Extreme return values (billions %) in alternating long/short position backtests

[myuanz]

Just run this:

import vectorbt as vbt
import pandas as pd

price = vbt.YFData.download(
    "BTC-USD", start="2024-03-01 UTC", end="2024-07-01 UTC"
).get("Close")
size = [1.0, -1.0] * (len(price) // 2)

pf = vbt.Portfolio.from_orders(
    close=price, 
    size=size,
    # lock_cash=True,
)

pf.plot().show()

And if enable local_cache, the result will be normal:

But the sequences $[1, -1, 1, -1, ...,, 1, -1]$ do not lead to oversold.

---

I use vetcotbt 0.26.2 and tested on Windows10 and Manjaro