 0.36503834380802713
Iteration: 0, Cluster: 17
R2: 0.33122571568647363
Iteration: 0, Cluster: 18
R2: 0.4300055781420823
Iteration: 0, Cluster: 19
R2: 0.6345313486858764
Iteration: 0, Cluster: 20
R2: 0.049646596894408423
Iteration: 0, Cluster: 21
R2: -0.42441372859884796
Iteration: 0, Cluster: 22
R2: -0.7335582640680249
Iteration: 0, Cluster: 23
R2: -0.9366929805616558
 
Maxim Dmitrievsky #:

I made ML work on low TFs (here 5-min EURUSD), forward from 2025. It took me a long time because of low learning speeds. I found a workaround :)

I still need to switch to ticks somehow, as Saber willed. The puzzle is serious.


in the tester 5% for 3 years - a serious breakthrough in ML, I did not suffer in vain ;-)

[Deleted]  
Maxim Kuznetsov #:

in tester 5% in 3 years - serious breakthrough in ML, didn't agonise for nothing ;-)

The fan-fluffers and clueless ones are too lazy to reply already
 
Maxim Dmitrievsky #:

in bot the best model (hour 0)


Night Scalper

No commissions on the account :)


I don't even know where to put this at the end of the year: in the failures of the year or in the joke of the year
[Deleted]  

[Deleted]  
ZY if someone has other good statistics on eurobucks that this model cannot detect - please let me know.
 
Maxim Dmitrievsky #:

Accounting for the real spread at the 0th hour would ruin this fairy tale)
[Deleted]  
secret #:
Taking into account the real spread at the 0th hour will ruin this fairy tale).
There with reserve, the spread is real (almost). On some other rates it really +- kills, but on Eurobucks it is less. You can't put it into the model, only then look for good trading conditions.
 

Maxim Dmitrievsky #:
On Eurobucks it's less
look for good trading conditions. 

I've already written up a product description

 

Started on Ranna 2025 on real ticks.

And looked in detail through TesterReport. It is hard there at the rollover hour, but I have never traded EURUSD at this hour, so I do not know how the execution. Of course, only limit orders should be tested there.

1...373337343735373637373738373937403741
New comment