Bedfordspurs
Paul Gascoigne
The are all a accurate within their own model
The key is what you use it for
The games between different ones is negligible
-
Dear Guest, Please note that adult content is not permitted on this forum. We have had our Google ads disabled at times due to some posts that were found from some time ago. Please do not post adult content and if you see any already on the forum, please report the post so that we can deal with it. Adult content is allowed in the glory hole - you will have to request permission to access it. Thanks, scara
*******Spurs v PSG*******Official OMT*******
- Thread starter LutonSpurs
- Start date
The key is what you use it for
The games between different ones is negligible
mudshark
Justin Edinburgh
That's just a reflection of the fact there's different methodologies for modelling chance creation and coming up with a 'quality' measure. The numbers don't have to be identical across methodologies, as long as you stay internally consistent.
Which doesn't make sense. If you have 5 different models giving 5 different answers it doesn't matter how internally consistent they are they are showing different results.
The old saying you put brick in you get brick out.
Gutter Boy
Tim Sherwood
It's an Americanisation of sports coverage. Because their sports are so boring and interrupted, they kind of need something else to do during the dead time. Whereas global sports have always had much more flow, and are about momentum rather than quantitative measures.
Bedfordspurs
Paul Gascoigne
It’s choosing one source as your output
Clubs don’t look at these different models and think what could we do different
They do it themselves with their own data teams
We could both have a tape measure and get slightly different results if all the variables aren’t fixed. They aren’t in football, that’s for sure
All data can be used up make better decisions when used in the right way and you understand the input
Bedfordspurs
Paul Gascoigne
This
I’ve heard a mate who is a spurs fan and an NFL fan claim that all the set pieces come from the NFL model
When I told him and the team who won the league in Sweden IIRC from maximising set pieces he didn’t believe me
braineclipse
Garth Crooks
It's a somewhat glorified shot statistic.
But it also tells us more than a shot statistic.
A stat doesn't have to be 100% accurate to be useful.
Bishop
Chris Perry
You can obviously draw some conclusions when you look at the numbers and the context of variables like you mention, factors like luck or goalkeeping quality are also just averaged, but as a raw stat its really just telling you the quality of the chance on average. That isn't particularly informative by itself, you need to place them versus the quality of your finishing or similar depending on what you are trying to determine. For me xG alone doesn't tell you much without combining it with some other data point.
Last edited:
braineclipse
Garth Crooks
Possession says more about style of play and game flow than quality.
Shots, and somewhat better xG tells a bit. It won't tell if a team will win or lose a game. But give me an xG for of 0.1 or 2.0 in any given game and I'll take the latter. Give us an average xG for of 1.5 for the rest of the season and we're almost certainly going to get better results than if it's 1.0 (assuming no changes to xG against).
Results are what matters. But results are over time clearly linked to performances. xG gives an indication of performance while obviously not being some perfect factual answer. I'd rather see a loss with xG numbers like against PSG than a loss with xG numbers like against Arsenal. At least the PSG game gives us some realistic reasons to think we'll do well in the future.
mudshark
Justin Edinburgh
That's a gross oversimplification. For example, there are also different ways of modelling the flow field and pressure distribution across a wing section under different conditions, at different Mach numbers and so on. They won't necessarily produce identical results in terms of raw numbers either, but the wings that are designed using them are still just as reliable.
parklane1
Teddy Sheringham
Simple as that really, XG is a buzz word that in all reality does not really show much. But it entertains the stat jockeys but overall means very little.
Baleforce
Chris Waddle
You still have to rely on what input is put in. Is it a person or ai?
Posession is determined by how many passes a team makes compared to the other. So does someone sit there with 2 clickers watching the game? How reliable would that be?
Results have to be quick so no going back double checking.
Even if it's ai. It's trying to read a 3d game from a 2d image converted and reconverted. With an infinite amount of variables.
It's a guide but is massively fallable. As are humans.
Bedfordspurs
Paul Gascoigne
They have cameras capturing a lot of it
Although… I don’t know how the stats companies do that
My nephew is doing it for derby county now so I have seen how he does it
It’s a series of short cuts on a key board watching the video
Useful when used with other things
Bishop
Chris Perry
Not to be pedantic but I’m gonna be pedantic.
If you have:
- 25 shots
- Total xG ≈ 2.0
- No individual chance over, say, 0.15 xG
Then what you really had was:
- Lots of low-probability efforts
- Probably many blocked shots, tight-angle attempts, or pot-shots from distance
- Pressure but not threat
So while the cumulative xG says “you created enough to score twice”, the experience of the match will tell us differently.
“We never really looked likely to score — we just took a lot of low-quality shots.” xG per shot, shot locations, big chances created, etc, is more informative not just the raw xG total.
What I find interesting is that in the period when we had Kane and Son I don't think we appreciated how far above the statistical norm their finishing was. Kane & Son were absolutely xG overperformers. xG models tell you what an average player would score. Kane and Son were not average players. We often had games where the xG said “1.1 expected goals,” but Kane would stick one in the top corner or Son would curl in something from 0.06 xG.
Muttley
Young-Pyo Lee
From the BBC article yesterday. Spurs have scored 20 times from an xG of 11.2, meaning they have scored about nine more than the average team would have, from the shots they have taken this season.
So here's a question. In fact two.
Why would I trust an indicator that is performing so incredibly badly? People get fired for way less.
Secondly, what will the score be at the weekend based on xG. Perhaps someone can tell me what they find.
I did find this though....Spurs were victorious in 53.2% of the Opta supercomputer’s simulations. A draw was the second most likely outcome, occurring in 24% of simulations, while Fulham came out on top in 22.8% of sims.
Even the super computer won't get off the fence and tell me the xG for the game.
It amazes me how everyone builds arguments around it. It's slightly more useful than a chocolate teapot but doesn't help me win at the bookies any more than my own judgment.
So here's a question. In fact two.
Why would I trust an indicator that is performing so incredibly badly? People get fired for way less.
Secondly, what will the score be at the weekend based on xG. Perhaps someone can tell me what they find.
I did find this though....Spurs were victorious in 53.2% of the Opta supercomputer’s simulations. A draw was the second most likely outcome, occurring in 24% of simulations, while Fulham came out on top in 22.8% of sims.
Even the super computer won't get off the fence and tell me the xG for the game.
It amazes me how everyone builds arguments around it. It's slightly more useful than a chocolate teapot but doesn't help me win at the bookies any more than my own judgment.
So a lot like var.