Quote:
Originally Posted by Elrazor
I don't think you can start picking arbitrary events in games and say they invalidate xG. Sendings off change games; soft penalties change games; key injuries during play change games; low xG goals change games.
So imo you can't just say "we can only count xG to here, then something something it doesn't count". The whole point of taking a statistical approach is to rule narratives out, and you just have to live with the deviation that exists in your numbers.
Why yes, yes you can. It's why hockey uses full strength close game state fenwick and corsi stats as the best indication for who the top teams are. It's why at the beginning of the season I was super skeptical of Swansea's numbers as they had played multiple games with a man up (5 games in they were like top5 in xgd).
It's why penalties and own goals are denoted in parenthesis on caleys and other xG guys maps. Red cards are pretty close to random (unless Mike dean is involved) so using half a game with a team a man down is a very bad way to compare relative strength.
The argument you should be making is that yes the red card grossly skewed things in spurs favor but the game state of -1 for them would have led to increased numbers anyway.
American soccer analysis even does exactly what I'm saying in their work in mls using even game state 11v11 stats for their match projections .