Scout Lab
122 posts
Scout Lab
@ScoutLabApp
Player/Team analysis • rankings • visualizations 🔗 https://t.co/7Lgb1EjhDn | https://t.co/bEyjGD1S8S Personal: @JoseSilva933
Katılım Temmuz 2024
328 Takip Edilen2.9K Takipçiler
@CymruComps @ef_Stats It captures the image based on the dimensions of your screen, which is why it doesn’t look good on mobile, for example. Your browser window is probably not very wide. Try hiding the sidebar before taking the screenshot in that case.
English
@academicismo @withqwerty A faster learning rate at the start of each season, gradually converging to the steady-state value, might be worth testing.
English
@withqwerty In the spirit of the discussion, saw an article arguing for an Elo reset at the start of each NBA season.
Not fully convinced, Elo rating that resets every season already exists: it is called "points".
Elo is memory.
Any thoughts on that?
English
@Novvigodly @PaRaLlEx99 I don't like Opta's definition for progressive carries. I use the one from Statsbomb, which is similar to the definition for progressive passes: carries that advance the ball at least 25% of the remaining distance towards the opponent's goal.
English
@PaRaLlEx99 ScoutLab probably has stricter criteria @ScoutLabApp i am right ?
English
x.com/cfc_krishna/st…
C. Palmer premier league 24/25 review
penalty merchant 🥸
7/19 Big chances was finished, ass ! but 24 BCC also great
#premierleague #palmer #football #data #stats
k☭ #PorVida@cfc_krishna
@Novvigodly palmer 24/25
English
x.com/PaRaLlEx99/sta…
KDB Premier League 17/18 review
excelent player, best midfilder in PL for me
#stats #data #football #kdb #pl #python
ParallaxPath@PaRaLlEx99
@Novvigodly Give me KDB 17/18 and 19/20. 🙏
English
@adnaaan433 I think it would be a good idea to calibrate an xPass model, using “under pressure” as one of the features. The visualization above provides useful context, but with xPass that context would already be captured in the efficiency metric.
English
Now I understand why Leeds fans can't believe this, he plays too safe, which is shown by numbers here.
First of all, he plays back passes mostly under pressure. Forward passing accuracy is low.
Also he has so less amount of under pressure actions comparing his minutes on the pitch. Thus it is true that he plays safe so that he doesn't appear more frequently under pressure. This helps him to gain more percentages with that low volume.
That's why I made this visualisation also to have an in depth understanding of the data.
Mohammad Adnan@adnaaan433
Leeds United missed their best, in fact, the best press-resistant player in the Premier League, Ilia Gruev today! #leeds #facup
English
@Johnsenzx Possible, but not very likely. I don’t like most of the metrics currently available for goalkeepers; xGOT from Opta, for example, doesn’t take into account shot velocity. And I don’t have the necessary data to improve those models.
English
@ScoutLabApp Great work.
But im just wondering, are you gonna add goalkeepers aswell?
English
Scout Lab updates:
🔹 New Player Summary page
🔹 Improved layouts for Similarity & Rankings
🔹 Minor tweaks for better readability
🔗 scoutlab.streamlit.app
English
@WizardAKA Yes, he played nearly 70% of his minutes on the right.
If it looks different from Sofascore’s heatmap, it’s because they color by absolute touch volume (areas turn red once they pass a certain threshold), while mine shows the % distribution of touches across areas.
English
@kashONdaCounter Ah, thanks. It’s simpler that I thought, then. The main way to change the player should be by using the box at the top right, where you can just type the name, but I hadn’t noticed that on mobile it got that small. I’ll fix the other bug as well.
English
@ScoutLabApp You can switch leagues , let alone choose players. Hope this video makes sense mate.
English
@kashONdaCounter But what player did you search that caused the error? And does the same error always happen when you search that specific player?
English
@ScoutLabApp Sorry as in the bug being talked about here where it is defaulting to Olise.
English
@kashONdaCounter @xenostat Can you be more specific in what is preceding the error?
English
@ScoutLabApp @xenostat This error is still around mate unfortunately pretty much all options are being defaulted to bundesliga and where appropriate (player metrics) are Olise.
English
@blue_cicada8 For me, at least, it became more readable due to the higher light level. But I toned the brightness down a bit.
English
@phalmer @PavlochitisSZN This can’t be used to assess league quality. Players who move aren’t representative, they’re top performers relative to their league and/or youngsters with high potential, and you only observe those who keep getting minutes after the move. I won’t respond anymore on this thread.
English
@ScoutLabApp @PavlochitisSZN I just saw this analysis posted by the raumdeuterapp guy. It looks at change in player performance when moving between leagues. Players moving from Brasileiro and the Championship to the PL show exactly the same %age drop-offs in their performance output:
x.com/fbobiano/statu…
Atreides@fbobiano
Tenho andado a trabalhar num major update e parei para fazer um exercício exploratório sobre a tradução da performance nas diversas ligas. Recalculei o performance index - retirando o peso da força das ligas - e comparei a variação da performance quando um jogador sai da Liga A para a Liga B. O mapa cruza a origem (eixo yy) com o destino (eixo xx), considerando apenas pares de ligas com >5 transferências por ano para evitar histórias de amostra mínima. Os eixos seguem o Opta League Power Ranking: ligas mais fortes no topo, descendo para competições mais fracas. O que o mapa sugere em geral é que boa parte dos saltos de liga, depois de retirado o peso competitivo, não deixa um resíduo enorme. Isso por si só é um resultado útil: confirma que grande parte da variação “óbvia” é mesmo contexto, não necessariamente “o jogador ficou melhor ou pior”. Tende a confirmar o Opta League Power Ranking, tirando algumas exceções. 🇸🇦 Saudi Pro League - um caso de estudo A coluna da Saudi Pro League chama a atenção: entradas vindas das Big 5 aparecem com tom alaranjado, próximos de uma variação neutra. Jogadores que descem no ranking da Opta para a Arábia Saudita tendem a manter o nível estatístico, não apresentando uma variação acentuada como poderia ser expectável. Isto pode representar um aumento do nível competitivo na Arábia Saudita tendo em conta o grande investimento nos últimos anos, ou uma falta de motivação ou ambição dos próprios jogadores que apenas veem a liga saudita como uma forma de "reforma antecipada". 🇵🇹Ligue 2 para Primeira Liga Apesar de um Opta Power Ranking bastante inferior, as entradas para a Primeira Liga de jogadores provenientes da Ligue 2 tende a ser positiva, mantendo o níveis estatísticos. É certamente uma liga que os clubes portugueses devem investir no recrutamento. 🌍 Outros pares que valem nota: Brasileirão para Argentina: A prova da diferença competitiva entre as ligas, sendo a liga brasileira a liga mais competitiva do continente sul-americano. Primeira Liga para Premier League: Um salto que pode não ser tanto abrupto quanto se podia pensar. A variância média das performances está ao nível da Serie A. Significa uma proximidade tática ou que os clubes da Premier League fazem bem o seu trabalho e sabem contratar os melhores da liga? BF, Ruben Dias, Mateus Fernandes e Nico são alguns dos exemplos de transferências de sucesso. Superliga para Bundesliga: entre as maiores variações negativas da tabela. A movimentação para a maior liga da Alemanha tende a ser muito punitiva para jogadores provenientes da 1º liga Dinamarquesa.
English
Made some UI/UX improvements to Scout Lab, applying what I learned from the recent Teams app updates. The core content presentation remains mostly the same, except for the Scatter Plot.
There may be a few bugs due to the changes. Just let me know if you spot anything.
English
@phalmer @PavlochitisSZN Deleted my previous reply, but you really need to read about how Elo systems work and think deeply whether they can adjust across leagues with almost zero proximity. Brazilian vs European matchups are <1/year on average. Blindly trusting a model isn’t a great idea.
English
@ScoutLabApp @PavlochitisSZN "Just go see what happened in the FIFA Club World Cup" - you need to read the methodology, it factors in club performances in inter-competition tournaments, so Club World Cup should be included:
globalfootballrankings.com/methodology
English
@ScoutLabApp player radar and player comparison sometimes bug out when trying to change from the default Olise to a player from another league. happens more often with the selector in the lower area of the UI.
English
@phalmer @PavlochitisSZN Doesn’t make it any less wrong. Just go see what happened in the FIFA Club World Cup; not only the results but how the brazilian teams performed against the betting lines. They also have more money than Championship teams; don’t even need to adjust for the size of the league.
English
@ScoutLabApp @PavlochitisSZN globalfootballrankings.com
Championship 7th, Brazil 9th.
That site is as neutral an overall aggregator as you can get.
English
@phalmer @PavlochitisSZN At the end of the day it’s still a personal project and I’m from Brazil. And there is no shot that the Championship is better than the brazilian league. The Portuguese league is also much better at the top.
English
@ScoutLabApp @PavlochitisSZN Honestly the Championship needs to be added anyway. It's already more stronger than other leagues you have included.
Btw, the "Show Top Players" setting on the new scatters is too narrow - I made a chart for every midfielder in the big 5 leagues, it highlighted only 1 player..
English