Pieter Robberechts

Top stuff once again. Bringing validity to the conversation which is too often overlooked in the football analysis space.

Thanks! 🙏 Happy to hear it resonated.

How Do We Know a Metric Is Good? A Case Study on VAEP In sports analytics, the primary goal is to develop indicators that quantify the skills, capabilities, and tendencies of athletes and…

Before an app launches, it undergoes usability testing. Why shouldn't an advanced sports metric?

In our latest blog, we propose 4 criteria:
1️⃣ Trustworthy
2️⃣ Novel
3️⃣ Intuitive
4️⃣ Useful

Then, we put VAEP to the test...

\w @jessejdavis.bsky.social & Maaike Van Roy

dtai.cs.kuleuven.be/sports/blog/...

You can first combine the xG of multiple shots in a possession as 1 − ∏(1 − xGᵢ).

Expected goals and the Monte Carlo Trap On a team level, xG is often used to assess the quality of the goal scoring opportunities that a team creates and concedes. That is, it…

We can use xG to compute post-hoc win-draw-loss probabilities.

Our new blogpost shows MCMC's estimates of these probabilities are unstable, computing them exactly is faster than approximating them, and MCMC's runtime strongly depends on its implementation.

dtai.cs.kuleuven.be/sports/blog/...

Looking forward to the #MLSA25 workshop next week at @ecmlpkdd.org! We'll have some exciting research on soccer, ultimate frisbee, basketball and table tennis. ⚽🥏🏀🏓

Want a sneak peek? The papers are already available on the workshop website 👇

The Carnegie Mellon Sports Analytics Conference is an annual event dedicated to highlighting the latest sports research from the statistics and data science community. Registration for the 2025 CMSAConference is open.
🔗 www.cmsaconference.com

YouTube video by Hudl Statsbomb Boot it: A pragmatic alternative to build up play, Lorenzo Cascioli | Hudl Statsbomb Conference 2024

“What if, when the opposition has 11 players behind the ball, a centre-back decided it was more worthwhile to ‘kick for touch’ rather than attempt to play through midfield?”

Oh boy do I have a research presentation for you

Excited to announce a new paper together with @sigridolthof.bsky.social entitled "Perspectives on data analytics for gaining a competitive advantage in football: computational approaches to tactics".

This is a synthesis article that has 3 key objectives

www.tandfonline.com/doi/epdf/10....

1 / 4

It should work for the 2025/26 season. You only need to disable the cache once to discover the new season (see also github.com/probberechts...).

PyData Eindhoven 2025 collaboration with PySport. December 9th 2025.

🔷 𝐏𝐲𝐒𝐩𝐨𝐫𝐭 𝐗 𝐏𝐲𝐃𝐚𝐭𝐚 𝐄𝐢𝐧𝐝𝐡𝐨𝐯𝐞𝐧 𝐂𝐚𝐥𝐥 𝐟𝐨𝐫 𝐏𝐫𝐨𝐩𝐨𝐬𝐚𝐥𝐬! 🔶

On December 9th @PyData Eindhoven and @pysport.org will organize 𝐏𝐲𝐃𝐚𝐭𝐚 𝐄𝐢𝐧𝐝𝐡𝐨𝐯𝐞𝐧 𝟐𝟎𝟐𝟓!

𝐌𝐨𝐫𝐞 𝐢𝐧𝐟𝐨: lnkd.in/eBGWKs3V

5 takeaways from the Women's Euro group stages The Women’s Euros kicked off at the start of July and Catalina Bush and I launched an online app using the Value Actions by Estimating Probabilities (VAEP) framework. The app is being updated after ev...

Your fave women’s footie analytics gals (@cata-bush.bsky.social and myself, ofc) are at it again!
Below is a VAEP inspired breakdown of the Women’s Euros group stages - read to see which players and teams have shown out or bombed (whoops, Netherlands 👀):
open.substack.com/pub/beyondth...

Predicting the 2025 Women's European Championship The UEFA Women’s Euro 2025 will be held in Switzerland in July. Will Spain capture the title after winning the World Cup in 2023? Or will…

Our Women's Euro 2025 predictions are out! 🔮

Spain 🇪🇸 is our model's favorite to win the title, with Germany 🇩🇪 close behind. Meanwhile, Belgium 🇧🇪 will most likely be eliminated in the group stage.

w\ @jessejdavis.bsky.social

dtai.cs.kuleuven.be/sports/blog/...

📊 Our interactive visualization provides detailed odds for each team advancing past the group stage and eventually winning the tournament. We will update the probabilities after each round.

dtai.cs.kuleuven.be/sports/euro2...

According to our simulations, the top 5 favorites are:

🇪🇸 29% 🇩🇪 26% 🏴󠁧󠁢󠁥󠁮󠁧󠁿 19% 🇸🇪 11% 🇫🇷 9%

We estimate each team’s strength based on historical performances using Elo ratings. Then, we run Monte-Carlo simulations to calculate the probability of every team reaching each stage of the tournament.

Predicting the 2025 Women's European Championship The UEFA Women’s Euro 2025 will be held in Switzerland in July. Will Spain capture the title after winning the World Cup in 2023? Or will…

Our Women's Euro 2025 predictions are out! 🔮

Spain 🇪🇸 is our model's favorite to win the title, with Germany 🇩🇪 close behind. Meanwhile, Belgium 🇧🇪 will most likely be eliminated in the group stage.

w\ @jessejdavis.bsky.social

dtai.cs.kuleuven.be/sports/blog/...

Football Machine Learning Engineer | Jobs at Sportec Solutions AG At Sportec Solutions, a joint venture between Deltatre and the DFL Group based in Munich, we develop innovative solutions in sports data and technology. Since our founding in 2016, we have been commit...

We're looking for an ML Engineer to join our team at Sportec Solutions. Lots of exciting projects in the pipeline, including ones with skeletal-tracking data. Apply or pass it on to someone who might be interested!

sportec-solutions.jobs.personio.de/job/2151299?...

🚀 Ingestify 0.7.0 is live

Parsing football data with kloppy is fun. The hard bit is fetching raw files and keeping them clean and version-safe.

That’s the gap Ingestify fills.

github.com/PySport/inge...

It’s a wearable that anyone can use, although the focus is currently on physiotherapists. Essentially, it’s an IMU worn in a belt around the hips. I used it on (almost) all my runs outdoors, and for me, one of its main strengths was the consistency of the insights across different terrains.

My Marathon Journey with Runeasi by Pieter Robberechts Discover my marathon journey with Runeasi by Pieter Robberechts, including my race day experience and new personal best of 2:31:59.

I strongly believe biomechanical analysis will be the next big leap in sports analytics—and not just for the pros. I got a glimpse by testing the @runeasi.bsky.social sensor (a @dtai-kuleuven.bsky.social spin-off) and wrote about how it helped during my marathon training → runeasi.ai/my-marathon-...

It's mostly hockey taking ideas from soccer, but one big takeaway for me was the potential in the overlap between data analytics and sports science. In soccer, they mostly live in separate silos.

Had the pleasure of attending the Linköping Hockey Analytics Conference this week. Thought “icing” was just for cakes, but somehow they still let me on stage 😅 Loved the welcoming community and saw lots of familiar ideas with huge potential across sports. #LINHAC

I'm familiar with ⨀ to denote the element-wise product of two matrices, but I wouldn't know how to interpret this particular formula without a definition. That said, any symbol can be used in maths as long as it's clearly defined within the given context. So, I don't see an issue with it per se.

I’m honored by @leanderalphabet.bsky.social excellent write up of our soccer research!

Lots of credit to my wonderful collaborators @janvanhaaren.be @wannesm.bsky.social @probberechts.bsky.social Maaike Lotte Tom @lorenzocascioli.bsky.social @hugoriosneto.bsky.social Max Arne Tim & so many others

The Belgian lab shaping modern soccer’s data revolution A small corner of one of the world’s oldest universities has moved the sport forward, and made the argument for research institutions writ large

The Belgian lab shaping modern soccer’s data revolution

👀 Something's changed at kloppy.pysport.org

We’ve completely rewritten and reorganized 𝐤𝐥𝐨𝐩𝐩𝐲's documentation, now with:
• 🚀 Quickstart Guide for new users
• 📖 API Reference for power users

Dive in & let us know what you think!
@pysport.org 🤩

The paper submission deadline is approaching!
Submit your work to our 𝐖𝐨𝐫𝐤𝐬𝐡𝐨𝐩 𝐨𝐧 𝐌𝐋 & 𝐃𝐚𝐭𝐚 𝐌𝐢𝐧𝐢𝐧𝐠 𝐟𝐨𝐫 𝐒𝐩𝐨𝐫𝐭𝐬 𝐀𝐧𝐚𝐥𝐲𝐭𝐢𝐜𝐬 at @ecmlpkdd.org by 𝐉𝐮𝐧𝐞 𝟏𝟒!

An experimental analysis of possession state value models: part 3 Conceptually, possession value approaches such as VAEP, PV, OBV, and g+ are all identical: they estimate the chances of scoring (and…

🚨The third and final blog post in our series on possession value models design decisions🔍: Can the features chosen to represent the game state inadvertently bias player ratings? w/
@probberechts.bsky.social @jessejdavis.bsky.social Lode Van Tente

dtai.cs.kuleuven.be/sports/blog/...

The workshop is co-located with ECML PKDD, so the location follows their venue. But I’m not complaining. 😁 I'm sure it will be a great spot and I hope to see you there!