It certainly has to be considered that SofaScore's algorithm is biased against Modric in particular. His ratings are really quite low in basically all of his UCL campaigns. For instance, Pirlo's 14/15 UCL campaign, where he was clearly at the end of the rope as a top player, is rated rated 7.15, while you have to go all the way back to 16/17 for Modric to have a performance rated equal of higher to it (7.17) WhoScored doesn't seem to rate Modric well either, at least in comparison to Kroos. I think for most people, based on eye test and intuition, would choose Modric over Kroos when they were playing together. So, perhaps these algorithms are missing something.
Modric honestly wasn't that brilliant at WC2018 (the KO stages especially) and 2017 was indeed his last truly great campaign and year for Real Madrid I think. Not 2022 certainly. He was then already 32 years old. Take him for his peak years. Sofascore is Croatian... Modric is demigod there...
Can you provide those ratings for the specific games you are referring to? Also, I noticed that Gazzetta was comically harsh on Pirlo in at least 1 season (2008/09, maybe his worst season) where Milan fans in the forums on XtraTime were frequently laughing at how Pirlo must have shagged the rater's wife. Moreso, Ambrosini was consistently given 6.5 or higher to the point where he must be shagging the rater himself. Goal.com were doing ratings concurrently but I don't have access to the them, however forum posters were noting that Pirlo was (correctly) rated higher in those on a few occassions.
I'll go through and calculate his average for the 13/14 campaign, because it doesn't show up on the player profile beyond 15/16
Just looking at SofaScore for 08/09 and Xavi is the best midfielder (not including Messi, so only competing with 2 other players usually) on his team for all but 5 games I think, only two of those in the knockouts. In 09/10, he was best mid for 8/12 games! In 10/11, best mid in 5/12 games. I see your point here for Xavi, But SofaScore doesn't seem to support the same for Modric in his winning campaigns. Pirlo in 06/07 was best midfielder (not including Kaka, who is hardly more of a midfielder than Messi, so Pirlo is competing against 3 other players usually) in 5/11 games, and 4/7 in the knockout stages, including the Final (even if you include Kaka, only Inzaghi is rated higher with his two goals). What I can say though is that these (other than perhaps Bayern first leg where Pirlo scored) aren't marquee performances but good performances, where Seedorf and Kaka probably had 2 marquee performances each in the QFs and SFs.
Also, midfields can be set up in such way that exactly the strengths of midfielders with clear weaknesses (that Kroos and Pirlo had) shine through. You maximize their strengths and cover the weaknesses. But Kroos is about five years younger than Modric. Modric his last really outstanding year was 2017 I think. Sofascore is Croatian. Rui Costa could be deputy for Kaka and Pirlo. Seedorf was deputy for everyone, including Shevchenko even. Not saying it Opta etc. is perfect but I think the grades are understandable.
I'm skeptical of them and all statistical algorithm ratings, but they're not nothing. Sometimes I agree with them more than journalist ratings after I watch games. Nothing is a substitute for actually watching games and being informed by context though.
I've tried this before and couldn't find those ratings but now it's available as you say. Anyway there seems to have less matches available Kroos played as a #10 on this Bayern side (and were the best Bayern midfielder until he got injured). Robben played as a winger. How were them "directly competing for a place"?
According to Sofascore, Kroos has been a better player than Modric for almost his entire career. Like it or not, that's the site's view. I think this is interesting information, but it would be better if the criteria were open and the equations were made public. In the NBA there are several statistics that evaluate a player's performance (PER, VORP, WINSHARE), all the equations are open. It's not just one rating either, but several.
I'm currently calculating the SofaScore averages for Pirlo, Xavi, Modric, and Kroos for UCL earlier than 15/16 and I can already tell you there are huge discrepensies between SofaScore's ratings and WhoScored's, especially for Xavi. SofaScore seems to favour Xavi's game, and probably Kroos' too. I suspect it could be a passing volume thing that SofaScore rates higher, as looking at Pirlo 12/13 vs 13/14, his higher ratings in the latter compared to the former coincide with higher passes completed per game.
Maybe, but Xavi his best rated games until the end of 2009 (many think 2008-09 was his peak) stand below a number of other midfielders. He does have a high proportion of 7.5+ rated games though. He has nothing like the 9.9 of Lampard vs Bayern.
So the conclusion, is your eye test / opinion matters more. Of course, they are whole lot of nonsense. You cannot divorce actions from the whole, and statistical aggregate sites are already extremely limited in terms of the scope of tabulated actions (Stats), but then bring the nonsense of an arbitrary equation to the mix.
There is probably some modicum of value from the raw stats themselves. Beyond that, there is genuinely no reason to provide credence to a number rating.
Yes, eye test and informed viewing is most important. I only use SofaScore ratings as a (limited) indicator in the case where watching the full matches with the required contexts is not possible or realistic. I don't think every member on this forum is going to the full seasons of Xavi or Pirlo or Modric, so these numbers can be used, in conjunction with other indicators such as eye witness opinions and media consensus, to paint a rough if incomplete picture of form and performances.
Media ratings are also flawed and influenced by a ton of things. In Italy the likes of Buffon got a high rating by default. Elo ratings for national team are a good predictor of results in the near future. Better than the Fifa ranking. Sofascore team averages and assessed team strength also a very good predictor. Much better than Whoscored or Fotmob. Seriously. One of the reasons why I think it is the best of the three. It is the best in predicting actual results.
How do you know Sofascore is the best in predicting results? I’m not saying you’re wrong, but I just wasn’t aware that anyone had looked at which one was more predictive, and am genuinely curious where that information comes from.
Voetbal International article of a few months ago. Have to see if I can find it back. The 9.9 for Lampard vs Bayern is btw easily explainable if you see he had 11 tackles, lost the ball only 18 times the entire match (passes, shots etc. going astray), scored two very good goals andsoforth. His only attempted dribble was succesful. It is an outlier but not like Morientes his grade or so.
Just saw Kroos perfect 10 performances. It would be interested to watch the actual matches. Kroos completed 157 (or attempted, can't remember) passes in 90 minutes against Arsenal with percentage over 90%! That's insane. Again, I think volume of passes is weighted quite heavy in the rating system.
Highest rated champions league campaigns according to sofascore Finalists only Lionel Messi 2014/15 8.52 Cristiano Ronaldo 2013/14 8.51 Robert lewandowski 2019/20 8.35 Lionel Messi 2010/11 8.33 Cristiano Ronaldo 2015/16 8.12 Cristiano Ronaldo 2017/18 8.09 Karim benzema 2021/22 8.05 Xavi Hernandez 2008/09 7.98 Mohammed salah 2017/18 7.86 Cristiano Ronaldo 2016/17 7.85 Highest rated CL KO stage campaigns according to sofascore Finalists only Cristiano Ronaldo 2013/14 8.48 Lionel Messi 2010/11 8.44 Lionel Messi 2014/15 8.30 Cristiano Ronaldo 2016/17 8.22 Robert lewandowski 2019/20 8.10 Karim benzema 2021/22 8.05 Xavi Hernandez 2008/09 7.98 Highest SF rating for Modric(when his team reached the UCL final) 7.70 in 2013/14
Yes. Sofascore certainly values passing moreso than whoscored. On football algorithms in general: They are formulaic in a way they evaluate performance. There are two limitations to football algorithms: 1. Capacity to capture/measure actions. Some things are lost in there.. 2. Necessity to categorize actions in a finite amount of categories like (normal pass, long pass, key pass, cross, etc.) and arbitrarily weighing in each category. In this process, it comes to misjudgement. For example, each successful dribble is given +0.10 points and unsuccessful one -0.05 by default. It lacks further, qualitative evaluation of actions. Some dribble might be worth 0.25 and other 0.05.. algorithm assumes average/expected value for all players. That is the whole story to it. So rather than looking at 8.5 rating as definitive rating of a player in a match, look at it as: based on measurable actions and assuming these actions have their average (weighted) value, the rating 8.5 is THE MOST LIKELY rating of the performance per model. Look at it with uncertainty. It is not 8.5.. it is 8.5 +/- 0.5. Sofascore rating tells us what would be an approximate rating for a player who did these actions assuming all of his actions are of average/expected quality. As well as any other rating model.
That's a good way to look at it. I actually think there are even more limitations, however, when it comes to contextualizing that the algorithms simply can't account for. For example, in a two-leg UCL tie, maybe a player scores one goal in a 2-0 win in the first leg, but wastes a big chance in the 90th minute by attempting something flashy (or worse yet, irresponsibly drifts offside and the goal is ruled out, which wouldn't even register in the algorithm). In the second game, there is a 3-0 comeback. That ruined chance should really count more than a "big chance" missed in the rating system. This scenario could even play out over 90 minutes. Another example, maybe a winger dribbles his defender 6 times in the second half, but actually the defender was injured in the 60th minute and was asked to play out the remainder of the match for the team, putting his body on the line in a show of character. The team eventually wins. Not quantified. A leader, a great captain like Puyol or Maldini, steps onto the field as a 70th minute substitute as they return from an injury. The score is 1-0 and they need to hold out to advance in the tournament. The presence of the great leader lifts the whole team and they fight harder for every ball, winning the game. The great leader makes few tackles or blocks, but their presence was worth so much. not quantified. I could go on. One question I genuinely have though, is does SofaScore take into account the level of opposition? Scoring against Madrid is not like scoring against Olympiacos, and scoring against Buffon is not like scoring against Karius.
Sofascore is stationed 500 meters from my home. I will apply for a job and if I get it, I will tell you. I am not sure any rating is be able to predict future consequences of an action and adjust accordingly as you suggest in your first example tho. I think it shouldnt be adjusted regardless. I think 2nd and 3rd example belongs to my 1st point about not being able to measure and quantify every little thing om the pitch. Such context and leadership are definitely such categories. Regarding your point against oppoistion, I am pretty sure they dont adjust and rightly so in my opinion because, we dont intuitively look at 9/10 performance against Real Madrid in the final the same as 9/10 against Eibar in a league match, so I dont see why would we need to adjust it further. It would be an overcorrection. Doing it twice. In a rating and in our heads when looking at them.