search

nateemma / strategies

Custom trading strategies using the freqtrade framework
318 stars 85 forks source link

Signal Cluster #33

Open just-nilux opened 1 year ago

just-nilux commented 1 year ago

Hey, have you thought about using all the TradingSignals and then finding clusters of signals to build some kind of consensus for buys and sells? And eventually then also use the DWT signals as confirmation. Just an idea! Whats your take?

tomjrtsmith commented 1 year ago

Haven't done too much coding lately but had much the same thought when I saw TradingSignals.py.

On Wed, Jul 5, 2023 at 7:32 AM nilux @.***> wrote:

Hey, have you thought about using all the TradingSignals and then finding clusters of signals to build some kind of consensus for buys and sells? And eventually then also use the DWT signals as confirmation. Just an idea! Whats your take?

— Reply to this email directly, view it on GitHub https://github.com/nateemma/strategies/issues/33, or unsubscribe https://github.com/notifications/unsubscribe-auth/AXXUNNOR6DGNRCAS4GG3O7DXORVU3ANCNFSM6AAAAAAZ6CRPRY . You are receiving this because you are subscribed to this thread.Message ID: @.***>

just-nilux commented 1 year ago

the function you have to find the best classifier could actually be base for that. Just a thought!

nateemma commented 1 year ago

I'm not quite sure what you mean.

The signals in TrainingSignals.py are all forward-looking, so I can't use them in the models themselves, I can only use them to train the models. Did you mean that I could combine various signals to create new 'compound' signals for training (e.g. combine both adx and fbb signals)?

Thanks,

Phil

On Tue, Jul 4, 2023 at 2:26 PM nilux @.***> wrote:

the function you have to find the best classifier could actually be base for that. Just a thought!

— Reply to this email directly, view it on GitHub https://github.com/nateemma/strategies/issues/33#issuecomment-1620758774, or unsubscribe https://github.com/notifications/unsubscribe-auth/ABD4X56BTZ7RBE4GJ7I546TXOSDADANCNFSM6AAAAAAZ6CRPRY . You are receiving this because you are subscribed to this thread.Message ID: @.***>

just-nilux commented 1 year ago

@nateemma

you mean that I could combine various signals to create new 'compound'

yes in a way. in a way like feature engineering, just in this case label engineering / selection. actually it boils down to that creating these sets and files manually like now is quite bad to maintain, if you want to make a lot of experiments its not practical and the best sets you may never discover.

are you using any pipeline to track the results, drift etc? optuna with a pipeline to track results would be a perfect solution here I think!