Football's New Stat - What is Expected Threat?
Vložit
- čas přidán 30. 08. 2021
- Try The Athletic for FREE for 30 days: theathletic.com/tifofootball
We’ve had Expected Goals (xG), we’ve had Expected Assists (xA), now we have Expected Threat (xT) - but what is it? What does it tell us? How can you spot it in-game?
Written by Tom Worville, illustrated by Craig Silcock
Follow Tifo Football:
Twitter: / tifofootball_
Facebook: / tifofootball
Instagram: / tifofootball_
Listen to the Tifo Football podcast:
The Athletic UK: bit.ly/TifoPodChannel
Apple Podcasts: bit.ly/TifoFootPod
Spotify: open.spotify.com/show/06QIGhq...
Watch more Tifo Football: Tactics Explained: • Tactics Explained | Ti...
Finances & Laws: • Finances & Laws | Tifo...
Tifo Football Podcast: • Tifo Football Podcast
Most Recent Videos: • Most Recent Videos | T...
1 Popular Videos: • Popular Videos | Tifo ...
About Tifo Football:
Tifo loves football. We create In-depth tactical, historical and geopolitical breakdowns of the beautiful game.
We know there’s an appetite for thoughtful, intelligent content. For stuff that makes the complicated simple.
We provide analysis on the Premier League, Champions League, La Liga, Serie A, Bundesliga, World Cup and more.
Our podcasts interview some of the game’s leading figures. And our editorial covers football with depth and insight.
Founded in 2017 and became a part of The Athletic in 2020. For business inquiries, reach out to tifo@theathletic.com.
Music sourced from epidemicsound.com
Additional footage sourced from freestockfootagearchive.com
#xG #xA #FootballStats - Sport
It feels weird to hear Joe being calm and normal after watching Tifo Irl
You could almost argue he's got split personality syndrome 😂
his analogies are making their way into tifo main but without the ridiculousness
I think for xT to be most effective it needs to be used in conjunction with Packing. Packing is basically comparing how many opposition players are between the ball and the goal when I receive the ball and when I release the ball. So essentially it measures how effective my pass or dribble (or both) was at taking opposition players out of the game. It's similar to xT but uses the other team as a reference point rather than an area of the pitch. Combining the two probably would be very effective in finding the most effective progressive players
There's a stat for that but I just can't remember what its called.
@@shredder807 a stat that combines the two?
Very interesting addition. And wetagree that that might be one way to improve this statistic.
fantastic idea, however set-pieces could skew with the results. A player like van dijk winning a header that passes 5-6 opposition players in the 18yard box will have a much higher value than someone like grealish playing a high risk pass to set an attacker 1 on 1 with a goalkeeper after beating 1 player with his dribbling. Grealish would obviously be the better ball progressor however the metric would indiate it to be van dijk. I think the spacing of the other 21 players on the pitch has to be integrated in some way.
Much more information needs to be used in order to create a realistic calculation of value added from these events. For example player movement speed while in action and opposition average position on the field.. but it seems as this is not a part of the data and therefore not considered in the stat. It is very hard to find the perfect measure for this though. But as data increases the value of these measures do as well.
Well this was 5 minutes of pure knowledge most of us didn’t have, or even know existed
Agreed
We knew, but we didn't quantify it
Expected Threat is the new stat that Arsenal fans have all got a degree on ever since Odegaard joined them 😂
Odegards threat is getting pegged by Lacazette
Or doing the pegging
@@_compa_santi_ lol
@@_compa_santi_ 😂😂😭
Yeah as an Arsenal fan I think this expected threat stuff is so ridiculous. Banter banter at the bottom of the table. Don't tell me we have good stats.
This is a really cool stat, and goes some way on exploring something I’ve always wanted to see more data for: the pass before the assist.
So many players get labelled with this idea that they’re amazing playmakers who don’t get many assists because “they get the pass that leads to the assist”, it’s always baffled me that we didn’t have some way of quantifying that.
I would also include things like tackles and interceptions into this. Winning the ball higher up the pitch definitely gives more threat than winning it in your penalty box.
Bergkamp would be an ideal example of this - traditional stats never got close to doing him justice at Arsenal
It reminds me of Messi waaaaay back at Barcelona when he often pre-assisted his own goals with that classic diagonal through pass to Jordi Alba and then sprinted towards the penalty spot to finish Alba's cutback.
@@Bruno-gj5vn You mean last year and the decade before that? 😏
@@SanxBile Yep.
What about the pass before the pass before the assist?
*_"Your balls are more valuable at the edge of your opponents box"_*
Tifo is always on point
"xD"
I can see the influence of American sports (Baseball, American football) on the evolution of the football stats. Stats tell alot of the stories...Maybe not everything, but alot of it. Brilliant video, as always. ⚽⚽⚽⚽⚽⚽⚽⚽⚽⚽⚽
I can so see the american influence
Moneyball?
We should have "expected relegation" stats...you know...for Arsenal 😀
Love your videos bruv
@@wilfordcolbert1333 🤣🤣🤣🤣🤣🤣🤣🤣
xT has been long needed in football. Would be interesting to see the xT numbers for the likes of Xavi and Iniesta
I think xT per time (for a move) will be interesting. Ie how fast the expected threat increased due to that Action or transition
d(xT)/dT now we getting mathsy
Football stats getting so complicated now that they sound like a GPU model.
at least football stats are available for the general public to access
@@parapluie000 wdym?
@@aldren707 football stats are easier to access than a GPU
And we're only getting started!
Lol football is a business with huge money involved. These are all designed to make sure that when clubs spend money on players, they’re getting what they paid for and they’re sure how good of a player they’re.
Really want football to bring in Basketball's concept of player gravity. It's always talked about e.g. players who draw away defenders in relation to space etc. but I'd love for it to be more specifically/statistically done
I'm excited for Kicky McKickface. He's got a real future
Face of a champion , no doubt!
He's gonna win young player of the year for sure!
Got him for cheap in my fantasy league this year, think I'm gunna walk it tbh.
Bottler! Extremely overrated.
I've seen my fair share of Tifo videos...the drunk guy at 0:35 is probably the most cursed animation on this channel so far :D
brilliant work!
What's amazing is I was able to identify each player by their silhouette without them being mentioned😅😅
Exactly 🤣
Ah, but did you identify Del Boy and Trigger at the end of the bar?
Great explaination, another great video. Been here for about 2 years on the channel I think, fantastic content all the way! Luv ya
It's when you sign CR7 and everyone starts to expect a threat in the opposition box
penal threat
@@R3in_Ch don't let modern day media fool you
also women’s expected threat that he’ll assault them
Aerial threat
@@R3in_Ch xP
This xT stat reminds me of American Soccer Analysis’ goals added (g+) model. They use some sort of machine learning model, the details of which go over my head, to estimate the net effect on expected scoring of every action a player takes during a game. xT seems to be the same process, but with a specific focus on the impact of passes and dribbles. By contrast, g+ also measures impact of shots, fouls, pass receptions, and “interruptions” (a catch all for defensive actions).
Lacazette -0
Lacks a threat
Joe's finally back... Can't wait for his Ronaldo video!
Unless I missed it, this video didn’t mention that the values are determined by 4 “iterations.” Meaning the probability of scoring after 4 “actions.” This makes much more sense as to why your own corner flag has ANY xT at all; it is based on the next four actions 👍🏻
What good timing! On my birthday and I’ve been waiting to see this piece on xT 😁
Interesting that most of the advanced stats out there are focus on midfield and attacking play. Few explain the art of defending and to some extent goal keeping which probably explains why top teams still struggle to assemble a solid backline.
Really clear! Thank you for that! This and the xG videos really helped
Wonder how much of your audience missed the Only Fools and Horses reference!
'talk about money, it's yuppies favourite thing to talk about'
'I saw one of them old five pound notes the other day'
'no trig, not yer coin collection, talk about your wealth'
'yeah but I ain't got none of that'
'well neither have any of these'
I’ve been preaching how under rated Hudson-Odoi is both by other fan bases and his own
The xT model actually shows how valuable CHO is for Chelsea
They dont even use him lol. Dortmund will tho
@@TheRealPingu No, Dortmund won't, because the transfer (resp. loan) didn't happen.
Chelsea wanted to sign Saúl from Atlético first in order to have replacement. But when they did so, some few minutes before midnight, the German transfer window was closed already.
The DFL calls it a day at 6 p.m. and nobody knows why. It's a a huge disadvantage for German clubs on the transfer market.
The analogy at the start is very Joe Divine-esque 😄
There are so many conditional probabilities involved in a model like this. I have not seen any documentation on it, but it seems like it has to make so many assumptions to avoid those. Regardless, my suspicion is that this stat is, at best, meaningful only at a team context (and not match-up), or at worst, over-fitting historical outcomes.
These animations are getting better and better
Surely an easier way to do this is just adding together the vector components in the direction of the opposite goal of players moving with the ball or passing the ball?
Benjamin Mendy is the first player ever to increase his xT by 0.9 in just one week. Dangerous.
I can see a lot of extra stats stemming from this: Defenders xXT (reject upon XT), overlap of the XT with the opponents XT to comprehend better how each team is performing in a span of 10 for example minutes, Conversion upon XT to evaluate attackers etc.
Feels like these modern stats have been introduced to gauge just how much better Leo Messi is than everyone else
Very well explained Joe 👍
I love this channel! :)
The evolution of football is becoming more complicated
Good work
I wonder how would xT apply to long balls from CB's that lead to goals, vvd for example. Wouldn't that be low xT situation which frequently converts to goals or chances.
Surely will be a stat using xT in conjuction with xG to showcase true player creating ability
Seems to me like a stat that is designed to interpret the game to overvalue those passes, while undervaluing a backwards 5 yard pass in the center of the pitch. Johan Cruijff turns in his grave.
I’m gonna name my new fifa career mode avatar Kicky McKickface
Kicky McKickyface really racking up the XT there
This could probably help revolutionize defensive stats too. Defensive prowess is nearly impossible to quantify with the current metrics but something along the lines "xT reduced" could probably help change the game. A tackle in the box should be a lot more valuable than a tackle in the midfield and I believe xT could help us predict that.
A tackle in midfield could be more useful because you are closer to the opposition goal, though. This is what modern pressing is about.
Those hands moving on the blackboard are mental
It's interesting to see how expected threat on right side tends to be higher than the left side. Would there be any theory to back the statistics up or is it just due to the fact that there are not enough stats to see the whole picture?
I thought Kicky McKickface would be much higher on the xT metric, to be honest.
Expected Threat makes use of Monte Carlo Simulation and cumulative probability to estimate the chance of scoring from the opponent's perspective. Looking past match data is a direct way for estimation but ignores the dynamic flow and shocks of a football game. This concept compensates for that flaw by adding the scoring chances when the player moves and passes the ball closer to the opponent goal post. Nonetheless, estimating scoring chance also has to look at players and clubs stats, since player training and club management can have a significant impact on their success in league matches.
if it doesnt include crosses and cross receptions it's not really an accurate representation of 'threat'
So, in essence what we're talking about is the legendary Alexander Hleb "pre-assist"
Joe is surprisingly sane in this video
Great video. I get that this metric is trying to identify some early positive work, whether or not a goal is scored, but does it totally exclude forward progress provided by an assist? I presume it does because Harry Kane was the leading assist in the PL last year and is not featured in the list. The metric definition needs to be clear on this. I can believe Grealish is top but where's Maddison, Kante, Tielmans, all of whom produce more threat than e.g. Henderson. I hope that the designers of this metric (which I think is a great concept) go back and review parameters to identify improvements.
This content is gold
There’s no many problems with looking at one stat on it’s own. Getting the ball on the edge of your box when they’ve just had a corner is far better than getting the ball on the edge of their box when they’ve got 11 men behind the ball
KickyMcKickFace really is an underrated baller...
Would be really interesting to see more about how this data is collected and processed. I wish I knew jobs like this existed when I was in school!
Data collecting manual and boring. Processing is done by inputting those data into formulas. Unless u mean coming up with the models/formula for it, those are done by sports analytics companies which typically hire people with advanced degrees in Maths or related subjects (ie. boring people) so I don't think you miss out on much :)
Benjamin Mendy : Expected Threat 99
What is the timeframe for a goal occuring after having been in that position necessary for it to be counted?
Could xT be applied to player combinations, to find the best midfielder partnerships? That would be super interesting
For me stats are important in football but it's not everything in football
Football attributes, big game performances and leadership are important
A couple of issues jump out immediately.
First is that this doesn't seem to take into account where the opponents and team-mates are. You could pass the ball forwards to a team-mate who receives it and then is mugged by the 3 opponents surrounding him. Making such a pass is a bad decision, but with this metric would count as 'good' because there are many other similar forward passes that increase threat. Similarly, a crowded winger switching play to a team-mate in space but in line has made a good decision but wouldn't get any kudos from this.
Second is, when compiling all this into a player, ignoring their fails makes it meaningless. Always going for risky passes that if successful are certain goals, but generally failing and hence not being counted will inflate a player's 'worth'.
Seems like a good stat but not taking into account position of defenders seems to defeat the purpose slightly. If I run into a better location but there are 2 defenders there who immediately take the ball off me, my xT has increased but in that instance the threat debately decreased.
Beyond this I question how well this will work looking at possession based teams vs counter attacking teams
I've always wondered, Who are all these stats for really? Because all of the things the numbers say are visible if you watch a game. Is it to speed up the workflow of scouts and pregame prep?
Partially yes, it allows you to spot outliers in lower leagues, but it also allows you to compare high level players on the smaller details of their game. This lets you potentially create your optimal line up for a team, or to see if some players would be better with different roles
This is all well and good but where does Joe get his trousers?
That Kicky McKickface is a promising prospect hehehe
Not sure if it can be considered a reliable goal scoring probability model. It should be an independent measure of how successful a team is at creating and attacking. You could have amazing build-up but get cut off by a good interception.
that's really a Kicky McKickface's analysis!
I'm confused as to why your remove crosses when looking at which players improve a teams scoring chance in possession.
Because even a half decent cross into the box which just goes out to a goal kick would have high xT but in reality, that cross isn't that threatening. So by that model, cross spammers and burnley players would suddenly look like world class players when in reality they are just lobbing the ball in the box.
What is the name of the app at the end?
I first heard the word "Expected Threat"in the pod Fantasy 606.Statman Dave was the guy to use it.
Statman Dave knows nothing about stats haha
is it how much time you spend with Mendy?
But this system doesnt take into account the positions of opponent players.. two exactly similar passes can have two very different impacts depending on the opponents position.. for example a simple pass out from the back Vs a break away pass sending your teammate on a 1v1
so... expected threats conceded is "xtc" then? sweet!
From where do you get the stats?
i think i'll just watch the game
this is exactly what i looove about soccer!
Packing is still my fav but it never really progressed in the last 5 years
Notice how no Arsenal player is in the list and it’s exactly what I have been complaining about all year.
Why does xT on the right side has higher value? Does the stat bend because of Mo Salah?
Honey, wake up! New football stat just dropped!
I'm interested to see if those xT of Grealish actually turn into goals this year... so far we know him only as a foul winning player :P
either way great video to kick off September :)
When are we gonna get the x throw-in stat?
What is the difference to Possession Value?
the more stats like this are used in football, the more players like Grealish are recognized for being good footballers despite not scoring or assisting as much; and the less likely that overrated goal poachers (like a certain Portuguese forward) are hyped for simply scoring and not contributing anything else to the game.
I like this.
Gentle reminder that pionel pandres pessi has been finished as a top level footballer since joining the fashion outlet that plays football part time in paris. But sources tell me that since Eibar will be gettig back to la liga next season, he will rejoin the league in his superb role of eibarman which he so comically entertained with all of us
Would love to see this stats...on every big team un europe..
Matic and fred: passing forward? We dont do that here..
so... will this be included in FM22? now that I know it, I want it.
This stat will confuse highlight watchers. Only people that watch the whole games will see how they match up. Thiago and Hendo being on the list proves my point. Highlight watchers wouldn't know about Thiago and Hendo being player's like that.
this model actually makes attacking roles players play could be measured. How good those players on building up their team attack, pressing, and opening up for link-up play.
This is the eye test stat basically.
Yeah it is. But only if you are a person who watched the game and not highlights.
I'm worried that this stat favours players who make hazard passed that sometimes work but usually just give the poll to the opposition who then has a high expectation to score.
If you don't give negative points for failed passes that help the opposition, the defenders who keep doing dangerous forward passes when the right thing would be to play the ball to the keeper, rack up more points than sensible defenders.
Not at all a problem. You don't analyse a player and come to a conclusion based on a single statistic.
I am not hundred per cent sure if there is a deduction of xT on possession lost (eg. if player loses ball on 0.002 area, the player's xT decreased by 0.002+potential opponent's xT), but it would be stupid not to have that kind of feature I suppose.
You should add toghether then divide by time.
Kicky McKickface
I see The Athletic are fellow men of culture
Who is Henry Cooke and why does he drink baked-bean juice? 1:03 Lmao
There will no more tactical analists, only softwares and drives.
The concept of xT always went over my head. This video made it pretty clear tho. But some football stats seem one-sided. For example, a player with brilliant movement can increase his team's chances of goals by pulling defenders towards him and thus giving his teammates free space. But the players xT wouldn't increase in this regard. Actually I don't see many stats which compliment off the ball movement. Football stats are great & gradually improving but there's still a long way to go.
Maybe those runs could be taken into account by tweaking the heat map of percentages depending on the positions of the opposing team members. But that would require way more data to go into it.
Similar with goalkeepers, you can't really calculate good positioning so keepers like Allison are underrated in terms of stats
0:24 how do they know what I look like when I go to the bar
Quite complex
More parameters are needed to be added to this theory. but its better than simple XG
Please do a video about Saul to Chelsea
So no-one is going to talk about Joe's body proportions in the opening?
I was really expecting kovacic to be higher based on how he plays
from where can i get football data
?????