Team shape, tactics and personnel

[MSvillain]

5 hours ago, Tomaszk said:

Teams never end up defying xG long term.

We're currently playing very well though, and deserve the results we're getting IMO. Martinez has been incredible as well, held absolutely everything near his goal.

If the good performances continue, it's more likely the xG starts to look healthier than the results turn. We do have a tough run-in, that's a fact. Draws and losses will come eventually.

Graham Potter managed to defy his for 2(?) seasons at Brighton then a further 1/2 season at a completely different club so maybe its possible to deviate for longer than people think!

But yeah i imagine there will be tweaks to improve the situation

[Tomaszk]

21 minutes ago, MSvillain said:

Graham Potter managed to defy his for 2(?) seasons at Brighton then a further 1/2 season at a completely different club so maybe its possible to deviate for longer than people think!

But yeah i imagine there will be tweaks to improve the situation

Not sure, they ended up performing so well Chelsea wanted to hire their manager.

[DJBOB]

The Chelsea vs Villa game is an interesting case study as is the Leicester vs Villa game. 
 
Potter has historically underperformed his xG but I theorized that it’s a result of his structure which is opposite of Unai. Control the ball in the opponents half, cut it back into the top of the 18 or just slightly in and pepper shots at goal at high efficiency areas. Really good at propping up the xG but as we saw, a dedicated defense limits them to mostly narrow angled shots through lots of bodies (aside from the Mudryk chances)

 Unai will historically not generate lots of shots, hence lack of xG, but will create high quality chances as the defense is stretched trying to press into our half. 
 
I would say unless we have premier talent, Unai teams will always slightly underperform their xG. Vice versa with Potter. 
 
And then of course Leicester created incredibly high xG chances because of the turnovers even though we plenty of the ball but difficult shots through lots of defenders. 

[CVByrne]

1 hour ago, MSvillain said:

Graham Potter managed to defy his for 2(?) seasons at Brighton then a further 1/2 season at a completely different club so maybe its possible to deviate for longer than people think!

But yeah i imagine there will be tweaks to improve the situation

The reason for that was not having an actual centre forward. xG is based off a sample based on years and years of PL data. The majority of shots come from goal scoring forwards. When your shots are coming from the less skillful finishers you expect to shoot under your xG in terms of conversion. 

Chelsea against us for example, lots of low value shots from range by players not adept at scoring from there against a GK who is elite at stopping shots from distance. You're asking for zero goals from their entire second half of pot shots. 

Remember xG measures of all the shots from that exact position your probability of scoring. If we put a CB in the position or a centre forward the xG isn't adjusted for the specific skill of the player taking the shot. 

[CVByrne]

Take Villa Chelsea game

We had 0.38 chance missed and a 0.38 chance scores by Watkins in first half.

They had 0.36 chance, 0.21 chance, 0.34 chance and a 0.14 all in firat half and scored none all their other chances we under 10% lots of pot shots we want them to take when we have an elite shot stopper.

We consistently create these 20%+ xG chances and we are creating them for our CF who is (and always has been) a good finisher. 

[tomsky_11]

25 minutes ago, CVByrne said:

The reason for that was not having an actual centre forward. xG is based off a sample based on years and years of PL data. The majority of shots come from goal scoring forwards. When your shots are coming from the less skillful finishers you expect to shoot under your xG in terms of conversion.

Potter's Brighton also didn't create as high an npxG per shot as we do under Emery. Average for 21/22 season was around 0.084 compared to us at 0.113 under Emery.

On defence, Potter's Brighton kept opposition to 0.084 npxG per shot, whereas we are worse under Emery at 0.119. Our number though is heavily skewed by our few defeats coming from a very high number of high xG shots. We actually have outperformed our opponents in this metric in a higher proportion of games than Potter did in 21/22:

 

[tomsky_11]

Possible bad sign for Saturday. Newcastle have the 5th highest npxG per shot in the PL this season, and the 2nd lowest npxGA per shot.

Possible bad sign for the following Saturday. Brentford come highest in the former and 3rd lowest in the latter stat, and therefore have the biggest positive difference between quality of shot taken and conceded.

Edited April 13, 2023 by tomsky_11

[DJBOB]

38 minutes ago, tomsky_11 said:

Possible bad sign for Saturday. Newcastle have the 5th highest npxG per shot in the PL this season, and the 2nd lowest npxGA per shot.

Possible bad sign for the following Saturday. Brentford come highest in the former and 3rd lowest in the latter stat, and therefore have the biggest positive difference between quality of shot taken and conceded.

To be fair, this is all from the style of play. Newcastle play a heavy press but defensive formation that is, sorry to say, reliant on tactical fouling to break down build ups and keep attacks slow into the final third. High npXg per shot because of interceptions high up the pitch. 
 
Same with Brentford who are elite at set pieces and especially goals within the 6 yard box which naturally have a very high xG. 
 
All in context. I trust Unai to know that these are their strengths and to gameplan against it. 

[tomsky_11]

1 minute ago, DJBOB said:

To be fair, this is all from the style of play. Newcastle play a heavy press but defensive formation that is, sorry to say, reliant on tactical fouling to break down build ups and keep attacks slow into the final third. High npXg per shot because of interceptions high up the pitch. 
 
Same with Brentford who are elite at set pieces and especially goals within the 6 yard box which naturally have a very high xG. 
 
All in context. I trust Unai to know that these are their strengths and to gameplan against it. 

The stats don't surprise me, especially Brentford given their massive focus on data and where it has taken them so far.

I've no doubt that we'll have a gameplan for them, but then so will they for us, and they are clearly good at preventing the kind of chances we want to, while being very good at creating high value chances themselves, something we have struggled more to prevent on occassion.

[CVByrne]

1 hour ago, tomsky_11 said:

Potter's Brighton also didn't create as high an npxG per shot as we do under Emery. Average for 21/22 season was around 0.084 compared to us at 0.113 under Emery.

On defence, Potter's Brighton kept opposition to 0.084 npxG per shot, whereas we are worse under Emery at 0.119. Our number though is heavily skewed by our few defeats coming from a very high number of high xG shots. We actually have outperformed our opponents in this metric in a higher proportion of games than Potter did in 21/22:

 

Yes but I said we create our chances for a premier League calibre centre forward who will finish the chances. They were creating the xG for sub standard finishers than the average for the league (which xG measures). 

Watkins has scored the exact amount of goals over the almost 3 seasons he's in the PL as his xG. We are creating the chances for him to finish them. 

If we had him crating space and chances for say Kamara we would have less goals than our xG 

[tomsky_11]

22 minutes ago, CVByrne said:

Yes but I said we create our chances for a premier League calibre centre forward who will finish the chances. They were creating the xG for sub standard finishers than the average for the league (which xG measures). 

Watkins has scored the exact amount of goals over the almost 3 seasons he's in the PL as his xG. We are creating the chances for him to finish them. 

If we had him crating space and chances for say Kamara we would have less goals than our xG 

Yes I understand that and wasn't contradicting you. My point being it looks like they weren't even creating good chances, just lots of poorer ones, irrespective of the ability of players having them. The players being bad only compounds this.

[AshVilla]

 

[Wes]

Think that Tifo analysis is spot on personally.

• From a statistician's point of view - our underlying numbers suggest we have been overperforming/lucky in a number of games to have the point tally we do.
• From a tactician's point of view - Emery has done a really good job of embedding in a system which plays to the strengths of our players (Watkins highlighted here).

Overall a fair summary; on a purely numerical basis we may be overperforming, but we have enabled all of that to happen because of the good work by Emery and the development of the players we are witnessing. I'm sure Emery and his staff are aware of this, it's why in all his interviews he continues to stress that we need to continue improving.

The trajectory is positive and that's the main thing.

[KMitch]

I think some here are missing the point of xG/xGA.  They're "leading indicators" and are used to look forward and predict outcomes/events as opposed to "lagging indicators" which look backward on whether or not the intended result was achieved.    

Example:

If someone wants to lose weight and the only thing they did differently was look at their scale every morning and record their weight, they might get some success just from paying more attention to the scale and being more conscious of their eating habits.  On the other hand, if you track the amount of calories you intake and the amount of calories you burn every day through exercise and keep them under your daily maintenance calories needed, you'll have a lot of success losing weight in the long term.  The leading indicators for losing weight are calories in minus calories burned compared to your daily maintenance target.  The lagging indicator is your body weight every morning when you get on the scale.

Translating this to football terms, if you consistently create more quality chances and reduce those of your opponent, you're going to win more games in the long term.  xG/xGA looks at the quality of the chances created and looks at how likely the team is to score or concede based on them.  

That being said, football isn't as cut and dry as the thermodynamics of weight loss.  There are a lot more factors which go into winning/losing games than purely xG/xGA and luck pays a huge part in winning or losing games.  You can lose to some incredibly unlucky goals or mistakes, just like you can win when the opposition has dominated the game statistically but doesn't finish off their chances.  Weight loss isn't like that.  In weight loss with all things considered normal, it's impossible to do everything right mathematically one day, then step on the scale and gain weight the following morning.  Thermodynamics doesn't have an element of randomness to it like that.  

On a long enough timeline (30+ games), the randomness/luck element of football is reduced significantly in the data and xG/xGA/xPTS is a good predictor of how teams should fare going forward.  Just looking at one specific game where the xG/xGA/xPTS didn't match up to the result doesn't mean the system is broken or doesn't work.  Hell, if it was that good of a predictor of games and an exact science, then the betting market would be completely broken and there would be no point in actually playing the games at all.

What does this mean for Villa and our current situation:  Yes, we're over performing at the moment.  No, this run of form isn't going to last forever.  No, Emeryball isn't breaking football statistics and the xG/xGA/XPTS system.  We're largely overperforming the statistics right now because Watkins and Martinez are in great form and are skewing both statistics in our favor.  At some point they'll drop off back to normal.  We should enjoy the ride this season and not expect European football, because we're incredibly lucky to even be considered in the conversation at this point in the season.  We still need to move forward with upgrading the squad with depth all around this summer, regardless of the improvement in form some players have shown under Emery so far.

[Indigo]

An unsurprising amount of disappointing replies to that video.

[Mark Albrighton]

2 hours ago, AshVilla said:

 

I enjoy a bit of statistical analysis and I find xG a fairly interesting subject.

The thing is, as he highlights the Man Utd game as one we were supposedly “a little bit lucky”…were we? 

I don’t see the defence splitting pass for Bailey’s opener as “lucky”. I don’t see Digne’s pin point free kick as “lucky”. And I don’t see Ramsey’s goal in the top corner as “lucky”.

I tell you what I think WAS “lucky” - Man Utd’s deflected goal. 
 

I don’t want to come across as a luditte, there is definitely room for this sort of stuff. But you can’t tell me the vast majority who watched that game, then subsequently thought the least deserving side won. 
 

[KAZZAM]

Just shows XG and XGA isn't an effective measurement on who should win a game. 

I mean the Leeds game numbers in that vid are a joke. They should've scored 2.4??? We dominated them until the last 15.

[Indigo]

58 minutes ago, Mark Albrighton said:

I enjoy a bit of statistical analysis and I find xG a fairly interesting subject.

The thing is, as he highlights the Man Utd game as one we were supposedly “a little bit lucky”…were we? 

I don’t see the defence splitting pass for Bailey’s opener as “lucky”. I don’t see Digne’s pin point free kick as “lucky”. And I don’t see Ramsey’s goal in the top corner as “lucky”.

I tell you what I think WAS “lucky” - Man Utd’s deflected goal. 
 

I don’t want to come across as a luditte, there is definitely room for this sort of stuff. But you can’t tell me the vast majority who watched that game, then subsequently thought the least deserving side won. 
 

Agree that sometimes the context of games can be missed when trying to make broad readings such as this. Similarly the West Ham draw is listed in green but their xG from that game is predominantly made up by the penalty they were awarded, if you put that to one side then Villa created the better scoring chances and put up a higher number in the balance of play so you can't exactly call it "luck" to have drawn.

Reminds me a bit of the 1-0 win away to United last season where even removing the penalty miss from Fernandes they put up a surprisingly high xG relative to the eye test, but that was because they had about 25 speculative shots from range of minimal value that added up to make it seem more of an unlucky result for them than the reality of the game.

Definitely value to it but as with anything it can't be solely relied upon. It's a bit like how FBRef has suddenly become the sole determination of judging a player now for some people as if there isn't many factors that go into those numbers from the standard of team they play for, style of play etc.  

Edited April 14, 2023 by Indigo

[tomsky_11]

2 hours ago, AshVilla said:

 

Feels a bit harsh to lump Everton, Spurs and perhaps even Man Utd in the lucky pile. Ran a quick expected results model myself using xG data from fbref and got the following result likelihoods for our wins...

...draws...

...and losses...

For Everton and Spurs the most likely outcome looks like a win, while the Utd game we appear marginally the more likely of the two teams to win.

Then with the draws, if the most likely results had occured we'd be a point better off. So being unlucky in the Wolves games has more weigh than being lucky in the West Ham game. Then there's a whole separate discussion around the West Ham penalty and the one we weren't awarded and who had more luck there given the more likely result without the penalty was a win for us.

Then I wonder if there is something in the outcome of the shots regards luck. Is it fair to say that we were lucky if we prevented goals via blocks or saves, or is that down to the individual abiltyof our players, and/or collective ability and tactical set up in the case of blocks?

Take the Chelsea game for example. They generated 2.18 xG, but for 1.45 of that a goal was prevented by a save or defensive block. We were perhaps more lucky the other 0.73 did not hit the target. They failed to stop any of our 0.91 xG worth of shots, with 0.4 of them resulting in goals. So were we lucky or did our defensive set up and player's abilities earn a result? There's obviously more that just this that plays into shots being blocked or saved, and maybe could do more detailed analysis split blocks and saves or using post shot xg for the later to determine goal keepers ability or luck?

[The_Steve]

3 hours ago, KAZZAM said:

Just shows XG and XGA isn't an effective measurement on who should win a game. 

I mean the Leeds game numbers in that vid are a joke. They should've scored 2.4??? We dominated them until the last 15.

Part of the problem when they run data models without actually watching the games they speak about.