OK, lets go back to the original poster. He bought 50 very small and very thin rectangular magnets. To my opinion this is a bad choice if you want to make a high pole number PM-generator for a squirrel cage rotor. I think you can better reduce the pole number up to eight and use 6 coils for a 3-phase winding. As he can't buy thick circular magnets, may be he can glue some of the small magnets together. Assume that you use only one armature sheet with magnets on the back side like I do for my axial flux generators. So in this case you need eight magnets. Assume you use 8 small magnets for one large combined magnet. So you need totally 48 magnets. You can't end up with a circular magnet but a square combined magnet is the next most optimum shape if you use circular coils. You have only two options to realise this, two layers with four magnets per layer or four layers with two magnets per layer. One should chose the option which results in a rectangular magnet which is closest to a square magnet as possible. A problem with gluing magnets together is that the magnets are repelling each other at the small side of the magnets, so in this direction the magnets have to be clamped together firmly.
To find the optimum winding for a certain DC loaded voltage at a certain rpm is a rather complicated procedure. I have explained this procedure in all of my reports about axial flux generators. You start with a test winding and measure the generator with this winding for different voltages. There is one voltage for which the matching with the chosen wind turbine rotor is optimal (you need to know the optimum cubic line of the windturbine and I don't know how to find this curve for a squirrel cage rotor). Next you modify the winding such that the wanted Pmech-n curve is gained for the wanted battery voltage. A problem with measuring of a generator is that it isn't enough to only measure the supplied electrical power. You should also measure the mechanical power and this requires measurement of the torque and the rpm which requires a rather sophisticated test rig. So if you can't measure the generator correctly, try and error is the only last option you have.
The idea of using eight poles and nine coils is useless if neodymium magnets are used and if stators are used with no iron in the coils. However, if only ceramic magnets are available, one sometimes uses iron cores in the coils to increase the magnetic flux. But if you use eight magnets and six coils, you get very strong preference positions every 15 degrees, so 24 preference positions per revolution. This will give starting problems for a rotor with a low starting torque coefficient and the generator may also become noisy at higher rotational speeds. But if you use eight magnets and nine coils with iron cores in the centre, you will get 72 preference positions and the torque fluctuations for 72 preference positions are much less than for 24 preference positions.