stable/freqtrade/persistence/trade_model.py

1347 lines
52 KiB
Python

"""
This module contains the class to persist trades into SQLite
"""
import logging
from datetime import datetime, timedelta, timezone
from decimal import Decimal
from typing import Any, Dict, List, Optional
from sqlalchemy import (Boolean, Column, DateTime, Enum, Float, ForeignKey, Integer, String,
UniqueConstraint, desc, func)
from sqlalchemy.orm import Query, relationship
from freqtrade.constants import DATETIME_PRINT_FORMAT, NON_OPEN_EXCHANGE_STATES, BuySell, LongShort
from freqtrade.enums import ExitType, TradingMode
from freqtrade.exceptions import DependencyException, OperationalException
from freqtrade.leverage import interest
from freqtrade.persistence.base import _DECL_BASE
logger = logging.getLogger(__name__)
class Order(_DECL_BASE):
"""
Order database model
Keeps a record of all orders placed on the exchange
One to many relationship with Trades:
- One trade can have many orders
- One Order can only be associated with one Trade
Mirrors CCXT Order structure
"""
__tablename__ = 'orders'
# Uniqueness should be ensured over pair, order_id
# its likely that order_id is unique per Pair on some exchanges.
__table_args__ = (UniqueConstraint('ft_pair', 'order_id', name="_order_pair_order_id"),)
id = Column(Integer, primary_key=True)
ft_trade_id = Column(Integer, ForeignKey('trades.id'), index=True)
trade = relationship("Trade", back_populates="orders")
# order_side can only be 'buy', 'sell' or 'stoploss'
ft_order_side: str = Column(String(25), nullable=False)
ft_pair: str = Column(String(25), nullable=False)
ft_is_open = Column(Boolean, nullable=False, default=True, index=True)
order_id: str = Column(String(255), nullable=False, index=True)
status = Column(String(255), nullable=True)
symbol = Column(String(25), nullable=True)
order_type: str = Column(String(50), nullable=True)
side = Column(String(25), nullable=True)
price = Column(Float, nullable=True)
average = Column(Float, nullable=True)
amount = Column(Float, nullable=True)
filled = Column(Float, nullable=True)
remaining = Column(Float, nullable=True)
cost = Column(Float, nullable=True)
order_date = Column(DateTime, nullable=True, default=datetime.utcnow)
order_filled_date = Column(DateTime, nullable=True)
order_update_date = Column(DateTime, nullable=True)
ft_fee_base = Column(Float, nullable=True)
@property
def order_date_utc(self) -> datetime:
""" Order-date with UTC timezoneinfo"""
return self.order_date.replace(tzinfo=timezone.utc)
@property
def safe_price(self) -> float:
return self.average or self.price
@property
def safe_filled(self) -> float:
return self.filled or self.amount or 0.0
@property
def safe_fee_base(self) -> float:
return self.ft_fee_base or 0.0
@property
def safe_amount_after_fee(self) -> float:
return self.safe_filled - self.safe_fee_base
def __repr__(self):
return (f'Order(id={self.id}, order_id={self.order_id}, trade_id={self.ft_trade_id}, '
f'side={self.side}, order_type={self.order_type}, status={self.status})')
def update_from_ccxt_object(self, order):
"""
Update Order from ccxt response
Only updates if fields are available from ccxt -
"""
if self.order_id != str(order['id']):
raise DependencyException("Order-id's don't match")
self.status = order.get('status', self.status)
self.symbol = order.get('symbol', self.symbol)
self.order_type = order.get('type', self.order_type)
self.side = order.get('side', self.side)
self.price = order.get('price', self.price)
self.amount = order.get('amount', self.amount)
self.filled = order.get('filled', self.filled)
self.average = order.get('average', self.average)
self.remaining = order.get('remaining', self.remaining)
self.cost = order.get('cost', self.cost)
if 'timestamp' in order and order['timestamp'] is not None:
self.order_date = datetime.fromtimestamp(order['timestamp'] / 1000, tz=timezone.utc)
self.ft_is_open = True
if self.status in NON_OPEN_EXCHANGE_STATES:
self.ft_is_open = False
if (order.get('filled', 0.0) or 0.0) > 0:
self.order_filled_date = datetime.now(timezone.utc)
self.order_update_date = datetime.now(timezone.utc)
def to_json(self, entry_side: str) -> Dict[str, Any]:
return {
'pair': self.ft_pair,
'order_id': self.order_id,
'status': self.status,
'amount': self.amount,
'average': round(self.average, 8) if self.average else 0,
'safe_price': self.safe_price,
'cost': self.cost if self.cost else 0,
'filled': self.filled,
'ft_order_side': self.ft_order_side,
'is_open': self.ft_is_open,
'order_date': self.order_date.strftime(DATETIME_PRINT_FORMAT)
if self.order_date else None,
'order_timestamp': int(self.order_date.replace(
tzinfo=timezone.utc).timestamp() * 1000) if self.order_date else None,
'order_filled_date': self.order_filled_date.strftime(DATETIME_PRINT_FORMAT)
if self.order_filled_date else None,
'order_filled_timestamp': int(self.order_filled_date.replace(
tzinfo=timezone.utc).timestamp() * 1000) if self.order_filled_date else None,
'order_type': self.order_type,
'price': self.price,
'ft_is_entry': self.ft_order_side == entry_side,
'remaining': self.remaining,
}
def close_bt_order(self, close_date: datetime, trade: 'LocalTrade'):
self.order_filled_date = close_date
self.filled = self.amount
self.status = 'closed'
self.ft_is_open = False
if (self.ft_order_side == trade.entry_side
and len(trade.select_filled_orders(trade.entry_side)) == 1):
trade.open_rate = self.price
trade.recalc_open_trade_value()
@staticmethod
def update_orders(orders: List['Order'], order: Dict[str, Any]):
"""
Get all non-closed orders - useful when trying to batch-update orders
"""
if not isinstance(order, dict):
logger.warning(f"{order} is not a valid response object.")
return
filtered_orders = [o for o in orders if o.order_id == order.get('id')]
if filtered_orders:
oobj = filtered_orders[0]
oobj.update_from_ccxt_object(order)
Order.query.session.commit()
else:
logger.warning(f"Did not find order for {order}.")
@staticmethod
def parse_from_ccxt_object(order: Dict[str, Any], pair: str, side: str) -> 'Order':
"""
Parse an order from a ccxt object and return a new order Object.
"""
o = Order(order_id=str(order['id']), ft_order_side=side, ft_pair=pair)
o.update_from_ccxt_object(order)
return o
@staticmethod
def get_open_orders() -> List['Order']:
"""
Retrieve open orders from the database
:return: List of open orders
"""
return Order.query.filter(Order.ft_is_open.is_(True)).all()
class LocalTrade():
"""
Trade database model.
Used in backtesting - must be aligned to Trade model!
"""
use_db: bool = False
# Trades container for backtesting
trades: List['LocalTrade'] = []
trades_open: List['LocalTrade'] = []
total_profit: float = 0
id: int = 0
orders: List[Order] = []
exchange: str = ''
pair: str = ''
base_currency: str = ''
stake_currency: str = ''
is_open: bool = True
fee_open: float = 0.0
fee_open_cost: Optional[float] = None
fee_open_currency: str = ''
fee_close: float = 0.0
fee_close_cost: Optional[float] = None
fee_close_currency: str = ''
open_rate: float = 0.0
open_rate_requested: Optional[float] = None
# open_trade_value - calculated via _calc_open_trade_value
open_trade_value: float = 0.0
close_rate: Optional[float] = None
close_rate_requested: Optional[float] = None
close_profit: Optional[float] = None
close_profit_abs: Optional[float] = None
stake_amount: float = 0.0
amount: float = 0.0
amount_requested: Optional[float] = None
open_date: datetime
close_date: Optional[datetime] = None
open_order_id: Optional[str] = None
# absolute value of the stop loss
stop_loss: float = 0.0
# percentage value of the stop loss
stop_loss_pct: float = 0.0
# absolute value of the initial stop loss
initial_stop_loss: float = 0.0
# percentage value of the initial stop loss
initial_stop_loss_pct: Optional[float] = None
# stoploss order id which is on exchange
stoploss_order_id: Optional[str] = None
# last update time of the stoploss order on exchange
stoploss_last_update: Optional[datetime] = None
# absolute value of the highest reached price
max_rate: float = 0.0
# Lowest price reached
min_rate: float = 0.0
exit_reason: str = ''
exit_order_status: str = ''
strategy: str = ''
enter_tag: Optional[str] = None
timeframe: Optional[int] = None
trading_mode: TradingMode = TradingMode.SPOT
# Leverage trading properties
liquidation_price: Optional[float] = None
is_short: bool = False
leverage: float = 1.0
# Margin trading properties
interest_rate: float = 0.0
# Futures properties
funding_fees: Optional[float] = None
@property
def buy_tag(self) -> Optional[str]:
"""
Compatibility between buy_tag (old) and enter_tag (new)
Consider buy_tag deprecated
"""
return self.enter_tag
@property
def has_no_leverage(self) -> bool:
"""Returns true if this is a non-leverage, non-short trade"""
return ((self.leverage == 1.0 or self.leverage is None) and not self.is_short)
@property
def borrowed(self) -> float:
"""
The amount of currency borrowed from the exchange for leverage trades
If a long trade, the amount is in base currency
If a short trade, the amount is in the other currency being traded
"""
if self.has_no_leverage:
return 0.0
elif not self.is_short:
return (self.amount * self.open_rate) * ((self.leverage - 1) / self.leverage)
else:
return self.amount
@property
def open_date_utc(self):
return self.open_date.replace(tzinfo=timezone.utc)
@property
def close_date_utc(self):
return self.close_date.replace(tzinfo=timezone.utc)
@property
def enter_side(self) -> str:
""" DEPRECATED, please use entry_side instead"""
# TODO: Please remove me after 2022.5
return self.entry_side
@property
def entry_side(self) -> str:
if self.is_short:
return "sell"
else:
return "buy"
@property
def exit_side(self) -> BuySell:
if self.is_short:
return "buy"
else:
return "sell"
@property
def trade_direction(self) -> LongShort:
if self.is_short:
return "short"
else:
return "long"
@property
def safe_base_currency(self) -> str:
"""
Compatibility layer for asset - which can be empty for old trades.
"""
try:
return self.base_currency or self.pair.split('/')[0]
except IndexError:
return ''
@property
def safe_quote_currency(self) -> str:
"""
Compatibility layer for asset - which can be empty for old trades.
"""
try:
return self.stake_currency or self.pair.split('/')[1].split(':')[0]
except IndexError:
return ''
def __init__(self, **kwargs):
for key in kwargs:
setattr(self, key, kwargs[key])
self.recalc_open_trade_value()
if self.trading_mode == TradingMode.MARGIN and self.interest_rate is None:
raise OperationalException(
f"{self.trading_mode.value} trading requires param interest_rate on trades")
def __repr__(self):
open_since = self.open_date.strftime(DATETIME_PRINT_FORMAT) if self.is_open else 'closed'
return (
f'Trade(id={self.id}, pair={self.pair}, amount={self.amount:.8f}, '
f'is_short={self.is_short or False}, leverage={self.leverage or 1.0}, '
f'open_rate={self.open_rate:.8f}, open_since={open_since})'
)
def to_json(self) -> Dict[str, Any]:
filled_orders = self.select_filled_orders()
orders = [order.to_json(self.entry_side) for order in filled_orders]
return {
'trade_id': self.id,
'pair': self.pair,
'base_currency': self.safe_base_currency,
'quote_currency': self.safe_quote_currency,
'is_open': self.is_open,
'exchange': self.exchange,
'amount': round(self.amount, 8),
'amount_requested': round(self.amount_requested, 8) if self.amount_requested else None,
'stake_amount': round(self.stake_amount, 8),
'strategy': self.strategy,
'buy_tag': self.enter_tag,
'enter_tag': self.enter_tag,
'timeframe': self.timeframe,
'fee_open': self.fee_open,
'fee_open_cost': self.fee_open_cost,
'fee_open_currency': self.fee_open_currency,
'fee_close': self.fee_close,
'fee_close_cost': self.fee_close_cost,
'fee_close_currency': self.fee_close_currency,
'open_date': self.open_date.strftime(DATETIME_PRINT_FORMAT),
'open_timestamp': int(self.open_date.replace(tzinfo=timezone.utc).timestamp() * 1000),
'open_rate': self.open_rate,
'open_rate_requested': self.open_rate_requested,
'open_trade_value': round(self.open_trade_value, 8),
'close_date': (self.close_date.strftime(DATETIME_PRINT_FORMAT)
if self.close_date else None),
'close_timestamp': int(self.close_date.replace(
tzinfo=timezone.utc).timestamp() * 1000) if self.close_date else None,
'close_rate': self.close_rate,
'close_rate_requested': self.close_rate_requested,
'close_profit': self.close_profit, # Deprecated
'close_profit_pct': round(self.close_profit * 100, 2) if self.close_profit else None,
'close_profit_abs': self.close_profit_abs, # Deprecated
'trade_duration_s': (int((self.close_date_utc - self.open_date_utc).total_seconds())
if self.close_date else None),
'trade_duration': (int((self.close_date_utc - self.open_date_utc).total_seconds() // 60)
if self.close_date else None),
'profit_ratio': self.close_profit,
'profit_pct': round(self.close_profit * 100, 2) if self.close_profit else None,
'profit_abs': self.close_profit_abs,
'sell_reason': self.exit_reason, # Deprecated
'exit_reason': self.exit_reason,
'exit_order_status': self.exit_order_status,
'stop_loss_abs': self.stop_loss,
'stop_loss_ratio': self.stop_loss_pct if self.stop_loss_pct else None,
'stop_loss_pct': (self.stop_loss_pct * 100) if self.stop_loss_pct else None,
'stoploss_order_id': self.stoploss_order_id,
'stoploss_last_update': (self.stoploss_last_update.strftime(DATETIME_PRINT_FORMAT)
if self.stoploss_last_update else None),
'stoploss_last_update_timestamp': int(self.stoploss_last_update.replace(
tzinfo=timezone.utc).timestamp() * 1000) if self.stoploss_last_update else None,
'initial_stop_loss_abs': self.initial_stop_loss,
'initial_stop_loss_ratio': (self.initial_stop_loss_pct
if self.initial_stop_loss_pct else None),
'initial_stop_loss_pct': (self.initial_stop_loss_pct * 100
if self.initial_stop_loss_pct else None),
'min_rate': self.min_rate,
'max_rate': self.max_rate,
'leverage': self.leverage,
'interest_rate': self.interest_rate,
'liquidation_price': self.liquidation_price,
'is_short': self.is_short,
'trading_mode': self.trading_mode,
'funding_fees': self.funding_fees,
'open_order_id': self.open_order_id,
'orders': orders,
}
@staticmethod
def reset_trades() -> None:
"""
Resets all trades. Only active for backtesting mode.
"""
LocalTrade.trades = []
LocalTrade.trades_open = []
LocalTrade.total_profit = 0
def adjust_min_max_rates(self, current_price: float, current_price_low: float) -> None:
"""
Adjust the max_rate and min_rate.
"""
self.max_rate = max(current_price, self.max_rate or self.open_rate)
self.min_rate = min(current_price_low, self.min_rate or self.open_rate)
def set_isolated_liq(self, liquidation_price: Optional[float]):
"""
Method you should use to set self.liquidation price.
Assures stop_loss is not passed the liquidation price
"""
if not liquidation_price:
return
self.liquidation_price = liquidation_price
def _set_stop_loss(self, stop_loss: float, percent: float):
"""
Method you should use to set self.stop_loss.
Assures stop_loss is not passed the liquidation price
"""
if self.liquidation_price is not None:
if self.is_short:
sl = min(stop_loss, self.liquidation_price)
else:
sl = max(stop_loss, self.liquidation_price)
else:
sl = stop_loss
if not self.stop_loss:
self.initial_stop_loss = sl
self.stop_loss = sl
self.stop_loss_pct = -1 * abs(percent)
self.stoploss_last_update = datetime.utcnow()
def adjust_stop_loss(self, current_price: float, stoploss: float,
initial: bool = False, refresh: bool = False) -> None:
"""
This adjusts the stop loss to it's most recently observed setting
:param current_price: Current rate the asset is traded
:param stoploss: Stoploss as factor (sample -0.05 -> -5% below current price).
:param initial: Called to initiate stop_loss.
Skips everything if self.stop_loss is already set.
"""
if initial and not (self.stop_loss is None or self.stop_loss == 0):
# Don't modify if called with initial and nothing to do
return
refresh = False if self.nr_of_successful_entries > 0 else refresh
leverage = self.leverage or 1.0
if self.is_short:
new_loss = float(current_price * (1 + abs(stoploss / leverage)))
# If trading with leverage, don't set the stoploss below the liquidation price
if self.liquidation_price:
new_loss = min(self.liquidation_price, new_loss)
else:
new_loss = float(current_price * (1 - abs(stoploss / leverage)))
# If trading with leverage, don't set the stoploss below the liquidation price
if self.liquidation_price:
new_loss = max(self.liquidation_price, new_loss)
# no stop loss assigned yet
if self.initial_stop_loss_pct is None or refresh:
self._set_stop_loss(new_loss, stoploss)
self.initial_stop_loss = new_loss
self.initial_stop_loss_pct = -1 * abs(stoploss)
# evaluate if the stop loss needs to be updated
else:
higher_stop = new_loss > self.stop_loss
lower_stop = new_loss < self.stop_loss
# stop losses only walk up, never down!,
# ? But adding more to a leveraged trade would create a lower liquidation price,
# ? decreasing the minimum stoploss
if (higher_stop and not self.is_short) or (lower_stop and self.is_short):
logger.debug(f"{self.pair} - Adjusting stoploss...")
self._set_stop_loss(new_loss, stoploss)
else:
logger.debug(f"{self.pair} - Keeping current stoploss...")
logger.debug(
f"{self.pair} - Stoploss adjusted. current_price={current_price:.8f}, "
f"open_rate={self.open_rate:.8f}, max_rate={self.max_rate or self.open_rate:.8f}, "
f"initial_stop_loss={self.initial_stop_loss:.8f}, "
f"stop_loss={self.stop_loss:.8f}. "
f"Trailing stoploss saved us: "
f"{float(self.stop_loss) - float(self.initial_stop_loss):.8f}.")
def update_trade(self, order: Order) -> None:
"""
Updates this entity with amount and actual open/close rates.
:param order: order retrieved by exchange.fetch_order()
:return: None
"""
# Ignore open and cancelled orders
if order.status == 'open' or order.safe_price is None:
return
logger.info(f'Updating trade (id={self.id}) ...')
if order.ft_order_side == self.entry_side:
# Update open rate and actual amount
self.open_rate = order.safe_price
self.amount = order.safe_amount_after_fee
if self.is_open:
payment = "SELL" if self.is_short else "BUY"
logger.info(f'{order.order_type.upper()}_{payment} has been fulfilled for {self}.')
self.open_order_id = None
self.recalc_trade_from_orders()
elif order.ft_order_side == self.exit_side:
if self.is_open:
payment = "BUY" if self.is_short else "SELL"
# * On margin shorts, you buy a little bit more than the amount (amount + interest)
logger.info(f'{order.order_type.upper()}_{payment} has been fulfilled for {self}.')
self.close(order.safe_price)
elif order.ft_order_side == 'stoploss':
self.stoploss_order_id = None
self.close_rate_requested = self.stop_loss
self.exit_reason = ExitType.STOPLOSS_ON_EXCHANGE.value
if self.is_open:
logger.info(f'{order.order_type.upper()} is hit for {self}.')
self.close(order.safe_price)
else:
raise ValueError(f'Unknown order type: {order.order_type}')
Trade.commit()
def close(self, rate: float, *, show_msg: bool = True) -> None:
"""
Sets close_rate to the given rate, calculates total profit
and marks trade as closed
"""
self.close_rate = rate
self.close_date = self.close_date or datetime.utcnow()
self.close_profit = self.calc_profit_ratio()
self.close_profit_abs = self.calc_profit()
self.is_open = False
self.exit_order_status = 'closed'
self.open_order_id = None
if show_msg:
logger.info(
'Marking %s as closed as the trade is fulfilled and found no open orders for it.',
self
)
def update_fee(self, fee_cost: float, fee_currency: Optional[str], fee_rate: Optional[float],
side: str) -> None:
"""
Update Fee parameters. Only acts once per side
"""
if self.entry_side == side and self.fee_open_currency is None:
self.fee_open_cost = fee_cost
self.fee_open_currency = fee_currency
if fee_rate is not None:
self.fee_open = fee_rate
# Assume close-fee will fall into the same fee category and take an educated guess
self.fee_close = fee_rate
elif self.exit_side == side and self.fee_close_currency is None:
self.fee_close_cost = fee_cost
self.fee_close_currency = fee_currency
if fee_rate is not None:
self.fee_close = fee_rate
def fee_updated(self, side: str) -> bool:
"""
Verify if this side (buy / sell) has already been updated
"""
if self.entry_side == side:
return self.fee_open_currency is not None
elif self.exit_side == side:
return self.fee_close_currency is not None
else:
return False
def update_order(self, order: Dict) -> None:
Order.update_orders(self.orders, order)
def get_exit_order_count(self) -> int:
"""
Get amount of failed exiting orders
assumes full exits.
"""
return len([o for o in self.orders if o.ft_order_side == self.exit_side])
def _calc_open_trade_value(self) -> float:
"""
Calculate the open_rate including open_fee.
:return: Price in of the open trade incl. Fees
"""
open_trade = Decimal(self.amount) * Decimal(self.open_rate)
fees = open_trade * Decimal(self.fee_open)
if self.is_short:
return float(open_trade - fees)
else:
return float(open_trade + fees)
def recalc_open_trade_value(self) -> None:
"""
Recalculate open_trade_value.
Must be called whenever open_rate, fee_open is changed.
"""
self.open_trade_value = self._calc_open_trade_value()
def calculate_interest(self, interest_rate: Optional[float] = None) -> Decimal:
"""
:param interest_rate: interest_charge for borrowing this coin(optional).
If interest_rate is not set self.interest_rate will be used
"""
zero = Decimal(0.0)
# If nothing was borrowed
if self.trading_mode != TradingMode.MARGIN or self.has_no_leverage:
return zero
open_date = self.open_date.replace(tzinfo=None)
now = (self.close_date or datetime.now(timezone.utc)).replace(tzinfo=None)
sec_per_hour = Decimal(3600)
total_seconds = Decimal((now - open_date).total_seconds())
hours = total_seconds / sec_per_hour or zero
rate = Decimal(interest_rate or self.interest_rate)
borrowed = Decimal(self.borrowed)
return interest(exchange_name=self.exchange, borrowed=borrowed, rate=rate, hours=hours)
def _calc_base_close(self, amount: Decimal, rate: Optional[float] = None,
fee: Optional[float] = None) -> Decimal:
close_trade = Decimal(amount) * Decimal(rate or self.close_rate) # type: ignore
fees = close_trade * Decimal(fee or self.fee_close)
if self.is_short:
return close_trade + fees
else:
return close_trade - fees
def calc_close_trade_value(self, rate: Optional[float] = None,
fee: Optional[float] = None,
interest_rate: Optional[float] = None) -> float:
"""
Calculate the close_rate including fee
:param fee: fee to use on the close rate (optional).
If rate is not set self.fee will be used
:param rate: rate to compare with (optional).
If rate is not set self.close_rate will be used
:param interest_rate: interest_charge for borrowing this coin (optional).
If interest_rate is not set self.interest_rate will be used
:return: Price in BTC of the open trade
"""
if rate is None and not self.close_rate:
return 0.0
amount = Decimal(self.amount)
trading_mode = self.trading_mode or TradingMode.SPOT
if trading_mode == TradingMode.SPOT:
return float(self._calc_base_close(amount, rate, fee))
elif (trading_mode == TradingMode.MARGIN):
total_interest = self.calculate_interest(interest_rate)
if self.is_short:
amount = amount + total_interest
return float(self._calc_base_close(amount, rate, fee))
else:
# Currency already owned for longs, no need to purchase
return float(self._calc_base_close(amount, rate, fee) - total_interest)
elif (trading_mode == TradingMode.FUTURES):
funding_fees = self.funding_fees or 0.0
# Positive funding_fees -> Trade has gained from fees.
# Negative funding_fees -> Trade had to pay the fees.
if self.is_short:
return float(self._calc_base_close(amount, rate, fee)) - funding_fees
else:
return float(self._calc_base_close(amount, rate, fee)) + funding_fees
else:
raise OperationalException(
f"{self.trading_mode.value} trading is not yet available using freqtrade")
def calc_profit(self, rate: Optional[float] = None,
fee: Optional[float] = None,
interest_rate: Optional[float] = None) -> float:
"""
Calculate the absolute profit in stake currency between Close and Open trade
:param fee: fee to use on the close rate (optional).
If fee is not set self.fee will be used
:param rate: close rate to compare with (optional).
If rate is not set self.close_rate will be used
:param interest_rate: interest_charge for borrowing this coin (optional).
If interest_rate is not set self.interest_rate will be used
:return: profit in stake currency as float
"""
close_trade_value = self.calc_close_trade_value(
rate=(rate or self.close_rate),
fee=(fee or self.fee_close),
interest_rate=(interest_rate or self.interest_rate)
)
if self.is_short:
profit = self.open_trade_value - close_trade_value
else:
profit = close_trade_value - self.open_trade_value
return float(f"{profit:.8f}")
def calc_profit_ratio(self, rate: Optional[float] = None,
fee: Optional[float] = None,
interest_rate: Optional[float] = None) -> float:
"""
Calculates the profit as ratio (including fee).
:param rate: rate to compare with (optional).
If rate is not set self.close_rate will be used
:param fee: fee to use on the close rate (optional).
:param interest_rate: interest_charge for borrowing this coin (optional).
If interest_rate is not set self.interest_rate will be used
:return: profit ratio as float
"""
close_trade_value = self.calc_close_trade_value(
rate=(rate or self.close_rate),
fee=(fee or self.fee_close),
interest_rate=(interest_rate or self.interest_rate)
)
short_close_zero = (self.is_short and close_trade_value == 0.0)
long_close_zero = (not self.is_short and self.open_trade_value == 0.0)
leverage = self.leverage or 1.0
if (short_close_zero or long_close_zero):
return 0.0
else:
if self.is_short:
profit_ratio = (1 - (close_trade_value / self.open_trade_value)) * leverage
else:
profit_ratio = ((close_trade_value / self.open_trade_value) - 1) * leverage
return float(f"{profit_ratio:.8f}")
def recalc_trade_from_orders(self):
# We need at least 2 entry orders for averaging amounts and rates.
# TODO: this condition could probably be removed
if len(self.select_filled_orders(self.entry_side)) < 2:
self.stake_amount = self.amount * self.open_rate / self.leverage
# Just in case, still recalc open trade value
self.recalc_open_trade_value()
return
total_amount = 0.0
total_stake = 0.0
for o in self.orders:
if (o.ft_is_open or
(o.ft_order_side != self.entry_side) or
(o.status not in NON_OPEN_EXCHANGE_STATES)):
continue
tmp_amount = o.safe_amount_after_fee
tmp_price = o.average or o.price
if o.filled is not None:
tmp_amount = o.filled
if tmp_amount > 0.0 and tmp_price is not None:
total_amount += tmp_amount
total_stake += tmp_price * tmp_amount
if total_amount > 0:
# Leverage not updated, as we don't allow changing leverage through DCA at the moment.
self.open_rate = total_stake / total_amount
self.stake_amount = total_stake / (self.leverage or 1.0)
self.amount = total_amount
self.fee_open_cost = self.fee_open * self.stake_amount
self.recalc_open_trade_value()
if self.stop_loss_pct is not None and self.open_rate is not None:
self.adjust_stop_loss(self.open_rate, self.stop_loss_pct)
def select_order_by_order_id(self, order_id: str) -> Optional[Order]:
"""
Finds order object by Order id.
:param order_id: Exchange order id
"""
for o in self.orders:
if o.order_id == order_id:
return o
return None
def select_order(
self, order_side: str = None, is_open: Optional[bool] = None) -> Optional[Order]:
"""
Finds latest order for this orderside and status
:param order_side: ft_order_side of the order (either 'buy', 'sell' or 'stoploss')
:param is_open: Only search for open orders?
:return: latest Order object if it exists, else None
"""
orders = self.orders
if order_side:
orders = [o for o in self.orders if o.ft_order_side == order_side]
if is_open is not None:
orders = [o for o in orders if o.ft_is_open == is_open]
if len(orders) > 0:
return orders[-1]
else:
return None
def select_filled_orders(self, order_side: Optional[str] = None) -> List['Order']:
"""
Finds filled orders for this orderside.
:param order_side: Side of the order (either 'buy', 'sell', or None)
:return: array of Order objects
"""
return [o for o in self.orders if ((o.ft_order_side == order_side) or (order_side is None))
and o.ft_is_open is False and
(o.filled or 0) > 0 and
o.status in NON_OPEN_EXCHANGE_STATES]
@property
def nr_of_successful_entries(self) -> int:
"""
Helper function to count the number of entry orders that have been filled.
:return: int count of entry orders that have been filled for this trade.
"""
return len(self.select_filled_orders(self.entry_side))
@property
def nr_of_successful_exits(self) -> int:
"""
Helper function to count the number of exit orders that have been filled.
:return: int count of exit orders that have been filled for this trade.
"""
return len(self.select_filled_orders(self.exit_side))
@property
def nr_of_successful_buys(self) -> int:
"""
Helper function to count the number of buy orders that have been filled.
WARNING: Please use nr_of_successful_entries for short support.
:return: int count of buy orders that have been filled for this trade.
"""
return len(self.select_filled_orders('buy'))
@property
def nr_of_successful_sells(self) -> int:
"""
Helper function to count the number of sell orders that have been filled.
WARNING: Please use nr_of_successful_exits for short support.
:return: int count of sell orders that have been filled for this trade.
"""
return len(self.select_filled_orders('sell'))
@property
def sell_reason(self) -> str:
""" DEPRECATED! Please use exit_reason instead."""
return self.exit_reason
@staticmethod
def get_trades_proxy(*, pair: str = None, is_open: bool = None,
open_date: datetime = None, close_date: datetime = None,
) -> List['LocalTrade']:
"""
Helper function to query Trades.
Returns a List of trades, filtered on the parameters given.
In live mode, converts the filter to a database query and returns all rows
In Backtest mode, uses filters on Trade.trades to get the result.
:return: unsorted List[Trade]
"""
# Offline mode - without database
if is_open is not None:
if is_open:
sel_trades = LocalTrade.trades_open
else:
sel_trades = LocalTrade.trades
else:
# Not used during backtesting, but might be used by a strategy
sel_trades = list(LocalTrade.trades + LocalTrade.trades_open)
if pair:
sel_trades = [trade for trade in sel_trades if trade.pair == pair]
if open_date:
sel_trades = [trade for trade in sel_trades if trade.open_date > open_date]
if close_date:
sel_trades = [trade for trade in sel_trades if trade.close_date
and trade.close_date > close_date]
return sel_trades
@staticmethod
def close_bt_trade(trade):
LocalTrade.trades_open.remove(trade)
LocalTrade.trades.append(trade)
LocalTrade.total_profit += trade.close_profit_abs
@staticmethod
def add_bt_trade(trade):
if trade.is_open:
LocalTrade.trades_open.append(trade)
else:
LocalTrade.trades.append(trade)
@staticmethod
def get_open_trades() -> List[Any]:
"""
Query trades from persistence layer
"""
return Trade.get_trades_proxy(is_open=True)
@staticmethod
def stoploss_reinitialization(desired_stoploss):
"""
Adjust initial Stoploss to desired stoploss for all open trades.
"""
for trade in Trade.get_open_trades():
logger.info("Found open trade: %s", trade)
# skip case if trailing-stop changed the stoploss already.
if (trade.stop_loss == trade.initial_stop_loss
and trade.initial_stop_loss_pct != desired_stoploss):
# Stoploss value got changed
logger.info(f"Stoploss for {trade} needs adjustment...")
# Force reset of stoploss
trade.stop_loss = None
trade.initial_stop_loss_pct = None
trade.adjust_stop_loss(trade.open_rate, desired_stoploss)
logger.info(f"New stoploss: {trade.stop_loss}.")
class Trade(_DECL_BASE, LocalTrade):
"""
Trade database model.
Also handles updating and querying trades
Note: Fields must be aligned with LocalTrade class
"""
__tablename__ = 'trades'
use_db: bool = True
id = Column(Integer, primary_key=True)
orders = relationship("Order", order_by="Order.id", cascade="all, delete-orphan", lazy="joined")
exchange = Column(String(25), nullable=False)
pair = Column(String(25), nullable=False, index=True)
base_currency = Column(String(25), nullable=True)
stake_currency = Column(String(25), nullable=True)
is_open = Column(Boolean, nullable=False, default=True, index=True)
fee_open = Column(Float, nullable=False, default=0.0)
fee_open_cost = Column(Float, nullable=True)
fee_open_currency = Column(String(25), nullable=True)
fee_close = Column(Float, nullable=False, default=0.0)
fee_close_cost = Column(Float, nullable=True)
fee_close_currency = Column(String(25), nullable=True)
open_rate: float = Column(Float)
open_rate_requested = Column(Float)
# open_trade_value - calculated via _calc_open_trade_value
open_trade_value = Column(Float)
close_rate: Optional[float] = Column(Float)
close_rate_requested = Column(Float)
close_profit = Column(Float)
close_profit_abs = Column(Float)
stake_amount = Column(Float, nullable=False)
amount = Column(Float)
amount_requested = Column(Float)
open_date = Column(DateTime, nullable=False, default=datetime.utcnow)
close_date = Column(DateTime)
open_order_id = Column(String(255))
# absolute value of the stop loss
stop_loss = Column(Float, nullable=True, default=0.0)
# percentage value of the stop loss
stop_loss_pct = Column(Float, nullable=True)
# absolute value of the initial stop loss
initial_stop_loss = Column(Float, nullable=True, default=0.0)
# percentage value of the initial stop loss
initial_stop_loss_pct = Column(Float, nullable=True)
# stoploss order id which is on exchange
stoploss_order_id = Column(String(255), nullable=True, index=True)
# last update time of the stoploss order on exchange
stoploss_last_update = Column(DateTime, nullable=True)
# absolute value of the highest reached price
max_rate = Column(Float, nullable=True, default=0.0)
# Lowest price reached
min_rate = Column(Float, nullable=True)
exit_reason = Column(String(100), nullable=True)
exit_order_status = Column(String(100), nullable=True)
strategy = Column(String(100), nullable=True)
enter_tag = Column(String(100), nullable=True)
timeframe = Column(Integer, nullable=True)
trading_mode = Column(Enum(TradingMode), nullable=True)
# Leverage trading properties
leverage = Column(Float, nullable=True, default=1.0)
is_short = Column(Boolean, nullable=False, default=False)
liquidation_price = Column(Float, nullable=True)
# Margin Trading Properties
interest_rate = Column(Float, nullable=False, default=0.0)
# Futures properties
funding_fees = Column(Float, nullable=True, default=None)
def __init__(self, **kwargs):
super().__init__(**kwargs)
self.recalc_open_trade_value()
def delete(self) -> None:
for order in self.orders:
Order.query.session.delete(order)
Trade.query.session.delete(self)
Trade.commit()
@staticmethod
def commit():
Trade.query.session.commit()
@staticmethod
def get_trades_proxy(*, pair: str = None, is_open: bool = None,
open_date: datetime = None, close_date: datetime = None,
) -> List['LocalTrade']:
"""
Helper function to query Trades.j
Returns a List of trades, filtered on the parameters given.
In live mode, converts the filter to a database query and returns all rows
In Backtest mode, uses filters on Trade.trades to get the result.
:return: unsorted List[Trade]
"""
if Trade.use_db:
trade_filter = []
if pair:
trade_filter.append(Trade.pair == pair)
if open_date:
trade_filter.append(Trade.open_date > open_date)
if close_date:
trade_filter.append(Trade.close_date > close_date)
if is_open is not None:
trade_filter.append(Trade.is_open.is_(is_open))
return Trade.get_trades(trade_filter).all()
else:
return LocalTrade.get_trades_proxy(
pair=pair, is_open=is_open,
open_date=open_date,
close_date=close_date
)
@staticmethod
def get_trades(trade_filter=None) -> Query:
"""
Helper function to query Trades using filters.
NOTE: Not supported in Backtesting.
:param trade_filter: Optional filter to apply to trades
Can be either a Filter object, or a List of filters
e.g. `(trade_filter=[Trade.id == trade_id, Trade.is_open.is_(True),])`
e.g. `(trade_filter=Trade.id == trade_id)`
:return: unsorted query object
"""
if not Trade.use_db:
raise NotImplementedError('`Trade.get_trades()` not supported in backtesting mode.')
if trade_filter is not None:
if not isinstance(trade_filter, list):
trade_filter = [trade_filter]
return Trade.query.filter(*trade_filter)
else:
return Trade.query
@staticmethod
def get_open_order_trades() -> List['Trade']:
"""
Returns all open trades
NOTE: Not supported in Backtesting.
"""
return Trade.get_trades(Trade.open_order_id.isnot(None)).all()
@staticmethod
def get_open_trades_without_assigned_fees():
"""
Returns all open trades which don't have open fees set correctly
NOTE: Not supported in Backtesting.
"""
return Trade.get_trades([Trade.fee_open_currency.is_(None),
Trade.orders.any(),
Trade.is_open.is_(True),
]).all()
@staticmethod
def get_closed_trades_without_assigned_fees():
"""
Returns all closed trades which don't have fees set correctly
NOTE: Not supported in Backtesting.
"""
return Trade.get_trades([Trade.fee_close_currency.is_(None),
Trade.orders.any(),
Trade.is_open.is_(False),
]).all()
@staticmethod
def get_total_closed_profit() -> float:
"""
Retrieves total realized profit
"""
if Trade.use_db:
total_profit = Trade.query.with_entities(
func.sum(Trade.close_profit_abs)).filter(Trade.is_open.is_(False)).scalar()
else:
total_profit = sum(
t.close_profit_abs for t in LocalTrade.get_trades_proxy(is_open=False))
return total_profit or 0
@staticmethod
def total_open_trades_stakes() -> float:
"""
Calculates total invested amount in open trades
in stake currency
"""
if Trade.use_db:
total_open_stake_amount = Trade.query.with_entities(
func.sum(Trade.stake_amount)).filter(Trade.is_open.is_(True)).scalar()
else:
total_open_stake_amount = sum(
t.stake_amount for t in LocalTrade.get_trades_proxy(is_open=True))
return total_open_stake_amount or 0
@staticmethod
def get_overall_performance(minutes=None) -> List[Dict[str, Any]]:
"""
Returns List of dicts containing all Trades, including profit and trade count
NOTE: Not supported in Backtesting.
"""
filters = [Trade.is_open.is_(False)]
if minutes:
start_date = datetime.now(timezone.utc) - timedelta(minutes=minutes)
filters.append(Trade.close_date >= start_date)
pair_rates = Trade.query.with_entities(
Trade.pair,
func.sum(Trade.close_profit).label('profit_sum'),
func.sum(Trade.close_profit_abs).label('profit_sum_abs'),
func.count(Trade.pair).label('count')
).filter(*filters)\
.group_by(Trade.pair) \
.order_by(desc('profit_sum_abs')) \
.all()
return [
{
'pair': pair,
'profit_ratio': profit,
'profit': round(profit * 100, 2), # Compatibility mode
'profit_pct': round(profit * 100, 2),
'profit_abs': profit_abs,
'count': count
}
for pair, profit, profit_abs, count in pair_rates
]
@staticmethod
def get_enter_tag_performance(pair: Optional[str]) -> List[Dict[str, Any]]:
"""
Returns List of dicts containing all Trades, based on buy tag performance
Can either be average for all pairs or a specific pair provided
NOTE: Not supported in Backtesting.
"""
filters = [Trade.is_open.is_(False)]
if(pair is not None):
filters.append(Trade.pair == pair)
enter_tag_perf = Trade.query.with_entities(
Trade.enter_tag,
func.sum(Trade.close_profit).label('profit_sum'),
func.sum(Trade.close_profit_abs).label('profit_sum_abs'),
func.count(Trade.pair).label('count')
).filter(*filters)\
.group_by(Trade.enter_tag) \
.order_by(desc('profit_sum_abs')) \
.all()
return [
{
'enter_tag': enter_tag if enter_tag is not None else "Other",
'profit_ratio': profit,
'profit_pct': round(profit * 100, 2),
'profit_abs': profit_abs,
'count': count
}
for enter_tag, profit, profit_abs, count in enter_tag_perf
]
@staticmethod
def get_exit_reason_performance(pair: Optional[str]) -> List[Dict[str, Any]]:
"""
Returns List of dicts containing all Trades, based on exit reason performance
Can either be average for all pairs or a specific pair provided
NOTE: Not supported in Backtesting.
"""
filters = [Trade.is_open.is_(False)]
if(pair is not None):
filters.append(Trade.pair == pair)
sell_tag_perf = Trade.query.with_entities(
Trade.exit_reason,
func.sum(Trade.close_profit).label('profit_sum'),
func.sum(Trade.close_profit_abs).label('profit_sum_abs'),
func.count(Trade.pair).label('count')
).filter(*filters)\
.group_by(Trade.exit_reason) \
.order_by(desc('profit_sum_abs')) \
.all()
return [
{
'exit_reason': exit_reason if exit_reason is not None else "Other",
'profit_ratio': profit,
'profit_pct': round(profit * 100, 2),
'profit_abs': profit_abs,
'count': count
}
for exit_reason, profit, profit_abs, count in sell_tag_perf
]
@staticmethod
def get_mix_tag_performance(pair: Optional[str]) -> List[Dict[str, Any]]:
"""
Returns List of dicts containing all Trades, based on entry_tag + exit_reason performance
Can either be average for all pairs or a specific pair provided
NOTE: Not supported in Backtesting.
"""
filters = [Trade.is_open.is_(False)]
if(pair is not None):
filters.append(Trade.pair == pair)
mix_tag_perf = Trade.query.with_entities(
Trade.id,
Trade.enter_tag,
Trade.exit_reason,
func.sum(Trade.close_profit).label('profit_sum'),
func.sum(Trade.close_profit_abs).label('profit_sum_abs'),
func.count(Trade.pair).label('count')
).filter(*filters)\
.group_by(Trade.id) \
.order_by(desc('profit_sum_abs')) \
.all()
return_list: List[Dict] = []
for id, enter_tag, exit_reason, profit, profit_abs, count in mix_tag_perf:
enter_tag = enter_tag if enter_tag is not None else "Other"
exit_reason = exit_reason if exit_reason is not None else "Other"
if(exit_reason is not None and enter_tag is not None):
mix_tag = enter_tag + " " + exit_reason
i = 0
if not any(item["mix_tag"] == mix_tag for item in return_list):
return_list.append({'mix_tag': mix_tag,
'profit': profit,
'profit_pct': round(profit * 100, 2),
'profit_abs': profit_abs,
'count': count})
else:
while i < len(return_list):
if return_list[i]["mix_tag"] == mix_tag:
return_list[i] = {
'mix_tag': mix_tag,
'profit': profit + return_list[i]["profit"],
'profit_pct': round(profit + return_list[i]["profit"] * 100, 2),
'profit_abs': profit_abs + return_list[i]["profit_abs"],
'count': 1 + return_list[i]["count"]}
i += 1
return return_list
@staticmethod
def get_best_pair(start_date: datetime = datetime.fromtimestamp(0)):
"""
Get best pair with closed trade.
NOTE: Not supported in Backtesting.
:returns: Tuple containing (pair, profit_sum)
"""
best_pair = Trade.query.with_entities(
Trade.pair, func.sum(Trade.close_profit).label('profit_sum')
).filter(Trade.is_open.is_(False) & (Trade.close_date >= start_date)) \
.group_by(Trade.pair) \
.order_by(desc('profit_sum')).first()
return best_pair