I was doing a little data analysis today and I think I found a data problem with vix coming from CBOE extension.
My three data sources were.
Tradestation data pulled into a chart then downloaded with copy all from WL
CBOE data pulled into a chart then downloaded with copy all from WL
CBOE.COM direct download
vix.x (TradeStation) IS the CBOE file. It's a full-OHLC mirror of the data you just pasted. Check your own download: 6/23/16 reads 19.54, 19.79, 17.25, 17.25 — that's exactly vix.x. And 3/16/2020 closes at 82.69, the real all-time high — exactly vix.x.
vix(WL-CBOE).csv is NOT the CBOE close. It's the CBOE open column, mislabeled as the close. I tested it against your files: in the entire pre-2020 block, 99.8% of vix.csv's "close" values equal CBOE's OPEN price (not the close). The smoking guns line up perfectly with your CBOE download:
6/23/16 — CBOE open 19.54, CBOE close 17.25. vix(WL-CBOE).csv shows 19.54. (it took the open)
2/5/18 — CBOE open 18.44, CBOE close 37.32. vix(WL-CBOE).csv shows 18.44.
3/16/20 — CBOE open 57.83, CBOE close 82.69. vix(WL-CBOE).csv shows 57.83.
That single fact explains everything we saw earlier. On calm days the VIX open and close are close together, so vix(WL-CBOE).csv looked almost right. On violent days (Feb '18, COVID) the open-to-close gap is enormous, so vix.csv was off by 10–25 points.
The file you thought was straight-from-CBOE (vix(WL-CBOE).csv). It appears someone built its history from CBOE's OPEN field and called it the close.
The file you were suspicious of (vix.x, TradeStation) is the clean, exact CBOE close.
My three data sources were.
Tradestation data pulled into a chart then downloaded with copy all from WL
CBOE data pulled into a chart then downloaded with copy all from WL
CBOE.COM direct download
vix.x (TradeStation) IS the CBOE file. It's a full-OHLC mirror of the data you just pasted. Check your own download: 6/23/16 reads 19.54, 19.79, 17.25, 17.25 — that's exactly vix.x. And 3/16/2020 closes at 82.69, the real all-time high — exactly vix.x.
vix(WL-CBOE).csv is NOT the CBOE close. It's the CBOE open column, mislabeled as the close. I tested it against your files: in the entire pre-2020 block, 99.8% of vix.csv's "close" values equal CBOE's OPEN price (not the close). The smoking guns line up perfectly with your CBOE download:
6/23/16 — CBOE open 19.54, CBOE close 17.25. vix(WL-CBOE).csv shows 19.54. (it took the open)
2/5/18 — CBOE open 18.44, CBOE close 37.32. vix(WL-CBOE).csv shows 18.44.
3/16/20 — CBOE open 57.83, CBOE close 82.69. vix(WL-CBOE).csv shows 57.83.
That single fact explains everything we saw earlier. On calm days the VIX open and close are close together, so vix(WL-CBOE).csv looked almost right. On violent days (Feb '18, COVID) the open-to-close gap is enormous, so vix.csv was off by 10–25 points.
The file you thought was straight-from-CBOE (vix(WL-CBOE).csv). It appears someone built its history from CBOE's OPEN field and called it the close.
The file you were suspicious of (vix.x, TradeStation) is the clean, exact CBOE close.
Rename
Thanks. I'll look into it.
It looks OK to me, just eyeballing a chart of VIX sources from our CBOE extension and Tradier.
Glitch, the problem is the older initial history that the provider has stored as as a starting point that's the problem. I'm working on it.
Your Response
Post
Edit Post
Login is required