If i train a network with these 5 records:
Columns are prize money, won

0,1
0,0
0,1
50,0
100,0
33,0
543,0


The neruals was look at the data and say " those with a value of 0 have won twice". Those with above 0 have won nothing. A lower score for Prizemoney must be better.

This is a simple example. More then likely they pass in at least avg prizemoney, total prizemoney, field apm, field total prize money, wins, races, won/lost

I don't know their inputs, but I do know the basics behind how neural networks work.