You missed the point of my post.
The outcome logic of collisions in NCAA/Madden is based upon their animations and ratings. [and someone please correct me if they have seen evidence otherwise]
The outcome logic should be based primarily upon the physics engine.
They don't need anything additional added to the game to generate physics-based outcomes because both Madden and NCAA already have physics engines. All that is needed is to compute the outcome and then generate the appropriate animation.
Example:
Jeff Demps tries to stiff arm a huge 300+ pound defensive tackle on a run up the middle.
The current animation engine may allow this to happen if the button is pressed at the right time. There are still at least three animations which take place though. One is the actual stiff-arm by Demps, another is the tackle animation by the defensive tackle, and then a transition animation bridges to the outcome animation which is either a defensive success (successful tackle completion and failed stiff-arm) or offensive success (failed tackle and stiff-arm completion).
Based on what I'm saying, the outcome decision engine needs to use the physics engine as the primary driver. After the stiff-arm animation by Demps and the tackle animation by the defensive tackle there is a transition animation which must take place if there is any contact between the two objects.
The resulting animation should quite frankly always result in Demps being crushed in one of many possible ways by this huge defensive tackle unless Demps is flying at full speed into a defensive tackle whose momentum is already taking them backwards or to the side. Speed, momentum, size.... these values are used in the physics engine. The final outcome animation still comes from the typical EA Sports animation library. I don't necessarily care how many ways that little guy can be crushed, but only that he does actually get crushed. Your comment that the game lacks animations simply refers to how many ways a player can be tackled, how many angles are accounted for, and how many situations can be accounted for. Even with one tackle animation and one broken tackle animation in the entire game, there has to be a winner/loser in a collision.
BackBreaker relies on physics to determine how the entire transition animation is simulated through the outcome animation which is generated on-the-fly. Madden/NCAA only seem to use transition animations to cover the gap between the player's current position and the starting position of their mo-cap animations. That's why we see so much player and football warping and herky-jerky motion to get players in that "perfect" position to pull off the final animation. I'm not referring to that problem because there's no right answer. I'm simply referring to how the outcome decision gets made within the game AI.