Expected Goals on the rise in the Media

It seems like Expected Goals are going to get a lot more play in mainstream football analysis this season. I love the idea of letting the reader judge a given chance to better grok the idea of the metric.

However, xG values work better in the aggregate, not necessarily judging a single shot. It’s good to tamp down on the “They should have scored there idea”, but perhaps it will go too far.

I’ve long advocated for some sort of error bars on xG, but this may be harder to convey to a general audience. Perhaps generic phrases “1 in 10”, “1 in 4”, “an unlikely shot” could and should be used by announcers, but media will likely put percentages on top of chances in the replays.

What do you think about how the mainstream (and us!) should talk about expected goals as compared to real goals? Also post any other places where it is being mentioned in the media.

It should also be noted that Opta is doing a lot of work to make these values available in the first place, fair play to them! Sam isn’t yet here, but @Worville also works at Opta, and I hope we can provide some feedback to them on how it can be good for analysts AND fans. Sam has more here: https://twitter.com/GregorydSam/status/895695401357103105

Also, I love this diagram and I think it definitely is crucial to understanding base xG, before adding in other variables.

I imagine they will go down the route of using xG on single shots and single games rather than using its predictive power which is a shame.

Hopefully the BBC with collaberation from Opta help educate the presenters and those involved with some basic statistical knowledge to prevent the misuse of xG as much as possible. I would encourage them to use xG as a reference point and upgrade or downgrade the chance bearing in mind that xG doesn’t take into account some variables. Similar approach should be taken to single game xG scores.

But if it doesn’t work out that way and we do get pundits dismissing and misusing xG, there’s probably a bigger positive to take from this. Some won’t but a lot of viewers will take an interest and research xG further. This can only be good for analytics and take it further into the game moving forward.

I think I’m less optimistic about xG working, or working regularly, in an MOTD setting than some. I think it should be used sparingly, because, let’s be honest, it doesn’t need to be used a lot on a highlights show.

Maaaybe have an early episode xG heavier as a primer to show off the fancy tools, and then otherwise use the concepts to educate commentators or pundits.

I imagine they will go down the route of using xG on single shots and single games rather than using its predictive power which is a shame.

In fairness to those who present games, talking about that single match is pretty much the point. If they’re going to use it on the game-day broadcast, they’ll need to use it in game-day context.

Now if we’re talking about weekly shows or previews or anything else, yeah, using it as the predictive measure makes way more sense, but is also far harder to illustrate to the mathematically uninclined.

It is a game-day broadcast but after a few minutes discussing the game the pundits discuss the season; retrospectively and going forward. They discuss performance whether it’s the team, manager or individual players. This is where xG is most meaningful.

Maybe they will use it in that sense but I just get the feeling it will be used most often on single shots and games as there’s no doubt this is the most popular application of xG (just look at the demand for @11tegen11 plots on Twitter).

Right, and again, I think this is just because it’s very difficult to easily and succinctly explain the notion of predictivity to the general public. Maybe it’s not even just the “explain” part, because I don’t doubt that it’s possible to do (though it still appears difficult).

Frankly, most people just don’t CARE about predictivity. That’s not the point for most sports fans. They want to know what happened and why, not so much what is going to happen. Knowing the future takes the fun out of it. That’s why people still watch sports live, rather than just watching a recording after the fact.

Heck, that’s why people hate sports analytics. We’re literally trying to take the “anticipation” away.

1 Like

The main ‘non-single game’ use I can see for it on something like MOTD is for a leading goalscorers chart which is either just in the corner as a thing of interest or briefly touched upon if someone notable is hugely over/underperforming.
“Rooney’s not had a great start at Everton has he Alan?”
“No, he hasn’t, but he’s been getting in good positions. If you look at the goal scorer table he’s down in 15th, but he’s been better than that. If you look at his Expected Goals he’s up in 6th, so if he keeps making chances for himself he’ll start banging them in”
[End Scene]


In case anyone hasn’t seen it across their Twitter timelines so far, xG has been used amongst the post-game stats that roll across the bottom of the screen as MOTD shows the manager post-game interviews. No explainer on what it is for anyone who didn’t already know, but then I don’t think many people take notice of those rolling stats anyway; maybe the BBC would say that the article on the BBC Sport website is adequate enough (and they didn’t have any good space for it in the show anyway)