We've noticed some inconsistencies in the quote data we serve from Kaiko so have taken a protective step to delete all quote data from the cloud. This will impact the fills your algorithm recieves if you're using minute data. If your algorithm was relying on poor quote data you might see a marked decline in performance.

We've sent a notice to Kaiko and will update this thread if/when the data is fixed. Otherwise for now we'll continue using purely trade data for fills. For the last 2 years the spread is small so it should not impact many users. The trade data still looks perfect and closely reflects the GDAX market.

You can model spread with the QuantConnect fill models. A slippage model which moves against you for market orders would reflect this well for small orders. See examples of the fill models for C# and Python in the documentation

Author