I think i can add a good perspective hence here’s my take:
If we only consider $w_i$ to set priorities then among the processes with same $w_i$, all processes will be given same priority. hence it may so happen that longer jobs r executed first and we know that this is not optimum as if the shorted jobs r executed first then average waiting times are minimised. Hence we need to prioritize smaller $t_i$ among equal $w_i$ hence we get to $\frac{ w_i}{t_i }$.
Now similarly if u only take $t_i$ then among the processes with same $t_i$ all r given same priority hence it may so happen that processes with bigger weights are executed last. Now these will be multiplied with $T_i$ given in the question which depends on waiting time(refer to Ayush sir’s ans for how this is true). hence basically if the bigger wieghts are executed later (among the same $t_i$ processes) then they will have to wait more. Hence it will be worse than executing the smaller $w_i$ processes first.