yimoe
5.7K posts
@A3per_ @patelsahil2002 @RK_DRAFT Nah it’s free throws.
Plus free throws more effected by luck as players rarely deviate from their true ft%
English
@RK_DRAFT Practically speaking 3p% is by far the easiest variable to control for so I get why it's done, I'm just not a big proponent of that tho
English
100% adjustment and padding with over 1k minutes what are we even doing here?
Fuck it lets just assume every lineup ever is a net neutral and any deviation is variance 👊
𝐔𝐍𝐂@notwashedcnu
I like my 3PT luck adjusted thank you very much
English
Just added 5 year windows for goals and xG
yimoe@Yim0e
G-RAPM now includes ucl data. Here is xG-RAPM as well. Improved my model so keeper ratings look better in the xG one. Can keep fiddling with it but ultimately going to add a prior next time g-RAPM is last 15 years i think. xG-RAPM is 2014/15–2025/26 docs.google.com/spreadsheets/d…
English
@M0MO_N Uh hard to explain but imagine this with 22 players instead of 10
substack.com/inbox/post/185…
youtu.be/5WRe6SoBh3g?si…
YouTube
English
G-RAPM now includes ucl data. Here is xG-RAPM as well. Improved my model so keeper ratings look better in the xG one. Can keep fiddling with it but ultimately going to add a prior next time
g-RAPM is last 15 years i think.
xG-RAPM is 2014/15–2025/26
docs.google.com/spreadsheets/d…
English
@kmedved @ARKNDJL Bad model I'm guessing. There's not a lot of stints to separate players properly on offense and defense and defenders/keepers get rated to high on offense. One idea is to 0 out the O values for keepers but that doesnt feel right. Tuning the weighting for keepers is probably
English
15 Year Soccer G-RAPM. Top 5 League Only don't pay too much attention to the O/D split. Might post stint data later.
Someone please let me know if my email is shared. Very rudimentary will improve it soon hopefully.
docs.google.com/spreadsheets/d…
English
@Loek_ty___2 Just goals hence G-RAPM. Currently working on an xG version. Only issue is the source I use doesnt have champions league data
English
@Yim0e Really cool, like what are these even based off of like xg
English
@Rajdeep_C_18 @Stan_Incarnated Yes. But over time the best creators rack up higher xa
You can show cases where xa capture creativity better than xag as well
English
xA without context is one of the most overhyped stats of all time. It doesn’t really capture ‘creativity’ the way people think.
The image below is a perfect example of how flawed xA can be-it overrates simple high-probability passes while undervaluing difficult & creative ones.
Diving 🇱🇧🇨🇦🇹🇷@DivingxDown
Olisé has 5 more xA than Yamal in league + UCL 👍 not all bcc are equal.
English
@imzispoms mismatched and lewa came clear, and peak real just as dominant as bayern were just real had a close competitor.
youtu.be/dachHO8kPcE?si…
YouTube
English
Lewandowski's stats presented without comment:
🇵🇱:
When the tie is level/losing
• 25 90s
• 10 goals (0.4)
• 1 assist (0.04)
🧲@legayushv2
Karim Benzema in UCL KOs (08-22) When leading • 27.6 90s • 14 goals (0.50) • 5 assists (0.18) When the tie is level/losing • 28.4 90s • 20 goals (0.70) • 2 assists (0.07) All timer UCL performer.
English
@imzispoms anyways think the best adjustments you can make are like team goal contribution % or clutch % while favoring players who scored more all else being equal.
English
@imzispoms his teams historically dominated and he won everywhere he went to and showed floor raising abilities with his country.
(never ask a benzema fan about his international pedigree #15)
English
🚨 3-Year Rim Assist RAPM Leaders
Measures how a player impacts the creation of assisted shots at the rim over a multi-year sample.
Offense: A positive value means increasing the lineup's rim assists.
Defense: A negative value means limiting opponents’ rim assists.
English