The micro physics of fat tail is identified to non-Markov statistics. We consider a simple history dependent random walk model and derive corresponding non-Markov distributions with fat tail. Some analytic properties are investigated and new option pricing methods that count more volatility risk are developed.