shots with a body in the way convert worse than xG thinks.
held distance fixed, split 39k shots by how much of the goal was blocked. on the most obstructed fifth @StatsBomb's model says 5.6%, reality is 3.9%.
if your team shoots into traffic, your xG total is flattering you.
@PSG_inside posted a Football Data Scientist vacancy this week. one of the few ligue 1 clubs with an active senior data science role in public right now.
https://t.co/SRAAybYu3h
@Arsenal is hiring an AI Research Engineer for the men's first team, focused on the application layer of first-team models. noted in this week's brief.
https://t.co/SRAAybYu3h
@karun1710 arsenal's AI Research Engineer opening for the men's first team is in this week's brief. application layer work, directly with the squad.
https://t.co/SRAAybYu3h
1/10
Mapping players across data sources is one of the classic, foundational problems in soccer analytics — and it's quietly hard. There's no unique key that joins a player across databases at 100% accuracy. Here's how to actually approach it 🧵
https://t.co/R5CAKcFeZ5
the lenovo platform co-developed with @FIFAcom is ingesting 500 data points per second from the smart ball across all 16 WC2026 venues. we wrote it up.
https://t.co/SRAAybYu3h
the Football AI Pro platform from @Lenovo is running at all 16 WC2026 venues. 500 data points per second off the smart ball, 2,000 per-match parameters. it is in this week's brief.
https://t.co/SRAAybYu3h
clean json is the easy 20 percent. coverage is the hard part. does it go past the top 5 leagues, and how granular are the events. everyone nails the endpoint, nobody has the lower divisions.
@footio https://t.co/BH9mGctIwt
yeah. the aggregates never show the ball lost under pressure, or the drift out of position. that stuff only shows up in the full 90s, never in the totals.
@FPL_Corazon https://t.co/rb72yFR3ro
There are many football data accounts that are praising Brentford for the Sangaré deal as the best deal of the summer, but I feel like those accounts never actually watched full 90s of him. Maybe just comps. His data is impressive but, watching him, I wasn't awed like they are.
£100m buys the highlight reel. his output lived in one hot stretch, the underlying numbers flatten out either side of it. getting into the position was never the issue, the finishing is.
@chizzygold67 https://t.co/IdC5VeSpd8
🚨 𝗗𝗘𝗕𝗔𝗧𝗘:Bradley Barcola might cost Liverpool or Arsenal over £100m. But it would be a coup.
Will Bradley Barcola be playing in the Premier League in the coming season?
Barcola is blessed with a broad range of attacking qualities, but it is his blistering pace that is most striking.
According to tracking data provider SkillCorner, Barcola ranked fourth across Europe’s top five leagues for ‘Peak Sprint Velocity’ last season, which measures the 99th percentile of his physical activities above 15km/h, clocking 32.7km/h.For a talent of his stature, football fans do not get to see enough of Bradley Barcola.
The 23-year-old forward would walk into most starting line-ups, but faces fierce competition for minutes at Paris Saint-Germain and with the French national team, possibly the most talented club and country pairing in world football.
@vshufinskiy name + dob + club-season gets you most of the way. it's the tail that hurts, loans, youth players, transliterated names. we lean on jersey number + team from video as a tiebreak. what's your fallback for the last few percent?
@Abhishek829300 fcpython is a great on-ramp. once you've got the event frames down, the fun part is everything between events, the off-ball runs the feed never logs. good start.
@TouchlineLabs the read where xG describes one shot. it doesn't. it's a rate over a pile of them. a lone 0.9 tap-in tells you basically nothing, and every match-report villain arc is built on forgetting that.
@AlexpropsHQ rolling xG gap is real, but raw xG lumps a 6-yard tap in with a 25-yard screamer. weight for placement and pressure and the gap gets a lot more tradeable.
@DigestDuffy@karun1710@PremierLacrosse porting xT to lacrosse is a fun one. hardest part with no tracking is filling the transition matrix, you need enough possessions per zone to trust the cell. video ball-tracking could bootstrap it if you've got the patience.
@afctag whoscored's a solid base, but it's event data. it can't see the run that dragged two defenders or the press that forced the bad pass. the numbers are real, just not the whole picture.
@brightsign5d the gap cuts both ways though. you can't build xG for pelé's era, no shot coords, no tracking, just the box score. some of the unreal is us only having the totals. 1000 goals is still 1000 goals.
@Jon_Mackenzie press coming good is usually a synchronization thing, not a design one. the question is whether iraola's rest-defence shape actually changed or the trigger timing just tightened up. those move very differently.
@ConnorHolden00 the tight-space vs open-field success split would settle this. players labelled as sole carriers usually have a big gap between the two. low turnover in congestion is the tell that he's a tempo-setter, not a runner.
@BeGriffis works right up until a striker actually commits to the keeper and the LB spot is 40 yards from anyone. love that vaduz don't even pretend to press it though