What is best for one set of conditions might not be the best for another. In considering configuration options in the competition context I would suggest starting with a review of:
1 - the missions to be completed in each competition. What does the ROV’s have to do, how much time is allowed, any lifting needed perhaps requiring multiple thrusters? What is the environment, operating depth and distance from the control point? Could be trade-offs here with various designs. Given the mission needs asked what can be achieved by 3 or 4 thrusters. 3 could be fine for a build with one as a spare or 4 with maximised vertical thrusts if need or 4 with simple sideward movement as a refinement of the basic 3 thruster design. Consider configurations against the mission needs.
2 - constraints on the allowed solutions e.g. some competitions mandate surface power sources and max voltage (assume 12v from your post), maximum spend, points for reuse, tether types (pair, multicore or other)
3 - whether the competition has a design process element and weighting on score and awards. This might make the effort described below very worthwhile and would in any case help to maximise the learning experience given the question of “what is best” has been asked.
4 - document existing builds and analyse pros and cons for each, try and figure out the rationale for the design (in support of 3 above as well), such as the great build from TCIII
5 - consider the teams experience, skills and time available. Go for something they have confidence in the ability to deliver in the time available, with some float on the schedule as well, and look to using that unused to make further minor improvements,
In respect of point 3) above I look to get students to try out different configurations e.g. 3 or 4 thrusters and configurations, to give different axis of movement up, down, forward, reverse and whether precision side crabbing has advantages or not on the mission, over a session or two. Try out each option and score them on a defined scale related to the mission objectives and other factors e.g. ease of learning, manoeuvring, time on mission etc, document the reason for the selected option. This is part of the development, experiment, learn and iterate to converge on a solution by consideration of alternatives. Make up a frame, set up thrusters in different configurations and say using surface power supply and switches / potentiometers rather can going straight into selecting a controller, programming etc, then look to see how the chosen configuration can be controlled.