Yeah, so “they each have qualities the other doesn’t have and I value their qualities higher” is a vague enough argument that you could basically make that argument about any two players, regardless of how much better your disfavored player actually is. And much of the rest of what you said is just really strong outcome bias—where you’re justifying an argument that they were better than Kane by reference to them winning things when they didn’t actually play as well as Kane overall in the competitions their team nevertheless won. The only real exception to that is potentially Villa in WC 2010. I think there is actually an argument that he was better than Kane has been at the WC. Even that seems debatable to me, though. Kane’s had some really good WC moments. He scored the winning goals in two KO-stage matches in this tournament, was fantastic in England’s hardest group stage match (against Croatia), etc. The knock on him is that he’s been relatively quiet in late KO-stage rounds, but Villa was not impressive in the 2010 semifinals or finals anyways. It’s not all that different IMO, except that Villa had a better team so his team won anyways. The only argument you’re left with is basically putting Torres and Villa ahead because of big “moments.” And like, okay, you’re free to do that I guess. But your initial argument was that they’re better players, and that seems like it’s something that obviously needs to be based on an assessment of more than a couple moments in their careers. It might make sense if you asserted that you think Torres and Villa will end up being more memorable players. I’m not sure I’d agree with it, but I could see the point. But asserting that they’re better players because of a very small number of isolated moments seems like a real exaggeration.
So if Messi wins no trophies, and played at Tottenham, he's the best player ever? Listen, maybe htat's your opinion, but it certainly is not my approach to professional sports. Re:qualities, I already discussed briefly in another post. Again, I could scour through your post history and show you relying on "outcome bias" but we both know how you operate: variable measuring sticks depending on the argument, hyperfocusing on specific details, and a generally obsessive nature to posting where you just can't let go and need a final word. Unless Kane wins a major trophy not called Bundesliga he will go into the archives of football history. For the record, he's an inferior player to Wayne Rooney - the best English striker of my lifetime.
I think this is true but like @lessthanjake I'm not wholly convinced it's accurate. I look forward to researching more and I think both have incredibly strong metrics/records. Di Stefano was known to be quite controlling of the Real Madrid dressing room and Puskas was surely at least past his physical peak by that stage plus has two years without training properly etc. Puskas clearly has a much stronger case in terms of international career but much of that was beyond Di Stefano's control. I definitely need to brush up on his South American years. It's absolutely possible that Ronaldo could have eclipsed them both and the come back to win a WC is sensational but the injury just took too much of a toll for me to make the case that as their careers panned out Ronaldo should be ahead. He really is the lost GOAT contender as for his era no one else makes the discussion. You have... Puskas/Di Stefano in the 50s Pele in the 60s Cryuff in the 70s Maradona in the 80s No one from the 90s but should be Ronaldo into the 2000s Then Messi and Ronaldo from the late 2000s onward. Tbh I would find ot hard to even have Ronaldo at 8 all time, he just wasn't the same after the injuries so his record and performances simply dont justify that ranking by the criteria I use. Theres also the hint of some of the issies that potentially stopped Maradona from reaching his peak which doesn't rwally apply to the other candidates (except Cryuff to a lesser degree). Would probably put Beckenbauer at 8 tbh.
Slow down. As I said, I would continue my response later as I didn't have time, so I rambled a few things on the fly and couldn't address everything. If you have a direct question (or a few) regarding something I said, you can ask me. I would appreciate it, because you are opening up conversation into so many directions with the last response that I can't possibly answer all of them deeply and coherently. Clearly, you are not a fan of AI, which is okay. And yeah, I read minds because I see the words you choose when you talk about it. I see behind them frustration, disgust, a bit of presumptiveness, combativeness, etc. Which is perfectly fine, btw. It is a mark of genuine and healthy skepticism, and the world would be better if everyone approached AI with the same attitude rather than being naive about AI. However, this could be your problem if you keep a closed-minded attitude, because you are unintentionally sabotaging your own options and opportunities in life. As with everything, there are pros and cons, and you are unintentionally throwing the baby with the bathwater if you don't understand it. I am open to talk about "overselling" and general trends of AI, and how it relates to the future of AI in football if you want. It is surely a very interesting topic and a lot to say about it. But let's keep that on a side for now. To directly respond to your curiosity about why the Sofascore model rates Pedri's performance as 9,0, it is because it rates his passing performance extremely highly in that game, as well as the dribbling performance. I can't tell you which plays exactly it rates highly, as I don't have access to the model, but even if I did, I wouldn't be able to tell you, because that is exactly how AI works: processing at least 100s of patterns and correlations to generate an evaluation. What I can tell you for sure is that it sees things that it "likes", in the passing performance in particular, because Sofascore, as of updates from this season, is breaking down performances across 4 major categories: shooting, passing, dribbling, and defending. His passing category is maxed out. Worth saying here is that it is not reduced to only those 4 categories and includes data points that don't belong in neither categories. Of the available information, his xA performance stands out the most. It is 1,23 xA, which is very high and extremely rare even for players who have a high rate of passing in possession-based teams. For example, no teammate of Pedri has more than 0,5 xA in that game. It is not just a passing volume. Laporte has 106 completed passes (20 more than Pedri) in that game with 0,13 xA. Independent of whether the Sofascore model gave a correct rating for Pedri, or if it is justified to rate a passing game that highly, the model is not hallucinating anything. Football models are not hallucinating at all. They are not a generative branch of AI like popular LLMs, which indeed hallucinate. Just to make that crystal clear. We can talk further about xA and a justification to use it in an evaluation process, its shortcomings, and so on. But Pedri's rating is not some random pattern that the model decided to hallucinate based on nothing. There are tangible reasons why it rated him so highly. The reasons that a human eye often fails to notice and register. Mostly, it comes down to failure to comprehend the extent of impact of many small positive actions. Brain is terrible at processing big amount of information. Also, I will correct you, because you are wrong. Even if you weren't consciously thinking about the context of matches, these two ratings are not equivalent AT ALL for one simple reason: available data based on which the Sofascore model(s) evaluate Pele's and Pedri's are fundamentally different. Notice, Pele's rating doesn't have any mention of xG, xA, or any other trendy metrics, meaning it is evaluating (estimating, if you like that word better) based on a more primitive dataset. Perhaps, if Pele's performance had a richer and more detailed dataset, like Pedri does, his performance might be evaluated as 10. The opposite might be true as well. If Pedri's performance was evaluated based on primitive metrics, he might have got worse rating because the model wouldn't know the xA of his passes. They are not the equivalent ratings that came from the same process. You can't compare current ratings with retroactive ratings 1 to 1. This is why it is important to actually understand what these models are doing behind the scenes. That understanding gives you insights into what models might overestimate or underestimate, and where you should "trust" them. Either way, consider this: The best-rated players by Sofascore are also the ones that end up on the top by newspaper ratings. The order is a bit different, but in general, Sofascore ratings tend to strongly correlate with human opinion across the 8-game tournament. It does so independently, with zero human interference. The Sofascore model, in particular, might get it wrong sometimes in extreme, unique cases (relative to human opinion - if we assume human opinion is a gold standard), but generally it gets it reasonably right. And we are still in the early days of such models. They will get incomprehensibly better very quickly. So, in my opinion, you are overreacting over a singular performance, while still not understanding that AI is much, much, much better at detecting specific patterns, and, on the flip side, you are not realizing how fallible and unreliable human opinion often is in certain aspects. EDIT: Also, worth highlighting: "If we look at what actually happened it turns out the magic model with all its so called objectivity was actually completely wrong and I was completely right. Pedri's 9 performance wasn't as likely to contribute to a win as Pele's 9 performance. We know this because Brazil actually won 4-1 whereas Spain drew 0-0." You fundamentally misunderstand probabilities. An outcome of a match or an outcome of an action doesn't change a probability model of the action. Ideally, objectivity is completely agnostic to outcomes. This is exactly why I say that humans are terrible at objectivity and isolating actions from outcomes. You are doing it unconsciously. They never fail at that.
If being combative is a sign of healthy skepticism and a good discussion then I'll go as fast or slow as I please, cheers! I take your point about it using more primitive data to rate Pele and more advanced metrics to evaluate Pedri. I'll also take this opportunity to repeat my point, in the real world where the games actually happened that more primitive data gave a much more accurate rating of Pele's performance than of Pedri's. This is going by your own account of what the number is measuring, I.e. the predicted contribtution to winning the game. As I said if you had told me only two specific statistics for each player I would have been able to give you a much more accurate prediction on which performance was more likely to contribute to a win than ai did using so much more data. What does that tell you? As for my skepticism of ai trust me it is very well earned and informed and not something that is holding back my life in any way. I dont see roko's basilik under my bed. Im not quite a luddite though and appreciate there are definitely use cases where it has strong applications but I'll wait until after blockchain has changed the world before drinking the ai kool aide. You asked about ai in general not for football models iirc. I didnt think the model was hallucinating. I would say to claim it as objective while saying humans have bias a d make errors when the ai was developed and designed by humans is an obvious glaring contrdiction. Regardless I genuinely appreciate your response and would be happy to have a more in depth discussion. I'll even try to hold off on the chippiness because I don't think you deserve it and it comes across more combative or snarky than I intend.
Ronaldo was amongst the best in the world in 1996, 1997, 1998 and 1999. Then he returned to the top in 2002 and was still a top elite player until 2003-04, with Real Madrid. He still managed to play as a starter at WC 06. Plus, he was already smashing it in Brazil earlier, being called up for the national team to dispute the World Cup in 1994, stealing the spot from players like Evair. So that would make it at least 6-7 years playing at a very high level, in total. I think that's enough for him to reach Top 8 in a ranking of greatest players. Again: the only reason he doesn't rise up to Top 3 or Top 4 is because of the injuries that hindered his longevity. Otherwise, we would be talking about Ronaldo in the same category as Pelé, Maradona and Messi. I believe his peak was higher than Puskas', as he pretty much DOMINATED the football environment in the late 90s. Puskas obviously had longevity on his side, but I don't see him having the impact in the early 50s that Ronaldo had in the late 90s. By the way: there is also the World Cup factor, of course. Which, like it or not, always counts (and hadn't even been mentioned yet). Ronaldo got it in his "few years" on the top, coming back from a serious injury, whereas Puskas didn't, even though he had an equally capable team in 1954.
So this is where one can and should draw a distinction between “greatest” and “best.” Messi might well be the best player ever still if he’d been on Tottenham. But he would be fairly unlikely to be the “greatest” since it would’ve been hard to win a whole lot at club level. Of course, if he *did* win a bunch for them anyways, then perhaps that would itself aid an argument for being the “greatest” (similar to people talking about Maradona’s success at Napoli). But yeah, since you’re also positing that he wins no trophies, then he probably wouldn’t be the greatest in that specific scenario. But anyways, I’m not sure the point of the analogy there. You didn’t say Torres and Villa are greater than Kane. Rather, you suggested they were “better” players than him. That’s an assessment that isn’t as trophy-focused as greatness is. As I’ve already said though, if you’d like to amend what you said to instead say that you think Torres and Villa are more memorable players because of certain specific big moments they’ve had, then that might at least be a somewhat defensible claim. It’s not one I’d agree with, and I certainly think it’ll be wrong when/if Kane wins the Ballon D’or in a few months. But it is actually true that the biggest moments those guys have had are bigger than Kane’s biggest moments. They’re not better players though. ______ EDIT: As for the Rooney comparison, I definitely see Kane as the better player, but that’s at least a closer comparison than Torres or Villa.
This is just your opinion, mate. You wouldn't. Let's play a game, shall we? Question 1: Who performed better? Player A: 0 goals, 0 assists. or Player B: 0 goals, 0 assists? Question 2: Who performed better? Player A: 2 goals, 0 assists. or Player B: 0 goals, 0 assists? Submit your answers, and then I will reveal you more information about them.
I think youre grasping at straws. Messi at the world cup finals isnt the best player ever. Hes not the best 100 players ever. Because his performance level (as any player) is contingent on supporting cast. So no. Nonsense argument on your part. Better/greater Im not eritimg essays on semantics. Im too old at this point.
whats your source that Sofascore relies on machine learning? also how did you determine journo x sofascore compatibility?? i see a lot of unsubstantiated claims
I dont need to play a hypothetical game. We are already using real world examples. You told us the model.is using a number to predict how a players performance gets translated into a number that represents their chances of contributing to winning the match, am I wrong in this interpretation? Here's two numbers for Pele and Pedri who had the exact same match rating. Pele one goal and 2 assists. Pedri 0 goals and 0 assists. My conclusion, Pele's performance is much more likely to contribute to a victory for his team. Real world outcomes, Brazil won 4-1 and Spain drew 0-0. The other example.posted in this thread perfe try fits this analysis as well. Ronaldo against Germany had 2 goals from whatever number of shots. Messi had 0 gaols from one blocked shot. The model tells you they both have the same chance of contributing to a victory by giving them the same rating. I tell you that Ronaldo is far more likely to have contributed to a victory. The real world.outcome is Brazil 2 Germany 0 and Spain 1 Argentina 0. You are right that it is my opinion that the 9.1 rating for Pele is far more indicative of his match winning potential than Pedri's 9.1 rating as it turns out this opinion is validated by Brazil's 4-1 win and Spain's 0-0 draw. I also have the opinion that goals win more games than models and Xa. It's just an opinion sure but it also happens to be reality.
Alfredo Di Stefano has a better record in major competitions than C. Ronaldo Di Stefano vs C. Ronaldo Di Stefano 100 % 1 America Cup disputed Attempts 1 Major Title in 1947 6 matches / 6 Goals 100 % Don Alfredo Di Stefano European Champions Cup ..records European Champions Cup editions 9 participations / 5 Major Titles 2 x runners up 1962 & 64 against Benfica & Inter Milan vs C. Ronaldo Uefa Champions League editions 20 participations / 5 Major Titles 1x runners up ... 2008/2009 season vs Barcelona Don Alfredo Di Stefano 56 % Titles won vs C. Ronaldo 25 % Titles won
Sorry I should actually note that the model gave Messi a higher score that it gave Ronalodo. Apparently all because of winning the ball high up the field once, that's definitely more valuable than scoring two goals!
no way man ! Puskas In terms of attacking metrics and playmaking actions, Puskás was quite similar to Maradona at his peak during the 1986 World Cup. But both contributed far more defensive work than Ronaldo Nazario de Lima .. Ronaldo Nazário's defensive work was limited to very sporadic actions—much like Romário, Adriano, Ronaldinho Gaúcho, or Baggio.
Thinking about this I should actually engage in good faith so I'll say player a in the second case and in the first it depends on their position. I'll also slightly reframe who performed better to who was more likely to contribute to their team winning, because according to you that is actually what the model is measuring unless I'm misunderstanding. That subtle distinction is important, its obviously possible for a midfielder to play better than a striker but if the striker is clinical with chances which increases their likelihood of contributing to a win. We have the example of Haaland at this WC to show that. Plenty of Norway players could have "performed better" while Haaland just needed to finish chances to be more likely to contribute to the team winning than them.
You don't need to do anything. Why would you not answer my questions tho? I can't engage with your points if you are not engaging with mine. If your goal was to just say your spiel on Sofascore, then okay. You've accomplished that. We can move on now.
Sofascore didn't do the ratings because they don't have all the games available, but Opta collected the data from the 1958 semifinal and final Back then, football involved fewer passes, which explains players with 39 or 30 completed passes. So maybe Sofascore's algorithm would punish past players a bit (they giving the same rating to Pedri against Cape Verde and Pelé in the 1970 final). What stands out is the high number of dribbles by Pelé and especially Garrincha. Since 1966, no player has reached the mark of 12 completed dribles in a single World Cup final, so Garrincha would be the record holder. And Pelé fills every column for key passes, shots, and dribbles.
Pele vs Messi in World Cup Finals Note: In the 2014, 2022, and 2026 finals, Lionel Messi played the full 120 minutes including extra time.
That's why I said he is the greatest STRIKER of all time, not the greatest defensive midfielder. At no point do I mention his defensive work, since that's irrelevant.
https://sofascore.helpscoutdocs.com/article/50-sofascore-statistical-ratings-explained “Calculated by a machine learning algorithm, the Rating assigns a value to each offensive, defensive, and goalkeeping action based on its context and impact in the match.”
I’m not sure what the point of this analogy is. Kane, Torres, and Villa played with great supporting casts and Kane clearly has played better. They all also played on lesser teams, and Kane was probably still the best one of them in those years too. He’s been better regardless of the context in terms of supporting cast.