theres also a contributor called "p-factor" which is the dissymetry of angle of attack of the prop blades. Imagine a plane climbing at an angle of attack of maybe plus 5 degrees (not 5 degrees nose up in relation to the ground, but 5 degrees positive angle of attack). Assuming a "normal" rotation prop, the descending blade of the prop on the right side of the model has a higher angle of attack compared to the ascending blade on the left side of the model, and hence it creates more drag,..thus rolling the model even more to the left. This is more pronounced in constant pitching maneuvers such as a loop, where the plane is accelerated in pitch to a high angle of attack and that high angle of attack is sustained throughout the maneuver.