BackTest with neuralnet

[d-pires]

Hey great software btw, Trying to work with https://github.com/zschro/gekko-neuralnet but without success, could you help me a bit ?

~/gekkojap/japonicus$ python3.6 japonicus.py -g --strat neuralnet GEKKO ██╗ █████╗ ██████╗ ██████╗ ███╗ ██╗██╗ ██████╗██╗ ██╗███████╗ ██║██╔══██╗██╔══██╗██╔═══██╗████╗ ██║██║██╔════╝██║ ██║██╔════╝ ██║███████║██████╔╝██║ ██║██╔██╗ ██║██║██║ ██║ ██║███████╗ ██ ██║██╔══██║██╔═══╝ ██║ ██║██║╚██╗██║██║██║ ██║ ██║╚════██║ ╚█████╔╝██║ ██║██║ ╚██████╔╝██║ ╚████║██║╚██████╗╚██████╔╝███████║ ╚════╝ ╚═╝ ╚═╝╚═╝ ╚═════╝ ╚═╝ ╚═══╝╚═╝ ╚═════╝ ╚═════╝ ╚══════╝ v0.70

The profits reported here depends on backtest interpreter function; interpreter v3: if > 0: = - else = japonicus.py -g --strat neuralnet Evolving neuralnet strategy;

evaluated parameters ranges: threshold_buy [0.1, 3.5]

threshold_sell [-3.5, -0.1]

learning_rate [0.01, 0.25]

decay [0.01, 0.1]

hodl_threshold [0.8, 1.0]

price_buffer_len [30, 300]

evolution candlestick dataset 2018-01-01 00:42:00 to 2018-05-29 02:42:00 CMT/ETH @binance

evolution candlestick dataset 2018-01-01 00:00:00 to 2018-05-30 02:21:00 ADX/ETH @binance

evaluation candlestick dataset 2018-01-01 00:00:00 to 2018-05-29 01:34:00 ADA/ETH @binance

======  EPOCH 0/800  ======

Locale1 first unevaluated: 30 0 individues removed due to equality [30] EPOCH 0 &30 Maximum profit -0.038 Average profit -0.038
Minimum profit -0.038 Profit variation 0
Population size 30 Max population size 30
Avg trade number 0 Avg sharpe ratio 0
Avg exposure time 0

my strategy_parameters/neuralnet.toml

threshold_buy = [0.10, 3.50] threshold_sell = [-3.50, -0.10] learning_rate = [0.01, 0.25] decay = [0.01, 0.10] hodl_threshold = [0.80 , 1.00] price_buffer_len = [30, 300]

my configStrategies.py ... "neuralnet": { "threshold_buy": (0.10, 3.50), "threshold_sell": (-3.50, -0.10), "learning_rate": (0.01, 0.25), "decay": (0.01, 0.10), "hodl_threshold": (0.80 , 1.00), "price_buffer_len": (30, 300) }

and my gekko config

//neuralnet settings config.neuralnet = { "threshold_buy": 0.125, "threshold_sell": -0.125, "method": "adadelta", "learning_rate": 0.01, "momentum": 0.1, "decay": 0.01, "stoploss_enabled": true, "stoploss_threshold": 0.85, "hodl_threshold": 1, "price_buffer_len": 100, "min_predictions": 1000 };

in gekko backtest that config works fine, so I assume its some configuration error I made in japonicus, it's not doing any trades. And how can I select only 1 Local to backtest ? like only ETH-TRX

thanks!

[d-pires]

With NNv2 no issues running will close thread.

[CryptoCoeus]

Hi, @desps-iscteiulpt is there any success with NNv2 on the real market ? Why you choose this strat to run ?