Xg....Question

I’ve seen us play at Swillsborough enough times to know with certainty that the best team doesn’t always score the most goals.

 

I think the first one you’re expecting him to save almost every time, he gets wrong footed which is unfortunate but I’d say it’s a definite mistake.

The third i think probably has a high expected save % statistically because of the deflection making the speed so much lower, but that’s one where the stats don’t really work because there are extenuating circumstances.

The rest I dont think you can expect him to save you’re right but I do think on average you’d say a keeper should be saving 1 or 2 of those and most games they probably would.

 

Even if we take your view of it, the best he could do is save 2 of them. But according to the Xg the "keeper is at fault for 3.87 goals statistically". There's no way on earth he's saving 5 of those shots. The Xg analysis is just wrong. And this isn't a case of things averaging out, this is actual analysis of the goals in that game. I'm now very sceptical of the figures it produces.

 

I mean i said the first he should be saving and explained why i think the third probably isn’t accurate in the stats. I then think usually you’d expect a keeper to save 2 of the other 5 on average. They were good goals, but not so incredible that they’re un-saveable. Id expect a competent keeper to save shots into the far bottom corner more often than not, I think.

These stats aren’t meant to be used for single games really though. They’re quite useful when used over a season or multiple seasons to see who regularly over-performs or under-performs.

 

Just checked the stats for the individual shots.

Goal 1: 0.1 xG and 0.11 xGOT
Goal 2: 0.27 xG and 0.61 xGOT
Goal 3: 0.1 xG 0.63 xGOT
Goal 4: 0.28 xG 0.85 xGOT
Goal 5: 0.04 xG 0.39 xGOT
Goal 6: 0.04 xG 0.08 xGOT
Goal 7: 0.01 xG 0.42 xGOT

meaning judging by the stats you’d expect a keeper to save goals 1,5,6,7 more often than not. So i was wrong about goal 3, i really thought that the deflection would have messed up the calculation, but i guess not.

 

It's a nonsense

 

The idea that things average out comes from using an unbiased model over a large period, not from being right or wrong in a specific game.

The only way long term xG would not be reliable is if these situations were common due to the tactics of a team. Or if a player consistently under/over performs it. As xG is based on data from a large sample of shots and situations, it essentially compares performance to the 'average' team. So, if a team did something unusual such as Bielsa's Leeds who played football very differently from average, they'd usually concede shots after their man to man press was exposed, leading to counters and attackers reaching the goal unopposed.

When conceding a shot after a counter attack, xG doesn’t capture it well because it doesn’t consider the speed of the attack. The shooter may more time with the ball to set up the chance, possibly with a teammate well positioned across the box for a pass, which the goalkeeper must account for. Additionally, defenders might be scrambling to block a cross or shot, unlike the average scenario where a defender is standing upright, covering a larger area.

Which is exactly why it should be used only as a tool and not a fact. It provides an insight but always requires context and critical thinking.

 

Wasn't having a go at you BTW. But one of those goals that the stats reckon the keeper should have saved is literally unsavable. Perfect shot clipping the inside post and the side netting.

I get the premise but the analysis is all wrong and if you don't analyse well it doesn't even itself out, it just get skewed.

 

I assume you mean the last one? I do agree that one is the best shot, but that’s also reflected in the stats. It’s a 0.01 xG chance that becomes a 0.42 xGOT shot.

I think it probably wasn’t quite as far in the corner as it looks on the video. It’s a bad angle for it really. I do agree that’s probably the one that looks the worst from the stat though.

I’m very sure it’ll average out over a longer timeframe though.