PT, you can't just throw out a number that doesn't fit your ideal model. That's why you back-test. You are trying to test how the model performs during real world market conditions. The ideal model will have you on the sidelines prior to such an event.
Maybe combining something with the cumulative A/D line on the NYSE could help, as that was flagging the 1987 drop from what I've read.
Market Data 
Markets 
AI
