NEIL ATKINSON tries to keep the peace as John Gibbons, Jay McKenna and Neil Docking go head to head on three things they’d love to ban.
Up for the ban bin this week:
– Merry-go-round managers.
– Which club would you ban from the Premier League.
– Blag football statistic.
Enjoyed this one again, though felt John was missing the point with xG – he is right about a lot of things about it though. People are dead moody with it. Basically people that tweet figures for it each have their own model, and to understand those a little further you’ll need a stats course under your belt, but it evaluates the quality of a chance; shot or non-shot, based on variables in the model. The models uses sample data as all do, which is one of the reasons for discrepancies between xG values from different peoples models – the other reason for this is that the models use variables to examine the chance quality. So, most models use distance from goal, most use type of finish (head, weak foot, strong foot) and then from there it goes a bit wild. Some use how many defenders are behind the ball, some use the type of pass leading to the chance – all these details change the xG. You get some good models, some bad models. It essentially evaluates a bit deeper than shots on target etc…
On his other point about how it predicted Leicester would drop off and how it didn’t happen, 2 things:
Sample size. 20 games, 30 games, 50 games is sometimes not big enough for an average to even out – huge problem with loads of stats.
System. Some systems ‘cheat’ xG, which some take as model deficiencies or others, as with your liverpool example, can come down to system. So with the pressing system we use, when teams beat the press – which doesn’t happen often cos we are dead good – the chances they lead to are generally central and with few defenders behind the ball, the nature of beating a high press.
Hope this helps
It isnt absolutely magic though for sure. It essentially puts the exact number on he scores that ‘x’ times out of ten.