PBP: EPL - Manchester United FC vs AFC Bournemouth - Jul 4, 2020 [R]

On the subj of Ole, I've got a post in another thread. I think somebody on the staff has been working on something. Patterns are emerging that weren't before.

The chances post Bruno absolutely were good enough to convert. Every goal bar the penalty had a low xG. Mason's 2nd was a good chance, maybe 25% of the time that's a goal. His 1st? not as good a chance. Martial's was probably a goal maybe 1% of the time (not for him). We were lucky to score 5 goals, I think 3 goals maybe 4 was fair for our output. 6 or 7 was only likely after we lucked out getting to 5.

 

The Fred chance where he slipped ang got up, that was saved?

 

The thing with xG is that for sure, some models have it wrong or not as accurate as can be, but others are incredibly specific.

For example, what a good model will do is calculate the percentage chance of a goal being scored from a shot/volley/header/other from that location with the attacker unmarked/marked loosely/marked tightly and factor if there are bodies between the shooter and the goal like w/ a FK or a shot from distance.

It seems alien when we are used to calling things half chances etc etc but it really is valuable analysis and it takes some of the doubt out of thinking about goal scoring. It isn't a guess, it is historical data used to calcuate the likely outcome given several variables.

 

Mason’s first goal was .07xG, which seems very low, and his second was .11. Not sure how to reconcile the first being the more difficult of the two.

 

Our xG for chances we didn't score on constitutes most of the xG for the game, and we get no xG for disallowed goals due to offsides, or for good passes that nobody quite got on the end of.

We could say we were lucky to get 5 based purely on the statistics provided but, on the basis that we are looking at the chances we didn't score rather than trying to subtract the ones that we did, I think 6 or 7 would not have flattered us.

 

It depends on the dataset, but some models are incredibly good. Basically statsbomb has 20 years of data based on leagues/matches they have footage of.

The models do not consider which foot is dominant.
a 30 yard shot is absolutely a very low chance event. Ronaldo put as many long shots in row M, section 6 as he put on target, much less in goal. The reason the data is invaluable is because we can all remember a screamer but it is a lot harder to recall all the times a shot went narrowly wide or was well saved. The biggest indicator of whether a goal will be scored is where you shoot from.

The best indicator of a player scoring? Touches facing goal inside the box, the closer the better. No coincidence Rashford and Martial having career scoring years with Ole at the wheel. He immediately came in and said he wanted them both to be scoring more goals inside the box.

The biggest thing xG fails at is communicating. I'd be if you could freeze play before a shot, even with knowledge of the player/dominant foot etc, you couldn't beat most models for predicting goal/no goal with a reasonable sample size.

 

For me the threat is more important than the result, you cant let that threat free so you commit a defender creating spaces. I call it steph curry effect who changed the game by just being able to do those long range 3s.

 

Beckham scored 56 freekick goals in his career. It may seem like he was automatic, but he missed way more than he scored. Just like everyone else.

There isn't a single data set used to my knowledge but there aren't a million datasets. The people releasing xG info are all buying their data from like 3 or 4 different sources. How they interpret the data is what is different. Their model may account for height of a volley or height of a header etc. etc.

It is a really good barometer. One of the things that we cannot account for is our bias. Paul Scholes scored 100 goals in his career and he was a brilliant player, but in 500 games he didn't score 50 goals from outside the box. Probably closer to 15 (I don't have data on that). Most of his goals probably came from good scoring positions. Why? Because that's how you score goals. You get into good scoring positions.

There's no secret why City score as many goals as they do. They take shots from inside the boxclose to or inside the 6, from low crosses. Those chances are objectively easier to convert and that's why they convert many of their chances despite many of them falling to Sterling a notoriously bad finisher. He didn't suddenly double his output because he learned to shoot, he stopped taking as many bad shots and started taking more high percentage chances.

 

NERDS!!!

 

Sorry, don't know how to embed Reddit videos. But I loved this, this pattern of switching from left to right in attack to stretch the defence and then have AWB play early ground crosses from the right. All in the lastportion of the game too, since James was subbed on.

I love how much work we're doing on our 'weaknesses', this is the sort of thing that adds 5-10 goals in a season if we have our fullback playing these early.

https://www.reddit.com/r/reddevils/comments/hlmdwk/3_beautiful_buildup_plays_using_low_driven/

xG +0 from all three by the way.

 

I’d have imagined that Carrick would have a great footballing brain after watching his play over the years .. I never imagined OGS would ever become a manager and I’m far from convinced now