CHANGELOG.md

current

• Add: resample_dataframe utility function. Useful for resampling dataframes with multiple columns that are not candlestick data

0.22.14

• Fix: Fix for inconcistencies in lending data downloads

0.22.13

• Update: Having pd.MultiIndex instead of per-pair pd.DateTimeIndex on candles slows down price access on backtesting considerably
• Update: Binance progress bar now shows individual progress for each pair when there are less than 5 pairs

0.22.12

• Update fetch_all_spot_symbols API call to Binance
• Add: LendingReserveUniverse.limit_to_protocol()
• Fix: If there is an API key error when saving exchange universe, abort the operation with an exception

0.22.11

• Add support for Aave v2 dataset

0.22.10

• Change pyarrow schemas of Candle and XYLiquidityto use double precision (float64) instead of float32
• Add GroupedCandleUniverse(forward_fill) option to the constructor
• Add PandasPairUniverse.iterate_tokens()
• Add get_prior_timestamp(series: pd.Series, ts: pd.Timestamp) as a separate utility function
• Add resample_candles(shift) argument
• Add format_price(decimals) argument
• Add Universe.get_default_chain() method
• Add resample_price_series() function
• Fix Binance data using local timezone
• Fix Binance multipair data resampling not working correctly

0.22.9

• Add remove_zero_candles() function
• Add DEXPair.exchange_name is set if PandasPairUniverse(exchange_universe) is given
• Add ChainId.centralised_exchange = -1 id that is used now for Binance loaded data

0.22.8

• Add fee argument to generate_pairs_for_binance()

0.22.7

• Fix for Binance lending data, change daily rates to annual

0.22.6

• Add PairNotFound(description) to have better error messages on pair look-up failures
• Stricter input type checks for exchanges when creating Universe
• Add UniswapV2MockClient.clear_caches for compatibility with Client in unit testing

0.22.5

• Add PandasPairUniverse.get_token_by_symbol()
• Add PandasPairUniverse.get_token(chain_id) argument

0.22.4

• Update error message for Binance symbols

0.22.3

• Add more Binance data sideloading related utility functions to tradingstrategy.binance.utils

0.22.2

Updates to BinanceDownloader

• Add BinanceDownloader.fetch_approx_asset_trading_start_date() to figure out when asset was listed on Binance
• Change BinanceDownloader.fetch_candlestick_data() and BinanceDownloader.fetch_lending_rates() to accept multiple pairs at once
• Add progress bars
• Add symbol validation

0.22.1

• Fix get_by_chain_and_factory() was returning ExchangeNotFoundError instead of raising it

0.22

Binance data downloading added. Python 3.11, Python 3.12 and Pandas 2.x compatibility fixes.

• Add: BinanceDownloader.fetch_approx_asset_trading_start_date() to figure out when asset was listed on Binance
• Fix: BinanceDownloader timestamps to be UTC
• Add: Example script how to download Binance data for multiple assets and write in a single Parquet file
• Upgrade to support Python 3.12 with breaking datetime.datetime.utcnow() changes
• Upgrade to support Pandas 2.x with breaking pd.Timestamp.utcnow() changes
• Upgrade to web3-ethereum-defi latest with Python 3.12 compatibilty
• Upgrade to pyarrow 14.0 for Python 3.12 compatibility
• datetime.datetime.utcnow() and datetime.datetime.utcfromtimestamp workarounds - we do not need timezones that unnecessarily bloat data size
• Add naive_utcnow for Python 3.12 compatibility
• Add naive_utcfromtimestamp for Python 3.12 compatibility
• Change from pd.Timestamp.utcfromtimestamp() to pd.Timestamp.utcfromtimestamp().tz_localize(None) when needed

0.21.1

• Fix estimate_accrued_interest crashing when there are gaps in data
• Fix resample_candles to retain pair_id column if one is present in data

0.21

• Add functionality for downloading and manipulating Binance candle and lending data to have alternative datasets to benchmark the strategies for overfit and data quality issues
• Add resample_series() method to work for single column pd.Series price data
• Fix CandleUniverse.get_candles_by_pair() (add return value)
• Fix resample_candles() method

0.20.18

• API update: CandleUniverse.get_candles_by_pair() accepts DEXPair besides pair_id as an argument
• forward_fill() can handle volume column

0.20.17

• Fix: DataFrame copy warning in estimate_accrued_interest
• Add: get_price_with_tolerance(pair_name_hint) to have better exception messages
• Fix: CandleSampleUnavailable exception message was tuple instead of newline separated string

0.20.16

• Add LendingMetricUniverse.estimate_accrued_interest so we can estimate the position cost and interest profit in backtesting

0.20.15

• Update candle mappings so volume doesn't get set to None

0.20.14

• Internal change: move make_clickable to module level, so it can be reused

0.20.13

• API change: Client.create_live_client(settings_file_path) argument allows to disable and reading of settings file, to avoid any confusion in Dockerised environments
• Internal change: Unit tests do not try to poke settings file anymore

0.20.12

• Internal change: Use parquet.read_table(memory_map=True) to decrease RAM usage

0.20.11

• Add: LendingReserveUniverse.limit_to_assets()
• Add: filter_for_trading_fee(pairs_dataframe)
• Internal change: Add more asserts to Universe creation to catch human mistakes early
• API change: get_by_chain_and_symbol() raises UnknownLendingReserve instead of returning None
• Add more asserts to Universe creation to catch human mistakes early

0.20.10

• Remove duplicate timestamp label on visualisation

0.20.9

• Add: DEXPair.get_link() to get a direct link to the trading pair page
• Add: LendingReserve.get_link() to get a direct link to the trading pair page
• Add: DEXPair.is_tradeable() for quick checks if a pair has decent volume
• Add: PandasPairUniverse.get_exchange_for_pair()
• Add: LendingMetricUniverse.get_single_value()
• Add: Aave Ghost (GHST) stablecoin
• Add: Jarvis Synthetic Euro (jEUR) stablecoin
• Add: format_links_for_html_output() helper to work with DataFrame objects
• API update: get_pair_by_human_description() can have exchange slug set to None matching the best trading pair across with the lowest fee across all DEXes
• API update: get_closest_pair(pair) can now now take DEXPair instead of a primary key as a lookup parameter and also display the pair name in error message
• Various exception message improvements
• Remove duplicate timestamp from candle labels

0.20.8

• Add: Progress bar to lending rate download
• Fix: JSONL download progress bar was missing the last refresh, leaving the progress bar a bit under 100%
• Use tqdm_loggable package everywhere to have more control over progress bars

0.20.6

• Multiple API changes to make working with lending easier
• Lending universe construction examples
• Add: filter_for_base_tokens()
• Add: filter_for_chain()
• Add: DEXPair.get_base_token() and DEXPair.get_quote_token()
• Add: LendingReserveUniverse.can_leverage(token)
• Add: LendingReserveUniverse.limit_to_chain()
• API update: Make all Universe members optional, as we can have lending only universes
• API update: get_rates_by_reserve() can take a lending reserve object as an argument
• API update: Universe checks constructor argument types are correct
• Deprecate: Useless methods on Universe

0.20.5

• Add concurrent write control to CachedHTTPTransport to allow multiple processes to write the same cached download file on the same system
• Enable parallel tests

0.20.4

• Rename UniswapV2MockClient.initialise_mock_data to be more descriptive

0.20.3

• Refactor some common code from GenericMockClient and UniswapV2MockClient into MockClient. Note: GenericMockClient resides in tradeexecutor due to its reliance on certain imports
• More lending data points

0.20.2

• Add vToken details to Aave lending reserves
• Add liquidation threshold and such denormalised information to Aave lending reserves

0.20.1

• Fix: Lending rates cache bug - different datasets were using the same cache filename

0.20

• Add lending reserves and candles universe. Now you can easily download and explore Aave v3 lending rates with functions available in tradingstrategy.lending module

0.19

• Change PandasPairUniverse.get_pair_by_human_description() signature to be easier to use
• More user friendly data not available / 404 error handling for lending candles

0.18.2

• Add exchange_universe argument to get_pair()

0.18.1

• GroupedCandleUniverse.create_from_single_pair_dataframe(time_bucket) parameter added
• TimeBucket.from_pandas_timedelta() added

0.18

• Add support for fetching lending candle data with fetch_lending_candles_by_reserve_id()
• Rename fetch_all_lending_protocol_reserves() to fetch_all_lending_reserves()
• Delete summarydataframe.py and its dependents. This includes backtrader.py and the TradeSummary.to_dataframe method. Backtrader has been unmaintained for a while.

0.17.6

• Add loaded trading pair data preprocesing and working around Parquet data problems
• Add DataDecodeFailed(Exception) that gives more context information if there is something wrong with the trading pair

0.17.5

• Fix test warnings by changing function calls from create_single_pair_universe() to create_pair_universe()
• Add PairNotFoundError and ExchangeNotFoundError for more helpful error messages
• Add last_supposed_candle_at data to fetch_trading_data_availability()

0.17.4

• Fix fetch_trading_data_availability() using wrong format for pair id array

0.17.3

• get_one_pair_from_pandas_universe(exchange) support

0.17.2

• Bump web3-ethereum-defi to version 0.21

0.17.1

• Add candle_decimals parameter make_candle_labels() that defaults to showing 4 decimal places on candlestick charts

0.17

• Optimisation: get_price_with_tolerance(): candle price lookup is much faster with sparse data match
• Added trading_strategy.utils.forward_fill module for dealing with sparse data

0.16.1

• Optimisation: get_price_with_tolerance(): candle price lookup is 40x faster for exact timestamp matches

0.16

• API change: Now get_single_pair_data() raises NoDataAvailable instead of returning an empty data frame by default if there is not enough data at source
• Fix unnecessary upper() in is_stablecoin_like
• Fix Anvil chain id not fitting to Parquet data frame (assumed 16 bit uint for chain id)

0.15.2

• Add more stablecoins
• Add DEXPair.volume_30d shortcut
• Add DEXPair.from_row shortcut

0.15.1

• Fix: Better error messages when fetch_trading_data_availability() cannot process input

0.15

• Add DEXPair.fee_tier to get the trading pair fee as 0..1 float
• Add PandasPairUniverse.create_pair_universe
• Deprecate PandasPairUniverse.create_single_pair_universe
• Deprecate PandasPairUniverse.create_limited_pair_universe
• Add USDT.e and USDC.e bridged stablecoins on Avalanche

0.14.1

• Add: caching to get_candles_by_pair() and get_single() methods. This can result in backtesting time being more than halved.
• Fix: Bad wick filtering code crashed on empty dataframes

0.14

• Add: tradingstrategy.utils.groupeduniverse.fix_bad_wicks to deal with candle data where high and low values are abnormal due to various price manipulation issues.
• Update: Automatically filter bad wicks when creating a candle universe
• Add a shortcut function tradingstrategy.pair.generate_address_columns() to generate base_token_address and quote_token_address columns
• Update API: resolve_pairs_based_on_ticker() supports HumanReadableTradingPairDescription
• Fix: Some error message polish

0.13.18

• Have classical Candle.volume field as buy volume and sell volume cannot be separated for Uniswap v3

0.13.17

• change resample_candles() to use timedelta instead of timebucket

0.13.16

• Fix JSON data loading for Uni v3 exchanges that lack trade count

0.13.15

• Add fee_tier option to get_one_pair_from_pandas_universe() and get_pair_by_human_description()

0.13.14

• Bump max pairs to load over JSONL to 75

0.13.13

• Chart visualization fixes
  • remove price_chart_rel_size and subplot_rel_size options (will all be specified in relative_sizing)
  • refactor and clean code in tradingstrategy/charting/candle_chart.py

0.13.12

• Fix SyntaxError caused by a typo

0.13.11

• We have now tradingstrategey.client.BaseClient as the base class for different client implementations
• Adding UniswapV2MockClient implementation that mimics Client, but reads all data from an EVM test backend. Makes live trading tests much easier.
• Small improvements to PairUniverse

0.13.10

• Fix detached indicator charting bugs for different volume_bar_modes
• Adds support for different subtitle font sizes
• Adds tests to make to catch charting bugs earlier in the future

0.13.9

• Bump depenedncies
• Add ChainId.anvil
• Add support for detached technical indicators
• Web3 6.0 fixes

0.13.8

• Match dependencies and Python version pindowns with Web3.py 6.0
• Adding Ethereum Tester chain id

0.13.7

• Silence pandas warning in PairGroupedUniverse.get_pair_ids()
• Rename "Binance Smart Chain" to "BNB Smart Chain"

0.13.6

• Add a new method Client.fetch_all_lending_protocol_reserves() for fetching data on supported decentralized lending protocols (only AAVE v3 at present).
• Improve validate_ohclv_dataframe() to check for date/timestamp index as well as column

0.13.5

• Fixes for filtering by chain_id in ExchangeUniverse

0.13.4

• Improve resolve_pairs_based_on_ticker() to be able to filter pairs with specified fee. Change default sorting criteria buy_volume_all_time to be function argument for more flexible filtering

0.13.3

• Add GroupedCandleUniverse.get_last_entries_by_pair_and_timestamp(pair_id, timestamp) for simplified candle data access
• Add PandasPairUniverse.get_pair(chain_id, exchange_id, base_token, quote_token) for simplified pair access
• Support passing exchange_universe in PandasPairUniverse construction, so we have all metadata needed to look up pairs
• Fixes to various token data accessor

0.13.2

• Same as 0.13.1 - fixing broken release

0.13.1

• Support Arbitrum L2 in ChainId class

0.13.1

A large impact optimisation update.

• Optimise PandasPairUniver.get_pair_by_id to cache full DEXPair objects to speed up strategies using large number of trading pairs

• Make DEXPair Python __slots__ based to make the object property access faster

• Added ResampledLiquidityUniverse for faster backtesting with luidity data

• Allow different granularity of OHLCV and liquidity data in Universe

0.13

• Release failed

0.12.5

• Add LiquidityUniverse.get_liquidity_with_tolerance

0.12.4

• Quick fix on get_pair_by_human_description error message

0.12.3

• Improve CandleSampleUnavailable error messge further to offer friendly advises how to fix
• Improve ExchangeUniverse interface with more helpful shorthand methods

0.12.2

• Add ExchangeUniverse.from_collection()
• Pass ExchangeUniverse instance as a part of Universe.exchange_universe instead just raw array of of univer.exchanges

0.12.1

• Better multipair lookup when typing out pair names by hand
• Add tradingstrategy.pair.HumanReadableTradingPairDescription
• Add PandasPairUniverse.get_pair_by_human_description

0.12

• Add default HTTP request retry policy to Client
• Add default HTTP user agent to Client
• More direct data feed infrastructure
• Fix create_requests_client(add_exception_hook=True) not working correctly
• Add GroupedCandleUniverse.create_from_multiple_candle_datafarames

0.11.1

• Add: Make DEXPair hashable
• Fix: Historical candle download for JSONL endpoint - make time ranges to work correctly
• Change: Make JSONL cache filenames more descriptive

0.11

• Adding Client.fetch_trading_data_availability API
• Moved types.py to properly Python 3.10 TypeAlias system

0.10.2

• Add TradeSummary.show to simplify usage in notebooks

0.10.1

• Fix missing dependencies because web3-ethereum-defi data extras
• Fix make_candle_labels not working correctly if some keys were not present in the OHCLV DataFrame
• Add VolumeBarMode option to visualise_ohlcv to allow different rendering modes for volume bars, in-chart or outside chart
• More implementation of direct-to-blockchain node data feeds

0.10.0

• Add the initial implementation for direct-to-blockchain node data feeds

0.9.1

• Add support for average duration statistic to be expressed in terms of bars
• Framework for direct, real-time, date feeds directly from a blockchain node
• Add uniswap_v2_incompatible exchange type

0.9.0

• Make the default behavior not to return the current candle if asked by timestamp

0.8.8

• Switch to tqdm-loggable to have better download status behavior on headless configurations
• Fine tune visualise_ohlcv label configuration

0.8.7

• Fix spelling visualise_ohlcv

0.8.6

• Add candle charts with visualise_ohclv

0.8.5

• Optimize groupeduniverse class, specifically get_timestamp_range() method

• Optimize get pair speed

• Remove the "no Content-Length header" warning if HTTP response lacks file size information

• Fix get_token() not returning symbol information

0.8.4

• Added Client.close() and CachedHTTPTransport.close() for explicit closing of streams
• Fix JUPYTER_PLATFORM_DIRS warning when importing
• Fixed bunch of Pandas warnings
• Add pytest -Werror to ensure the lib does not raise any warnings

0.8.3

• Escape hatch for interactive API key setup loop
• Added Candle.generate_synthetic_sample for mocking test data
• Added GroupedCandleUniverse.get_price_with_tolerance to safely get the latest price in backtesting

0.8.2

• Make matplotlib optional in Client.setup_notebook
• Added DEXPair.exchange_type to support Uniswap v3 pairs

0.8.1

• Change: List requests as an explicit dependency for the client (Might have been causing issues with Pyodide builds)

0.8.0

• Change: Remove a lot of dependencies and cleaning up old code. The package with its dependencies should be now much more compact. Any old Backtrader or QSTrader code is behind an extra dependency in the Python package definition. There is no longer dependency to Ethereum or crypto packages that had C and were problematic install.
• Change: preparing for Pyodide support.
• Change: Bumped Python 3.10 because 3.10 is minimum for Pyodide

0.7.2

• Fix: PairGroupedUniverse did not have OHLCV data sorted by timestamp, but instead of was pair_id, timestamp causing some failures of managing expectations when accessing data.
• Fix: Update tests to reflect new datasets from Trading Strategy API
• Added Client.fetch_candles_by_pair_ids that loads trading data for certain pairs
  data over JSONL endpoint and avoids loading large Parquest candle and liquidity files. This makes it possible to run backtests on low memory environments.
• Added Client.fetch_candle_dataset and read_parquest(filters) to filter Parquest files when loading candle datasets to optimise memory usage. (At the moment this path is not used outside the tests.)
• Change ChainId base class to IntEnum instead of Enum for better type hinting
• Change PandasPairUniverse.build_index to use Pandas DataFrame frame transpose, greatly speeding up the creation of the index
• Fix: get_closest_price gets a closest price even if the timestamp and
  and time bucket are not in the same internal
• Added environment/jupyterlite for the upcoming WebAssembly integration

0.7.1

• Added is_candle_green(), is_candle_red() helpers
• Added TimeBucket.to_pandas_timedelta() helper
• Added upsample_candles() helper
• Added unknown and osmosis blockchains

0.7

• Add Token presentation
• Add get_single_pair_data shortcut
• Add PandasPairUniverse.get_by_symbols shortcut
• Add PandasPairUniverse.get_all_tokens shortcut
• Add get_exchange_by_id shortcut
• Add summarydata.as_duration Pandas summary table cell formatter
• Fix filter_for_quote_tokens to be more strict (was checking either side of the pair, not quote token)
• API change create_single_pair_universe now to accept pick_by_highest_vol
• API change: Rename Universe.time_frame -> Universe.time_bucket to be consistent
• Make download progress bars more human friendly

0.6.9

• Add check_schema option when creating PyArrow table exports for trading pairs
• Fix type hint for token1_decimals
• Better error message for get_one_pair_from_pandas_universe
• Make token0_symbol and token1_symbol optional as not all tokens have a symbol

0.6.8

• Patch the previous release to token0_decimals and token1_decimals instead of base_token_decimals to be more aligned with the other Uniswap pair data
• Cleaning up API and API documentation

0.6.7

• Export base_token_decimals and quote_token_decimals in trading pair datasets

0.6.6

• New time points for the time buckets

0.6.5

• Add exchange_address as a part of trading pair datasets
• Add get_by_chain_and_factory() to decentralised exchange universe

0.6.4

• Add buy_tax, transfer_tax, sell_tax to trading pair datasets

0.6.3

• Fix download retries in live trading
• Add stablecoin module for upcoming stablecoin support fuctions
• Add DEXPair.convert_to_dataframe
• Add filter_for_stablecoins

0.6.2

• Fix broken build

0.6.1

• Fix Macbook M1, macOS, arm64 compatibility. Updated Scipy, Pyarrow and Numpy dependencies.

0.6

• Renaming the package tradingstrategy -> trading-strategy to be consistent with the package naming practices. See the new Github repository.

0.5.3

• Fix download responses being buffered in CLI
• Test for get_trading_pair_page_url() in the downloaded dataset
• Added support for Ganacher tester chain

0.5.2

• Added Client.create_live_client for live trading
• More helper functions to help unit testing
• tradingstrategy.utils.time to check for incompatible pandas.Timestamp formats
• Added tradingstrategy.pair.filter_for_quote_tokens
• Added get_all_samples_by_range, iterate_samples_by_pair_range shortcuts to consume trading pair data in strategies
• Added get_prior_timestamp to calibrate your clock with the existing time index
• Added Client.clear_caches to ensure fresh data downloads
• Added DEXPair.exchange_slug and DEXPair.pair_slug so we can point to web page URLs
• Renamed Sushiswap -> Sushi as per their branding

0.5.1

• Adding chain id 61 for Ethereum Classic / Ethereum Tester
• Adding GroupedCandleUniverse.create_empty() and GroupedLiquidityUniverse.create_empty()

0.5.0

This is a release to get Trading Strategy client towards live trading.

• Created tradingstrategy.universe.Universe helper class to encapsulate trading universe
• Added convenience method PairUniverse.create_from_pyarrow_table_with_filters so creating different trading universes is less lines of code
• Added convenience method tradingstrategy.utils.groupeduniverse.filter_for_pairs_on_exchanges
• Added helper methods to construct single trading pair universe, see PandasPairUniverse.create_single_pair_universe
• Separate QSTrader to an optional dependency. Please install as pip install tradingstrategy[qstrader]. This is because QSTrader depends on Seaborn that depends on SciPy that does not work so good on Apple M1 macs.
• Add the default connection and read timeout 15 seconds to all client downloads

0.4.0

• Reworked how QSTrader framework integration works. QSTrader was originally designed for stock markets, so it required few changes to make it more suitable for 24/7 cryptomarkets. Warning: QSTrader and its integration are both in beta. More work is needed to be done e.g. in fee calculations to make the integration smooth. See example on PancakeSwap momentum trading algorithm.
• Added portfolioanalyzer to make timeline of portfolio construction over the time. See example on PancakeSwap momentum trading algorithm.
• Include seaborn as a dependency as it is required for qstrader plot output, otherwise qstrader was crashing.
• PairUniverse supports indexed lookups to deal with the high pair count.
• More convenience methods for candle and liquidity manipulation.

0.3.4

• Adding the preliminaty price impact analysis. See the example notebook. Read the blog post about the liquidity formation.

0.3.3

• Make fastquant dependency optional, as it was causing installation issues with its ccxt dependency

0.3.2

• Updated inline interactive setup from within Jupyter notebook

0.3.1

• Polish README, release process and docs

0.3.0

• Multiple changes to make the multichain backtesting possible.

• The documentation code examples will be updated to reflect multichain support over time and may work incorrectly at the moment. Getting Started tutorial is already updated.

• This is a major release deprecating activity flags in the trading pair universe. The multichain trading pair data is too big to include inactivate trading pairs (800k+ total trading pairs). Thus, the pair universe set only contains active trading pairs after this release, making the trading pair universe less than 100k trading pairs again, making it more feasible to download the data.

• ExchangeUniverse.get_by_chain_and_slug() is now canonical way to refer to an exchange

• PairUniverse.get_pair_by_ticker_by_exchange() is now canonical way to refer to a trading pair

• Clarify primary keys may not be stable and should no longer referred permanently

• Exchange.homepage and Exchange.active_pair_count fields added

0.2.14

• Make ChainId database more deterministic by loading all supported chains once and only once

0.2.13

• Add chain_name output field for exchange data

0.2.12

• Adding slug lookups for chains

0.2.11

• Add Binance Smart Chain and Polygon logos

0.2.10

• Fix the default API endpoint URL

0.2.9

• Adding embedded chain list repository
• Adding support for Binance Smart Chain

0.2.8

• Adding more metadata information on ChainId objects

0.2.7

• Adding exchange_slug and chain_slug on the Exchange object

0.2.6

• Allow to customise color range for the trade success distribution diagram

0.2.5

• Added custom DEXTrader base class, as the default logic with Backtrader was a bit insufficient

• Added trade analysis support for Backtrader strategies (earlier was QSTrader only)

• Reworked trade analysis summaries and views

• Added advanced example for Double 7 trading strategy

0.2.4

• Polish API key welcome setup

0.2.3

• Fixed not reading settings file from Google Drive