To save his own job as head coach, that’s why. For those of you who are Bucs fans, do you agree with prior probability that the Tampa Bay Buccaneers lost their first two games this season, not because of their quarterback, but because of bad coaching decisions made by Schiano late in both games? Also, consider the following excerpt from Bill Barnwell’s insightful essay in Grantland:
By replacing Freeman now, Schiano basically buys himself time to get through the bye with the implied argument of “You can’t fire me until we see whether the rookie can solve our problems at quarterback!” It’s not scapegoating Freeman, since he hasn’t played well and probably deserves to be benched, but the timing is pretty curious, especially considering Schiano was calling Freeman his starter several days ago.
In addition, from prior probability‘s perspective, this conflict between the head coach and the quarterback is puzzling and important because, in theory, since both actors share the same goal (they want their team to win games and reach the playoffs), they should be able to cooperate with each other to achieve this common goal. But in reality, we observe a deep conflict of interest between both men (similar to the conflict that developed between Tim Tebow and Coach Rex Ryan of the N.Y. Jets last season). This state of affairs thus poses a fascinating and difficult research question: why do we see conflict where we would expect to find cooperation? What do you think?
prior probability 
Absolute power corrupts absolutely – Schiano is not looking for another chief in Freeman. Schiano is the Chief and expects Freeman to be a good little Indian. Unfortunately, maybe Freeman was expecting shared authority i.e. Co-Chief. “No no no Not in my house” as Mutumbo would say.
That is a good point: the Quarterback is expected to lead his team, and this puts the QB in tension with the head coach, who has final authority over the team … but my research question still stands: why are most QB-coach relationships mostly cooperative?