Liquidations
This example shows how to resample time series. It uses 5 minutes interval, in this interval we count how many 10 seconds spans had liquidations, and if the number breaks the threshold - we start and new trading session.
The liquidations.csv looks like this:
ts,trade_time,symbol,side,order_type,order_status,time_in_force,original_quantity,price,average_price,last_filled_quantity,filled_accumulated_quantity
1674374499889,1674374499884,BTCUSDT,SELL,LIMIT,FILLED,IOC,0.005,22806.20,22896.60,0.005,0.005
1674374553356,1674374553353,BTCUSDT,SELL,LIMIT,FILLED,IOC,0.009,22792.58,22881.50,0.008,0.009
1674374563396,1674374563393,BTCUSDT,SELL,LIMIT,FILLED,IOC,0.569,22785.32,22874.30,0.016,0.569
from cipher import Cipher, Session, Strategy
class LiquidationsStrategy(Strategy):
def __init__(self):
self.latest_direction = None
def compose(self):
df = self.datas[0]
liquidations_df = self.datas[1][["side", "filled_accumulated_quantity"]]
df["buys"] = self.count_10_seconds_samples(
liquidations_df[liquidations_df["side"] == "BUY"]
)
df["sells"] = self.count_10_seconds_samples(
liquidations_df[liquidations_df["side"] == "SELL"]
)
df["long"] = df["sells"] > 10
df["short"] = df["buys"] > 10
df["entry"] = df["long"] | df["short"]
return df
def count_10_seconds_samples(self, source):
"""How many 10 seconds samples have liquidations."""
df = source.resample("10s").agg({"filled_accumulated_quantity": "sum"})
df["exist"] = df["filled_accumulated_quantity"] > 0
df["exist"] = df["exist"].replace({True: 1, False: 0, None: 0})
return df.resample("5min").agg({"exist": "sum"})["exist"]
def on_entry(self, row: dict, session: Session):
if row["long"]:
if self.latest_direction != "long":
session.position += "0.01"
self.latest_direction = "long"
else:
if self.latest_direction != "short":
session.position -= "0.01"
self.latest_direction = "short"
def on_long(self, row: dict, session: Session):
session.position = 0
def on_short(self, row: dict, session: Session):
session.position = 0
def main():
cipher = Cipher()
cipher.add_source("binance_futures_ohlc", symbol="BTCUSDT", interval="5m")
cipher.add_source("csv_file", path="data/liquidations.csv")
cipher.set_strategy(LiquidationsStrategy())
cipher.run(start_ts="2023-01-20", stop_ts="2023-01-28")
cipher.set_commission("0.00075")
print(cipher.stats)
cipher.plot(rows=[["ohlc", "sessions"], ["signals"], ["balance"]], limit=1700)
if __name__ == "__main__":
main()
---------------- ------------------- -----
start 2023-01-20 00:00
stop 2023-01-27 23:55
period 7d 23h 55m
trades 19
longs 9 47.4%
shorts 10 52.6%
period median 4h 15m
period max 15h 15m
success 11 57.9%
success median 1.40603500
success max 3.70829250
success row 5
failure 8 42.1%
failure median 2.88356375
failure max 5.30065500
failure row 3
spf 1.375
pnl -1.96706250
volume 0.38
commission 21.7160625
exposed period 3d 17h 22m 40s 46.0%
balance min -12.904080000000022
balance max 3.2301750000000027
balance drawdown 12.904080000000022
romad -0.1524372524038906
---------------- ------------------- -----