It really was a close run thing.

I have analysed the difference between my predictions and the A.I. formula.

Here are the results per division and the average difference in place per team over all 5 divisions.

Division Human A.I.
1 64 68
2 112 112
3 76 78
4 112 118
5 110 102
total 474 478
Position 3.95 3.98

So what does this mean , my predictions were out by around 3.95 per team , the A.I. by 3.98 per team. So I did slightly better.

D1 and 3 were the most accurate divisions, it was trained using D3 data.

So what to conclude and next steps!

The A.I. did really well considering. Next season I will test using the A.I. formula on the main categories only GK,DEF,MID,ATT, Top18 and AVE alongside my own. I will also tweak the A.I. formula to create a formula for each division to see if it would have done any better, if it would not have made any difference I won't worry. I will then look at introducing the depth and balance categories the A.I. suggested while keeping the point system the same for consistency.