Please review my response to this post here to give you a better idea of why even larger differences will arise between the NSE data and our data.
If the NSE IV data is giving better correlation results despite it…
- Being a less accurate model for implied volatility
- Displaying stale data for non-ATM strikes near expiry
…then I would quadruple check that there isn’t look-ahead bias (point #2) or spurious correlation.