Real tests are best way, but most times results slow and expensive. In order to "fence in" the tests samples a mathematical model could be applied.
This would reduce the propellers to be tested.
The calculation chain would be:
1-] ROV total resistance in a preset speeds interval (Rt(v) ), paying special attention on projected cruise speed.
2-] Set propeller restrictions (mainly diameter).
3-] Set blades number
4-] Set desired RPM
5-] Calculate "ideal" propeller pitch for desired RPM and Thrust (Thrust ≈ Rt(v) )
6-] Calculate "ideal" propeller Torque (Prop Torque ≈ Motor torque at desired RPM and Load(Amps) )
7-] Change ideal prop until fitting Torque and Thrust.
8-] Select closest to "ideal" comercial propeller.
9-] Calculate real propeller+motor working conditions.
10-] Repeat from point -7- until approaching the best selection.
11-] Start real tests with closer propellers to the calculated one.
This way, may be only three/four propellers would have to be tested, reducing time and money.