I just posted the study. The basic conclusion is that when all 3 of my RAIM models have active scalps, then the likelihood of getting creamed are remote.
The data as presented on the blog is skewed, I haven't figured out how to keep stuff in its own columns.
In the 2.7 year backtest, the max drawdown is only 1.17%. Although there are some intrascalp dates where drawdown was as low as 4.71%.
Market Data 
Markets 
AI
