chickenbaltipie
Goalkeeper
Sunderland's Results this Season based on xG:
Sunderland 1 - 2 Charlton (L)
Luton 1 - 1 Sunderland (D)
Sunderland 3 - 1 Scunthorpe (W)
Gillingham 2 - 1 Sunderland (L)
Wimbledon 2 - 1 Sunderland (L)
Sunderland 1 - 1 Oxford (D)
Sunderland 2 - 2 Fleetwood (D)
Burton 2 - 1 Sunderland (L)
Sunderland 2 - 1 Rochdale (W)
Coventry 1 - 2 Sunderland (W)
Bradford 2 - 1 Sunderland (L)
Sunderland 2 - 1 Peterborough (W)
Shrewsbury 2 - 0 Sunderland (L)
Doncaster 2 - 2 Sunderland (D)
Sunderland 1 - 2 Southend (L)
Plymouth 2 - 1 Sunderland (L)
W: 4 D: 4 L: 8 F: 22 A: 26 GD: -4 Pts: 16
(We'd be just inside the relegation zone)
The model got the score correct 3 times and the result correct a further 3 times. In all, the model has a 28% accuracy after 16 games.
The xG model assumes that a player has a certain percentage chance of scoring a goal for any given opportunity. This is based on retrospective goal data gleaned form all levels of football. So if a player goes 1-on-1 with the keeper, the model predicts the same likelihood of a goal, whether it's Sergio Aguero, or Gammy-Leg Bob playing for the Dog 'n Duck. The rationale is that the defence and goalkeeper are appropriately skilled for the ability of the striker. So Sergio is facing Jordan Pickford, which Bob is facing One-eyed Pete from the King's Arms, therefore there's parity in the Att. v Def. ability, and the percentage chance of scoring is the same. Now obviously, if the King's Arms face Man City, the model will go to ratshit, because you're not dealing with parity of ability, and the likelihood of Aguero knocking one passed One-eyed Pete is considerably higher than for Bob.
So if Sunderland have notably better players, than League One average, it knacks the model up because you don't have parity of ability. The author dismissed this quite casually, suggesting (in his limited experience) he'd seen nothing to suggest that Sunderland have demonstrably better players than the rest of the league. So he's backing the model, and looking for other reasons to write-off the wild discrepancy in actual performance. He's basically settled on "blind luck" as an explanation, which I think is what has rattled people.
But... Sunderland's matchday squad of 18 players on Saturday contained 8 full senior internationals and a further 4 players who've been capped internationally at youth level. Furthmore, Cattermole, Watmore, Oviedo and Love weren't in the squad, and all have international caps at either youth or senior level. No other team in League One can claim anything close to this calibre. Opposition fans constantly whinge about our "big budget" and parachute payments. Why? Because, in their view, it gives us disproportionate power to sign / retain the highest quality players.
There is SOMETHING in the stats which I think most Sunderland would agree with. We've yet to really dominate a game and give a team a proper good hiding. But when a model suggests we should have lost 2-1 at home to Southend, this is when the people on the ground watching the games can rightly scoff at the Excel-jockeys.
Sunderland 1 - 2 Charlton (L)
Luton 1 - 1 Sunderland (D)
Sunderland 3 - 1 Scunthorpe (W)
Gillingham 2 - 1 Sunderland (L)
Wimbledon 2 - 1 Sunderland (L)
Sunderland 1 - 1 Oxford (D)
Sunderland 2 - 2 Fleetwood (D)
Burton 2 - 1 Sunderland (L)
Sunderland 2 - 1 Rochdale (W)
Coventry 1 - 2 Sunderland (W)
Bradford 2 - 1 Sunderland (L)
Sunderland 2 - 1 Peterborough (W)
Shrewsbury 2 - 0 Sunderland (L)
Doncaster 2 - 2 Sunderland (D)
Sunderland 1 - 2 Southend (L)
Plymouth 2 - 1 Sunderland (L)
W: 4 D: 4 L: 8 F: 22 A: 26 GD: -4 Pts: 16
(We'd be just inside the relegation zone)
The model got the score correct 3 times and the result correct a further 3 times. In all, the model has a 28% accuracy after 16 games.
This was my point to the author. Champ's game engine is based on individual player ability (albeit, based on the subject opinion of the game's scouts and researchers).Is it a better or worse predictor than just running the games through Champ Manager by the way?
The xG model assumes that a player has a certain percentage chance of scoring a goal for any given opportunity. This is based on retrospective goal data gleaned form all levels of football. So if a player goes 1-on-1 with the keeper, the model predicts the same likelihood of a goal, whether it's Sergio Aguero, or Gammy-Leg Bob playing for the Dog 'n Duck. The rationale is that the defence and goalkeeper are appropriately skilled for the ability of the striker. So Sergio is facing Jordan Pickford, which Bob is facing One-eyed Pete from the King's Arms, therefore there's parity in the Att. v Def. ability, and the percentage chance of scoring is the same. Now obviously, if the King's Arms face Man City, the model will go to ratshit, because you're not dealing with parity of ability, and the likelihood of Aguero knocking one passed One-eyed Pete is considerably higher than for Bob.
So if Sunderland have notably better players, than League One average, it knacks the model up because you don't have parity of ability. The author dismissed this quite casually, suggesting (in his limited experience) he'd seen nothing to suggest that Sunderland have demonstrably better players than the rest of the league. So he's backing the model, and looking for other reasons to write-off the wild discrepancy in actual performance. He's basically settled on "blind luck" as an explanation, which I think is what has rattled people.
But... Sunderland's matchday squad of 18 players on Saturday contained 8 full senior internationals and a further 4 players who've been capped internationally at youth level. Furthmore, Cattermole, Watmore, Oviedo and Love weren't in the squad, and all have international caps at either youth or senior level. No other team in League One can claim anything close to this calibre. Opposition fans constantly whinge about our "big budget" and parachute payments. Why? Because, in their view, it gives us disproportionate power to sign / retain the highest quality players.
There is SOMETHING in the stats which I think most Sunderland would agree with. We've yet to really dominate a game and give a team a proper good hiding. But when a model suggests we should have lost 2-1 at home to Southend, this is when the people on the ground watching the games can rightly scoff at the Excel-jockeys.