Yep, you are right.
I prefer to eliminate 1999 - 2002 data which gives, what I'll refer to as a premium on the NDX, an average of 20% better annual performance over the NDX.
In other words, if the NDX performed 10% during a given year, this system would expect to produce 30%. Pretty sweet.
Eliminating the data from those years is only to help me manage my expectations for what the system might prodcue going forward. It should not be taken out for studying the system's effectiveness.
Market Data 
Markets 
AI
