they still don't account for defensive positioning or goalkeeper positioning last i checked
They only predict the past, but not future performances though. Games change and vary tactically. Just because you have had high xG/many chances in one game doesn't mean you will get it in another one.
They actually work pretty well as predicting tools. Again, "professional" gamblers use them extensively, if it didn't work they wouldn't
I think if you just gamble based on stats you going to get things wrong a lot, but it might work on average. You can't predict things like Pools dominant win this year and Leicesters win with just stats.
You can, and a lot of people in fact did(including several people on this here board, including yours truly), predict leicester's fall based on xG/xGA
Agree. There are still so many parameters which simply cannot be accurately reflected by any of the models. xG, same as other stats, can be referred to as a reference to prove a certain point (e.g. a more clinical finisher, a better shot stopper), but it's inadequate to summarize a team's overall performance.
Once again, no. It is actually pretty good. Not my opinion, there's usually a strong correlation between xG, performance and results
Once again, no. It is a good tool to prove a certain point but it's still deficient to summarize a team's overall performance. Maybe one day people could add more parameters to the model and include every action on the pitch, then it's more relevant. Every year top teams are big talking points because they chase for the title, it has nothing to do with xG.
Ah so you predicted their 2 best players would get injured
From the way he sets up his argument he'll always win. When a team drops, the xG always predicts its coming; when it doesn't, it is an outlier. I'm a fan of xG and I quote it occassionally myself. The reason why people hate this kind of advanced statistics is that some misuse/misinterpret it deliberately/unknowingly.
Chances created and results naturally correlate, but you need to predict it for future results as well. Thus you need to analyse tactics and how teams play.
Hm. Yes - and it worked out well this time around. But the exact same logic (they can't sustain it) failed spectacularly not too long ago when Leicester actually maintained their unsustainable form long enough to - sensationally - win the PL. If you placed any bets on Leicester's ultimate league position that year in - say - October - based on xG, you'd feck up rather badly.
Yep exactly. Besides I think anyone could predict that a team like Leicester would miss their best players and dont have the same depth as the top clubs, and that they'd eventually start to get tired and have a wobble at some point.
Does it matter that they "reverted back to the mean the season after" though? What matters is that they won the league. I'm sure we'd all love it if United won the league and then reverted back to the mean and were the 3rd best team.
They were already reverting back to the mean before the injuries though
Hence why xG models are based on large samples, and why people who know how to use will tell you it takes roughly 1/3 of a season before it becomes truly reliable for predicting purposes
Yeah, gamblers usually rely more on recent information than season long numbers
Generally speaking, a team that creates a lot of chances against bottom half teams has a good attack, so it stands to reason that they will create a few against top half sides. Usually the difference here is that top half sides have better defences - but we have xGA to account for that too
Situational xG is the biggest point mentioned to explain liverpool's dominant season, particularly vs City. Liverpool were dominant at tied score, extremely dominant when chasing games, and took the hands off the wheel when leading by 2+ goals. City on the other hand really struggled to chase games when down in the score. Yes, we do have the data for that too
Yeah using it together with lots of other things certainly works well. Just as you look at past results, goals scored etc.They were already reverting back to the mean before the injuries though
Hence why xG models are based on large samples, and why people who know how to use will tell you it takes roughly 1/3 of a season before it becomes truly reliable for predicting purposes
Yeah, gamblers usually rely more on recent information than season long numbers
Generally speaking, a team that creates a lot of chances against bottom half teams has a good attack, so it stands to reason that they will create a few against top half sides. Usually the difference here is that top half sides have better defences - but we have xGA to account for that too
Situational xG is the biggest point mentioned to explain liverpool's dominant season, particularly vs City. Liverpool were dominant at tied score, extremely dominant when chasing games, and took the hands off the wheel when leading by 2+ goals. City on the other hand really struggled to chase games when down in the score. Yes, we do have the data for that too
You mean they were already tired and not in incredible form for the entire season? No way, nobody could have predicted thatThey were already reverting back to the mean before the injuries though
As @Pagh Wraith said Leicester's over-performance in the early part of the season was tied to Vardy out-performing his xg by a ridiculous amount.
It wasnt though. They had other players playing out of their skin and in the second half of the season not so much. Maddison, Soyuncu and Ndidi all found it tougher in the 2nd half of the season. Maddison wasnt scoring anymore and Soyuncu and Ndidi were finding it harder to defend. Their performances dropped without anything to do with Vardy's finishing
Switch in formation, absence of players (e.g. Maddison, Chilwell, Pereira, Soyuncu), dipped form of players (e.g. Barnes) all contribute to the fall of Leicester (and Vardy). "Regression to mean" is just oversimplifying the situation and is no more than a lazy explanation looking from hindsight. Fact is, xG gives no idea about when the regression is coming and you can only tell that from the context of the game, which is much more important than xG itself.
i must admit i didn't know a whole lot about xg and the naysayers will obviously state the obvious and say that the game has a lot more variables to it than what one metric will indicate but i was swayed by how it predicted that our fortunes would swing at the end of last season and that was exactly what happened. If i'm not wrong re Leicester there was an observation that it's predictive ability was more limited when it came to counter attacking teams like Leicester who will create fewer but maybe more easier chances to score on the break.They were already reverting back to the mean before the injuries though
Hence why xG models are based on large samples, and why people who know how to use will tell you it takes roughly 1/3 of a season before it becomes truly reliable for predicting purposes
Yeah, gamblers usually rely more on recent information than season long numbers
Generally speaking, a team that creates a lot of chances against bottom half teams has a good attack, so it stands to reason that they will create a few against top half sides. Usually the difference here is that top half sides have better defences - but we have xGA to account for that too
Situational xG is the biggest point mentioned to explain liverpool's dominant season, particularly vs City. Liverpool were dominant at tied score, extremely dominant when chasing games, and took the hands off the wheel when leading by 2+ goals. City on the other hand really struggled to chase games when down in the score. Yes, we do have the data for that too
That's one way to look at it. The point being made is that xG metrics were telling us Leicester's results didn't line up with their performances and unless the latter dramatically improved, it was unlikely the results could be sustainable
Efficiency just as performances can change though. Just to assume the xG will stay the same, but the efficiency will not is a bit odd.
It isn't because xG are built on massive amounts of data. xG aren't the be all end all, infallible metric that solves football, no. But they are usually pretty reliable. The vast majority of the time, xG, performances and results tend to line up with only relatively small deviations - those deviations mostly being down to player quality (hence it's normal for great teams to overperform and shite teams to underperform)
It was always unlikely that a team could play so well all season without doing it in previous seasons. But Leicester did it before and won the league so I guess it was possible.
I agree the metric works well to quantify the quality of chances in games. It is not perfect with that either, but it is doing a very good job at it.
I bet in 2013 we did. Worst Fergie side to win a title. Mostly down to RVP’s incredible finishing. I think the 99 side that won 3 titles in a row, two of them easily less so.
Except that's not exactly true.
It's great as a predicting tool over a long stretch of games. Because it's based on averages, and those tend to play put over a whole season, exceptions tend to be rare
Using your logic, Leicester would have to improve their general performance and improve their defence in their title winning campaign or it was likely that they wouldn't keep shutting out opponents and results would suffer, because they significantly outperformed their xGA. The result speaks for itself.
Now you're just being facetious.
In this sense we can't blame Moyes, van Gaal or Mourinho for the decline in recent years. It's just "regression to mean" according to the xG extremists and we can do nothing about it.
That's a little simplistic but the rot definitely started in Fergie's later years, which of our top players from the 08 CL winning side did he successfully replace before leaving in 2013? He overestimated the quality of many of the replacements he signed.
There's only one xg extemist round here that I can see... ffs
