Live Trading Stops After Few Hours

[sirtificate]

Live Trading Stops After Few Hours

[rud3ls87]

Hello, just checking if this issue was resolved or if there is a potential fix/work around? :) my bot stopped trading after 12hours or so, was working great up till then.

[SirTificate]

As I already told you on discord, I suspect that the onTrade() function fails at some point and the bot gets stuck, because the state of this.prevAction is wrong. You could add a debug function which prints this.prevAction every 100 candles or so to see, if there's something wrong.

[questionguy123]

Were you able to find a fix?

[questionguy123]

Maybe something to do with Momentum or Decay settings messing with the learning rate as time goes on? Going to test this now

[rud3ls87]

Hey! No didnt figure it out just yet. Okay awesome, yes please let me know if you are able to figure it out :) at the moment im just stopping and restarting the gekko every 10 hours or so. Regards Rudy

…

On Sat, 04 Aug 2018, 15:27 questionguy123, ***@***.***> wrote: Maybe something to do with Momentum or Decay settings messing with the learning rate as time goes on? Going to test this now — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#4 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AnctHyYr8vVjRVESQMMLsd5doB2J1AhRks5uNaFTgaJpZM4S9ZZ5> .

[questionguy123]

According to the 30min of youtube videos i watched, decay will gradually effect the learning rate over time.

On the backtest:
Changing the learning rate in the backtest effects the outcome
Changing the decay in the backtest has no effect on the outcome

therefore it doesn't make sense how decay can change the learning rate without changing the outcome? Maybe on the live bot decay reduces the learning rate lower and lower until the bot no longer wants to make trades.

30min of youtube doesn't make me an expert but going to try the bot with no decay and will post what happens

[rud3ls87]

Fantastic! Thank you for doing the research! Much appreciated. If you find that the 0 decay works then I will be over the moon :P was about to write python script to restart the bot every x hours, lol. Regards Rudy

…

On Sat, 04 Aug 2018, 16:10 questionguy123, ***@***.***> wrote: According to the 30min of youtube videos i watched, decay will gradually effect the learning rate over time. On the backtest: Changing the learning rate in the backtest effects the outcome Changing the decay in the backtest has no effect on the outcome therefore it doesn't make sense how decay can change the learning rate without changing the outcome? Maybe on the live bot decay reduces the learning rate lower and lower until the bot no longer wants to make trades. 30min of youtube doesn't make me an expert but going to try the bot with no decay and will post what happens — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#4 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AnctH81gfjtHrsVXTvCOhx5S7EXNIQnEks5uNatFgaJpZM4S9ZZ5> .

[g1tt]

I know this is closed but I was experiencing the same problems.
SirTificate's comments on Mar 29th were exactly the cause of my problem.
I have fixed it (a very low-tech fix but it works!) and have created a pull request here -
#11