I'm tweaking my earnings momentum model and have found earnings is based on about 5 months of data leading up to earnings ( my theory is that "historical data determines the future".
So i ran my earnings model after AMZN miss tonight and it timed really well with the miss and showed a very strong negative momentum trend into late January which i interpret as the model predicting the earnings miss. I feel it's getting very close to being effective at predicting pricing movement at earnings events.