Some questions about training and optimization

[maledicente]

I always use docker. I created an image with the requirements in the Dockerfile, it works well. Can I train TS_Gain normally or is there a specific configuration in zsh?

back_b_ia: image: freqtrade_ia restart: on-failure container_name: back_b_ia volumes:

• "./user_data:/freqtrade/user_data" command: > backtesting --datadir user_data/data/${EXCHANGE:-binance} -c user_data/config-exchange-general.json -c /freqtrade/user_data/bot_binance_ia.json -c user_data/pair_lists/api_binance.json -c user_data/pair_lists/blacklist-binance.json -c user_data/pair_lists/api_binance.json -c user_data/config-telegram-general.json --max-open-trades ${MAX_OPEN_TRADES:-6} --strategy-list TSPredict TS_Gain -c user_data/pair_lists/pairlist-backtest-static-binance-spot-usdt.json --data-format-ohlcv feather --timeframe 5m --dry-run-wallet 100 --timerange ${TIMERANGE:-20220101-20231207} CPU: '48' mem_limit: 500G memswap_limit: -1

  I have tried to use your script in a ubuntu machine, but zsh: can't open input file: user_data/strategies/scripts/test_strat freqtrade_local/freqtrade$ zsh user_data/strategies/scripts/test_strat -n 600 TSPredict TS_Gain

[nateemma]

Oops, the instructions are slightly wrong - you need to add .sh to test_strat:

zsh user_data/strategies/scripts/test_strat.sh -n 600 TSPredict TS_Gain

If you haven't changed the indicators, you (usually) don't need to re-generate the model

Thanks,

Phil

On Thu, Dec 7, 2023 at 3:52 PM Luiz Paulo Nievola @.***> wrote:

I always use docker. I created an image with the requirements in the Dockerfile, it works well. Can I train TS_Gain normally or is there a specific configuration in zsh?

back_b_ia: image: freqtrade_ia restart: on-failure container_name: back_b_ia volumes:

• "./user_data:/freqtrade/user_data" command: > backtesting --datadir user_data/data/${EXCHANGE:-binance} -c user_data/config-exchange-general.json -c /freqtrade/user_data/bot_binance_ia.json -c user_data/pair_lists/api_binance.json -c user_data/pair_lists/blacklist-binance.json -c user_data/pair_lists/api_binance.json -c user_data/config-telegram-general.json --max-open-trades ${MAX_OPEN_TRADES:-6} --strategy-list TSPredict TS_Gain -c user_data/pair_lists/pairlist-backtest-static-binance-spot-usdt.json --data-format-ohlcv feather --timeframe 5m --dry-run-wallet 100 --timerange ${TIMERANGE:-20220101-20231207} CPU: '48' mem_limit: 500G memswap_limit: -1

I have tried to use your script in a ubuntu machine, but zsh: can't open input file: user_data/strategies/scripts/test_strat freqtrade_local/freqtrade$ zsh user_data/strategies/scripts/test_strat -n 600 TSPredict TS_Gain

— Reply to this email directly, view it on GitHub https://github.com/nateemma/strategies/issues/41, or unsubscribe https://github.com/notifications/unsubscribe-auth/ABD4X52GWOGGGDZOQJ275PLYIJJETAVCNFSM6AAAAABAL3RSKWVHI2DSMVQWIX3LMV43ASLTON2WKOZSGAZTCNRZHEYDGOA . You are receiving this because you are subscribed to this thread.Message ID: @.***>

[maledicente]

Okay, here -n 600 TS Predict TS Gain -n = 600 days? In Docker, can I directly train Ts_Gain with the commands above in back_b_ia? I'm slowly studying your algorithms, it's quite dense to understand hehe

[nateemma]

I don't quite understand your question. What is back_b_ia?

Anyway, not sure if this will help:

• the command: zsh user_data/strategies/scripts/test_strat.sh -n 600 TSPredict TS_Gain is running the script test_strat.sh over the last 600 days, and is running the strategy TS_Gain.py in directory TSPredict

• if a model file already exists (in user_data/strategies/TSPredict/models/TS_Gain) then it will load that and use it as the basis for making predictions. If it does not exist, then it will train a new model and save it - in this case it will not make any predictions. I usually only run for long time periods when I am creating a new model. I would normally only test for about 30 days (or less)

• TSPredict.py is the base class of all of the strategies in this directory, and provides the main framework for loading/saving models, backtesting, populating buy/sell signals and handling custom stoploss/exit functionality.

• There are several 'families' of strategies: TS_Gain, TS_Simple, TS_Coeff and TS_Wavelet right now. Each of these overrides functions of TSPredict to implement a type of approach. For example, TS_Gain just predicts the future gain, it doesn't use any other indicators.

• there are variants within each family that might use different prediction algorithms (as in TS_Gain and TS_Simple), or different ways of modelling the signal (as in TS_Coeff and TS_Wavelet)

Hope that helps. Feel free to ask more questions, I know it's pretty complicated

Thanks,

Phil

On Thu, Dec 7, 2023 at 5:08 PM Luiz Paulo Nievola @.***> wrote:

Okay, here -n 600 TS Predict TS Gain -n = 600 days? In Docker, can I directly train Ts_Gain with the commands above in back_b_ia? I'm slowly studying your algorithms, it's quite dense to understand hehe

— Reply to this email directly, view it on GitHub https://github.com/nateemma/strategies/issues/41#issuecomment-1846343627, or unsubscribe https://github.com/notifications/unsubscribe-auth/ABD4X5YX7WRSTTKN7OWEOMLYIJSBNAVCNFSM6AAAAABAL3RSKWVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTQNBWGM2DGNRSG4 . You are receiving this because you commented.Message ID: @.***>