Surya K

💻 Head of Data Sci @ CricViz 🎓 Imperial College London ~ opinions my own

[CricViz @CricViz]

Have you been keeping a keen eye on the Women’s T20 World Cup?
If so, you will have noticed the introduction of the Skill Scales model.

This new model for the Women’s T20 World Cup was commissioned by the International Cricket Council (ICC) broadcast team and developed by CricViz to characterise player skills and styles for a global audience.

The Skill Scales model evaluates players across the defining skills and styles of T20 cricket.
For batters: Power, Control and Innovation.
For bowlers: Threat, Control and Variations.

Each is rated on a 0-100 scale, calibrated across all women’s T20 players since the start of 2023, with a minimum of 180 balls faced or bowled.

The full Skill Scales methodology can be found on our website 👉
https://t.co/eIkukXcyaJ

Or you can watch the ICC video explainer 👉 https://t.co/R1sTYXksV7

And read the BBC’s coverage 👉 https://t.co/EoBqLbc8P6

#CricViz #WT20WC2026 #WomensT20WorldCup #CricketAnalytics

[Surya K @sssskoch]

📢 Hiring for my team at CricViz!

You'll get to build the ML systems behind models like WinViz - used by teams and broadcasters globally. Looking for solid experience with batch and real-time orchestration systems. Role sits in Data Science team.

https://t.co/KmfmgvV816

[Surya K @sssskoch]

@himganj153 @DeepLongLeg fake news

[CricViz Analyst @cricvizanalyst]

Hi Stuart - in short, it's a white ball metric, which compares a player's actual performance to what we'd expect of a batter/bowler in their role. There's more detail in this explainer if you'd like to learn more.

For Archer in this case, it means he's costing his teams just over 6 runs more per innings than we'd expect from an "average" bowler in T20 cricket.

[Surya K @sssskoch]

@soumithchintala Now imagine if the most used messaging app in the world did the same when you get a new phone 😭😭 (@WhatsApp 👀)

[Surya K @sssskoch]

@TonyElHabr @lastrowview @robwhickman @AnEnglishGoat Ah ok makes sense - just applying a smoothing filter (Sav-Gol?) to the positions array fixed it?

[Surya K @sssskoch]

@AnEnglishGoat But I'm not sure that's the best way to do it as there's many adversarial cases, e.g. teammates close together appear to have no PC of their own, as said teammate would have controlled that area anyway

[Surya K @sssskoch]

@AnEnglishGoat This is perhaps a more general question - as the only implementation of individual player PC contributions I've seen takes the difference between PC with all players, and PC with player in question removed from field

[Surya K @sssskoch]

@HemmenKees Interesting! Would be interesting to see the number of shots for each block, I wonder if the yellow blocks have lower sample sizes - making overperformance more likely. And the raw xG for the long shots - must be very low to begin with?

[Surya K @sssskoch]

@AnEnglishGoat Awesome, thank you @AnEnglishGoat 👏!

Any thoughts on input smoothing, wrt. speed and necessity?
Looks like the FoT version uses a Savitzky–Golay filter on the positions to calculate velocity.

[Surya K @sssskoch]

@btsportfootball commentators should get into betting, they're predicting everything before it happens!
...or get the technicians to fix audio sync
#btsport

[Surya K @sssskoch]

#fot @Soccermatics https://t.co/XLfgRxNX2h
My submission for the Liverpool Challenge

[Surya K @sssskoch]

@citizenmatt @AlexPl292 @ideavim Thanks :)
Reason I needed ideavimrc at all is for line wrapping (set whichwrap)

This is a primary difference between vim and normal editors - so would be nice to have this since ideavim is kind of a hybrid - any updates on this/workarounds?

[Surya K @sssskoch]

@ideavim @citizenmatt wasn't sure where to reach out - but sourcing ideavimrc never seemed to work for me!
Does it just go in your home directory? (Pycharm, Mac)

[Surya K @sssskoch]

@citizenmatt @AlexPl292 @ideavim Thanks for your help! - figured it out (kinda)
I was trying all this in a Pycharm jupyter tab - where it seems broken. But works as expected in normal tabs. (using `set nu rnu` and `set nonu nornu` to test this out).
Is this lack of functionality in jupyter tabs known?

[Surya K @sssskoch]

@citizenmatt @AlexPl292 @ideavim Haha yep :)
~/.ideavimrc

[Surya K @sssskoch]

@AlexPl292 @ideavim @citizenmatt Yep I'm trying to use a very basic one just to get it working first
So right now it's just `set relativenumber`