freqtrade_origin/docs/data-analysis.md
2019-09-14 09:54:40 +02:00

7.1 KiB

Analyzing bot data with Jupyter notebooks

You can analyze the results of backtests and trading history easily using Jupyter notebooks. Sample notebooks are located at user_data/notebooks/.

Pro tips

  • See jupyter.org for usage instructions.
  • Don't forget to start a Jupyter notebook server from within your conda or venv environment or use nb_conda_kernels*
  • Copy the example notebook before use so your changes don't get clobbered with the next freqtrade update.

Fine print

Some tasks don't work especially well in notebooks. For example, anything using asynchronous execution is a problem for Jupyter. Also, freqtrade's primary entry point is the shell cli, so using pure python in a notebook bypasses arguments that provide required objects and parameters to helper functions. You may need to set those values or create expected objects manually.

Task Tool
Bot operations CLI
Repetitive tasks Shell scripts
Data analysis & visualization Notebook
  1. Use the CLI to

    • download historical data
    • run a backtest
    • run with real-time data
    • export results
  2. Collect these actions in shell scripts

    • save complicated commands with arguments
    • execute multi-step operations
    • automate testing strategies and preparing data for analysis
  3. Use a notebook to

    • visualize data
    • munge and plot to generate insights

Example utility snippets

Change directory to root

Jupyter notebooks execute from the notebook directory. The following snippet searches for the project root, so relative paths remain consistent.

import os
from pathlib import Path

# Change directory
# Modify this cell to insure that the output shows the correct path.
# Define all paths relative to the project root shown in the cell output
project_root = "somedir/freqtrade"
i=0
try:
    os.chdirdir(project_root)
    assert Path('LICENSE').is_file()
except:
    while i<4 and (not Path('LICENSE').is_file()):
        os.chdir(Path(Path.cwd(), '../'))
        i+=1
    project_root = Path.cwd()
print(Path.cwd())

Load existing objects into a Jupyter notebook

These examples assume that you have already generated data using the cli. They will allow you to drill deeper into your results, and perform analysis which otherwise would make the output very difficult to digest due to information overload.

Load backtest results into a pandas dataframe

from freqtrade.data.btanalysis import load_backtest_data

# Load backtest results
df = load_backtest_data("user_data/backtest_results/backtest-result.json")

# Show value-counts per pair
df.groupby("pair")["sell_reason"].value_counts()

Load live trading results into a pandas dataframe

from freqtrade.data.btanalysis import load_trades_from_db

# Fetch trades from database
df = load_trades_from_db("sqlite:///tradesv3.sqlite")

# Display results
df.groupby("pair")["sell_reason"].value_counts()

Load multiple configuration files

This option can be useful to inspect the results of passing in multiple configs. This will also run through the whole Configuration initialization, so the configuration is completely initialized to be passed to other methods.

import json
from freqtrade.configuration import Configuration

# Load config from multiple files
config = Configuration.from_files(["config1.json", "config2.json"])

# Show the config in memory
print(json.dumps(config, indent=2))

For Interactive environments, have an additional configuration specifying user_data_dir and pass this in last, so you don't have to change directories while running the bot. Best avoid relative paths, since this starts at the storage location of the jupyter notebook, unless the directory is changed.

{
    "user_data_dir": "~/.freqtrade/"
}

Load exchange data to a pandas dataframe

This loads candle data to a dataframe

from pathlib import Path
from freqtrade.data.history import load_pair_history

# Load data using values passed to function
ticker_interval = "5m"
data_location = Path('user_data', 'data', 'bitrex')
pair = "BTC_USDT"
candles = load_pair_history(datadir=data_location,
                            ticker_interval=ticker_interval,
                            pair=pair)

# Confirm success
print(f"Loaded len(candles) rows of data for {pair} from {data_location}")
candles.head()

Strategy debugging example

Debugging a strategy can be time-consuming. FreqTrade offers helper functions to visualize raw data.

Define variables used in analyses

You can override strategy settings as demonstrated below.

# Customize these according to your needs.

# Define some constants
ticker_interval = "5m"
# Name of the strategy class
strategy_name = 'SampleStrategy'
# Path to user data
user_data_dir = 'user_data'
# Location of the strategy
strategy_location = Path(user_data_dir, 'strategies')
# Location of the data
data_location = Path(user_data_dir, 'data', 'binance')
# Pair to analyze - Only use one pair here
pair = "BTC_USDT"

Load exchange data

from pathlib import Path
from freqtrade.data.history import load_pair_history

# Load data using values set above
candles = load_pair_history(datadir=data_location,
                            ticker_interval=ticker_interval,
                            pair=pair)

# Confirm success
print(f"Loaded {len(candles)} rows of data for {pair} from {data_location}")
candles.head()

Load and run strategy

  • Rerun each time the strategy file is changed
from freqtrade.resolvers import StrategyResolver

# Load strategy using values set above
strategy = StrategyResolver({'strategy': strategy_name,
                            'user_data_dir': user_data_dir,
                            'strategy_path': strategy_location}).strategy

# Generate buy/sell signals using strategy
df = strategy.analyze_ticker(candles, {'pair': pair})

Display the trade details

  • Note that using data.tail() is preferable to data.head() as most indicators have some "startup" data at the top of the dataframe.
  • Some possible problems
    • Columns with NaN values at the end of the dataframe
    • Columns used in crossed*() functions with completely different units
  • Comparison with full backtest
    • having 200 buy signals as output for one pair from analyze_ticker() does not necessarily mean that 200 trades will be made during backtesting.
    • Assuming you use only one condition such as, df['rsi'] < 30 as buy condition, this will generate multiple "buy" signals for each pair in sequence (until rsi returns > 29). The bot will only buy on the first of these signals (and also only if a trade-slot ("max_open_trades") is still available), or on one of the middle signals, as soon as a "slot" becomes available.
# Report results
print(f"Generated {df['buy'].sum()} buy signals")
data = df.set_index('date', drop=True)
data.tail()

Feel free to submit an issue or Pull Request enhancing this document if you would like to share ideas on how to best analyze the data.