This is fantastic. I have been pouring over this for the past couple days. Thank you for your work in developing this. On my next BTC buy in will toss a donation your way.
I feel like I have a bit of learning to do in regards to getting this to work before I propose any formal questions for you regarding its operation. I will quickly mention that I installed everything with no problems to speak of but it only gets as far as displayed down below (ive let it sit for hours). The buy advice (at time of advising) was actually 50cents higher then the highest bid so obviously Im missing something but felt like composing my troubles anyways.
Thanks again. Cheers....
Thanks! Just make sure to tune it in a way that it's profitable. Else it can be a pretty expensive toy.
I am still improving the data retrieving part of the bot (when I have some free time), but I
think that the way Gekko asks data is sime time behind the websocket (the source Clark Moody is using). I'll look into this.
I agree, truly fantastic work.
Instead of having the bot pull fixed parameters from the config.js, would it be possible to adjust it so it also accepts parameters via the command line?
And would it also be possible to have the bot auto-append the profit report results to a .csv table, along with the settings used?
Something like:
node gekko --EMA1 10 --EMA2 21 --buy_threshold 0.25 --sell threshold -0.25 --fee 0.6 --output results.csv
That way you could cycle through different parameters very comfortably via a batch file and find the optimal settings.
Edit:
it would be even more comfortable, if there was an option to set upper and lower values and an increment for each parameter in the config.js
Example:
buy_threshold_upper 0.5
buy_threshold_lower 0.1
buy_threshold_increment 0.01
output results.csv
The results of each iteration could be written to a table like this:
buy_threshold sell_threshold amount_of_trades simulated_profit
0.50; -0.25; ...; 200; 1500%; ...
0.49; -0.25; ...; 210; 1510%; ...
0.48; -0.25; ...; 220; 1490%; ...
...
0.10; -0.25; ...; 200; 1200%; ...
Of course, the more variables there are, the more exponential grows the number of possible iterations, I'm aware of that.
But it would be awesome ;-)
Good point, I was thinking about this the other day and I think that I will implement a version with an increment first and log all results in a CSV. Thanks for your thoughts! I'll put it on my todo list when I have some more time.
For a long term goal I was thinking about either
particle swarm optimization or an
genetic algoritm to optimize the parameters automatically. But at that time I want to have a lot more indicators and a way to use many at the same time (maybe neural network). But this will take a lot of time.
Whats up with the gay picture?
Gekko is named after
this pretty rich guy. Profile pic is
some other guy who can fly.
