A few observations from the stats (Stoke)

[Uganda Blade]

I take the raw stats/data from the most respected analytics companies (such as Opta) and then add in my contextual interpretation, very much from a blades fan perspective. I am unaware of anyone else doing a similar review of our games.

Is there data on power output, distance covered etc?
Analysed by first and second halves, that might confirm/refute questions over fitness.

 

[Ned the Donkey]

XG is certainly not just a subjective opinion. The calculation by each analytic company (such as Opta) uses massive datasets and machine learning models, so the probability values are derived from empirical evidence not opinion. Factors usually include shot location, angle, head/feet used, defensive pressure, and sometimes goalkeeper positioning.

But some subjectivity can be evident. Different providers (Opta, StatsBomb, etc.) use slightly different variables and weightings, and some models include factors such as goalkeeper data or shot speed, whilst others don’t. So while the underlying concept is objective, the implementation can vary.

For the Stoke game, Fotmob has the xG as 0.59 - 2.50 and those are the most widely quoted figures, but, for instance, Footy Stats has the xG as 0.99 - 1.49. That's a pretty significant difference.

It seems reasonable to conclude that Opta, Statsbomb, Understat, and the rest will have different stats (if they have stats at all for the Championship - I don't know the extent of their coverage).

I've no idea of the relative predicitive accuracy of these organisations, but I think it helps illustrate the idea that there is no such thing as objective xG, there is only xG according to...

 

[wavy-blade]

View attachment 228094

A lot of this is a hazy memory but Jay Socik explained this a few years ago, using the example above of Leon Clarke in a 1v1 with the keeper against Middlesbrough.

“He’s got to score from there.”

But if you analyse 1000s of games with 100s of players in similar situations it turns out that the situations Leon was in players only score about 40% of the time. So the xG is 0.4, depending on who you ask.

(In this case relevant factors might be there are recovering defenders and the goalkeeper is well positioned.)

This last point is important: Although xG is quoted as though it’s set in stone it is still, ultimately, an opinion.

In the above situation Company A could give an xG as 0.35 and Company B could give an xG as 0.45. I say Company A because this sort of information tends to be commercial.

The fact that we won 2-1 last night is not really a matter of opinion. The statement that the xG was 0.53 - 2.05 is, ultimately, a matter of opinion, and the details depend on who you ask.

View attachment 228096

Fuck me .....Way over my head

Games Gone

 

[Coolblade]

I am unsure I can add much more to the discussion. I agree there can be variances and have explained why.

As I’ve said, the calculation of xG is based on empirical evidence and the most respected sites have little variance. And minor variances don’t really impact on ultimate conclusions, as they are merely one stat of many, all of which need context before any fair conclusions can be drawn.

I try to use the most advanced versions available. Looking at just two you mention

Opta is one of the industry-standard providers used by broadcasters, analytics platforms, clubs, and official services like FBref. Its published xG model uses machine-learning (XGBoost) trained on hundreds of thousands of historical shots to estimate the probability that a given shot results in a goal. It includes many situational variables, as listed before.

FootyStats xG is broader, simpler, and less transparent. It displays xG figures across thousands of leagues and teams worldwide, including lower-league competitions many other providers don’t cover. However it doesn’t publicly document a detailed proprietary xG model in the same way Opta does and likely uses a simplified model or aggregates data from widely available sources instead of generating xG from raw event tracking. This often means the xG can reflect simpler factors like shot location and results rather than the full Opta contextual depth. So xG values on FootyStats can be less precise and one-dimensional compared to Opta, especially at the individual shot level.

So I tend to favour the more respected analytical sites, like Opta.

 

[ISZA ⚔️]

A few observations from the stats (sadly only watched on tv this time!):

Team Set Up: We changed formation by lining up in a 4-4-2 out of possession, this time with a double pivot of Riedewald and Soumaré screening the half-spaces. Width was again our route to goal, with more focus on the left. Burrows pushed high and aggressive, while Seriki provided the outlet on the right but less frequently than in recent matches. The front two of Campbell and Cannon stretched Stoke’s back line, supported by O’Hare in the half-spaces.

The numbers tell the story: we won the xG battle 0.53 vs 2.05, shots 7 vs 15, shots on target 3 vs 4 and touches in their box, 21 v 32. Stoke had more possessions 53.4% vs our 46.6% and more corners 6 vs 2. So, Stoke had more of the ball, but we created far more danger. We didn’t just edge them; we doubled their attacking output in key zones.

First Half: Our shape was compact and disciplined. The chances were there even if the breakthrough wasn’t. We maintained defensive structure and created opportunities, but the final ball or finish eluded us before the interval.

Second Half: Usual territory loss as Stoke pushed possession up, but our defensive structure held. Clearances, 36 vs their 23, Interceptions: 14 vs 6, Recoveries: United 61 vs Stoke 52, with good numbers in midfield third (33 for us).

Defensive: Bindon made 12 clearances, 3 interceptions, 2 aerial wins, and was great under pressure; Tanganga with 6 clearances, 2 tackles, 2 interceptions was solid reading danger. But given his attacking contribution, Burrows with 5 clearances, 5 interceptions was a stand out, a complete LB performance.

Midfield: Riedewald anchored everything: 46 passes at 95.7% accuracy, 3 tackles, 1 interception, and a goal. He screened the back four, recycled possession, and popped up at the right moment to score. Soumaré supported well with 2 key passes and important defensive contributions in midfield duels, helping us win second balls despite Stoke’s aerial advantage.

The foul count stayed low (4 each), but Stoke’s aerial edge (20 vs 14) meant we had to scrap for second balls. That’s where Soumaré and Tanganga stepped up with key duels.

Creativity: We looked organised and threatening, and the stats back it up. We had far more key passes: 11 vs 3, spread across Burrows (3), O’Hare (3), Campbell (2), Seriki (1), Soumaré (2). With crosses, we attempted 7 vs Stoke’s 25, but our accuracy was better. Burrows attempted 5 with 2 accurate; Seriki delivered 1 accurate cross. Stoke’s 25 crosses were largely speculative from wide positions; ours came from dangerous overloads and created genuine threat.

Offence: Campbell & Cannon combined for 6 shots (Campbell 4, Cannon 2), with Cannon scoring and Campbell assisting, O’Hare chipped in with 2 shots and 3 key passes, linking midfield to attack.

Our goals: one from pressure inside the six‑yard area (Riedewald), one from a corner from cross pattern (Burrows to Cannon). Our xG of 2.05 was built on purposeful box entries and varied delivery angles, with left overload, central combinations, and quick switches.

Strategic thoughts:

1. Left-Side overload this time worked well. Burrows and O’Hare gave us penetration and control down the left. As mentioned after the Wrexham game, the ability to flip our attacking focus is key, and it worked brilliantly tonight.

2. Second-Ball still needs work . Stoke’s aerial edge (20 vs 14) meant we defended deep at times. Our 36 clearances show resilience, but we need a better midfield screen to stop repeat entries. Win the scraps, and we stop the panic defending.

3. Efficiency over volume: 15 shots, 11 key passes, 2 goals from 2.05 xG shows we were reasonably clinical but could do better.

4. Discipline under pressure. The compact back four prevented the equaliser through late corner pressure. Pearson’s red at 88’ took the edge off Stoke’s final push and gave us breathing room.

5. Game management Improved, although not perfect: Our pass accuracy (76.9%) and recoveries (61) show we stayed calm under pressure despite ceding possession. Generally we defended with organisation and attacked with purpose.

Looking forward:,

1. Set Pieces: We turned a corner into a cross into the winning goal. We are presumably pre-planning one back-post overload and one near-post screen per game to raise our set-piece xG

2. Compressing the lanes at corners. When Stoke piled on the pressure between 63’-67’ and 78’-83’, our shape coped, but we invited second phases. Add a fixed exit runner (the fresh Chong/Bamford profile helps) to turn clearances into counter-attacking territory.

3. Use the 74’ triple sub as a template for better control. This time it worked better (again not perfectly) Bamford for hold-up, Chong for carry and press, Arblaster for recycle: these roles map to protect a one-goal lead away from home.

This was a performance built on defensive solidity and attacking efficiency. We didn’t dominate the ball, but we dominated the box. When Burrows and O’Hare linked with Campbell and Cannon, we looked dangerous. When Riedewald anchored, we looked secure.

Away at Stoke is never easy and this was a crucial bounce back.

Happy new year to all!

UTB

Welcome to the forum, Guilherme.