ranaroussi
bug: yf.download() of 1d EURUSD=X data has the same open and closing price
` yf.download('GBPUSD=X', interval='1d', period="10d")
[100%**] 1 of 1 completed
Open High Low Close Adj Close Volume
Date
2022-09-20 1.002959 1.005025 0.995708 1.002959 1.002959 0
2022-09-21 0.997079 0.997407 0.986914 0.997079 0.997079 0
2022-09-22 0.983072 0.990619 0.981104 0.983072 0.983072 0
2022-09-23 0.984155 0.985222 0.970393 0.984155 0.984155 0
2022-09-26 0.968992 0.969989 0.958231 0.968992 0.968992 0
2022-09-27 0.962371 0.967006 0.959453 0.962371 0.962371 0
2022-09-28 0.959619 0.968626 0.954016 0.959619 0.959619 0
2022-09-29 0.970817 0.978732 0.963828 0.970817 0.970817 0
2022-09-30 0.982956 0.984999 0.973606 0.982956 0.982956 0
2022-10-03 0.977230 0.984640 0.975515 0.982994 0.982994 0
` yfinance 0.1.74 pandas 1.5.0 Python 3.9.12
I agree that's wrong but this is what Yahoo has on their website so nothing we can do. Other than wait for Yahoo to fix, because I'm sure it was fine recently so maybe Yahoo messing around.
There are actually hundreds of candles with this issue, doesn't seem to be a temporary problem https://github.com/ranaroussi/yfinance/issues/1075#issue-1396356550
What I mean is, assuming that Yahoo was displaying correct candles until recently (which I think is true), that Yahoo has recently introduced a change or bug in their backend that either (i) mixed up the fetching of daily open/close from their database, or (ii) overwrote their database.
Just because an old candle (e.g. 2020) is being displayed today incorrect doesn't mean it was displayed incorrect in 2020.
Response from Yahoo received Thu, Oct 6 (eastern).
NOTE: This is a problem for more than just EURUSD=X, other currency pairs have incorrect 'close'/'adj close' DAILY values. They may not be exactly the same as the open but they are only a few PIPs off (which are also incorrect).
I realize this is not a yfinance problem (indirectly it is) but it was never resolved. I would leave it open so that everyone sees the issue.
Probably Yahoo never resolved because the fix is simple: close = open tomorrow. Because market is 24 hours.
import numpy as np
close = np.roll(df["Open"].to_numpy(), -1)
close[-1] = df["Close"][-1]
df["Close"] = close
I'll let someone else polish & submit a PR. Might be a PR not possible because must restrict to 24-hour markets.