xG, xG2, “Lucky”Liverpool, defenders putting pressure — story of an intriguing twitter thread!
This is the story of a twitter thread!
So, the story begins on 4th February, 2020 with Mark Taylor ( @MarkTaylor0) of infogoal app posting this -
#LFC 10 game rolling xG & xG2 created follow hand in hand since 2018. xG & xG2 allowed does the same until this season when the on target opportunities that get through to Alisson/Adrian drop like a stone. How? Why? Huge reason #LFC are 22 points ahead of #MCFC
Data @InfogolApp
Here’s #MCFC for comparison over the same timeframe.
xG & xG2 bouncing around in a fairly tight band, most likely just randomness. Typical of virtually every other team bar #LFC
This gave rise to a long long thread with very interesting ideas and materials thrown around!
This is a summary of that saga -
Basic concept of xG2:
xG2 is post shot as well as location based. xG2 only considers attempts that are on target. Includes post shot attributes — Pace, swerve, deflections, placement. Therefore you can include such post shot things as how well the ball was struck, which part of the net it is heading for, is it swerving, deflected etc.
All off target attempts therefore have an xG2 of 0.
Usually xG & xG2 models align (the difference is sometimes equated with finishing ability).
So #LFC ‘s opponents finishing ability has plummeted in 2019/20.
From data analyst Dan Kennet (@DanKennett): “My untested theory would be sitting deeper with a lead allows you to be more compact and have more opportunity to apply significant defensive pressure to opposition shots. Plus having the fastest back 4 in Europe really helps.”
Dan also provided some additional information in view of his watching the videos -
35 big chances conceded because that’s 59% of the XGA (understat) 13.9 of 23.6 XGA are big chances, XGA/Shot = 0.4 (!!) only 7 of 35 opposition chances have been scored (20%) 3/9 at tied 1/14 at -1 (7% !!!)
Of the 26 BC not scored at 1/0/-1/-2 game states, 15 were from a cross of some kind (low or high): 13 missed, 2 saved 6 headers, 9 foot shots XG/Shot=0.43
19 big chances that weren’t goals (XG/Shot=0.44) 6/19 were complete mis-kicks/shins/fresh air shots under ZERO defensive pressure (XG/Shot = 0.52) Ings Saints A Dunk Brighton H Sarr Watford H McBurnie Sheff Utd H Pereira Man Utd H Jota Wolves A. Only 5 of the 19 big chances could be classed as the shooter being under “moderate” or “high” defensive pressure. If we are assuming that all MISSED, WOODWORK and BLOCKED shots have an XG2 of zero then thats 7.78 XG degraded.
Another idea came from John Harrison (@jhdharrison1) of @gkintelligence. In his analysis, John shows how Liverpool’s defenders seem to “channel” opposition shots into the central area of the goal, where the GK (in this case Alison) can make the (easier) saves. (More on this later)
The discussion changed to if we have seen this before -
Did Burnley did a Liverpool!?
It was considered that Burnley were doing something similar when they finished 7th.
Pope also overperformed against the shots that did need saving. ~3% chance an average keeper equaled or bettered his goals allowed.
However, team’s slowly worked them out, & they got less fortunate.
All dried up next season, plus xG allowed rose considerably & Burnley went from 7th to 15th dropping 14 points.
Then some additional data was provided by Mark — xG to xG2 at different score differentials.
“Cutting the data fine here, but there’s at least 50 attempts in each of 0, +1,+2. #LFC most defensively efficient when leading by 1.”
Here’s all #LFC PL games. 18 where their xG 2 differential has been better than their xG differential. Average improvement is ~0.7 xG per game. No one else comes close, next best is Villa & Newcastle at half that value. MC xG = xG2
Additionally, From Sports analytics and data science expert, Laurie Shaw’s (@EightyFivePoint) article — “Bodies on the Line: Quantifying how defenders affect chances.”
In summary this was mentions — The results quantify the difficulty of scoring against a packed defense. The probability of scoring a shot from outside a crowded penalty area is less than 5%, essentially wasting possession. A well-organised, ultra-defensive team can reduce the probability of scoring from outside the six-yard box to less than 20%, or one in five. In that situation, the ability to maneuver defenders out of position is clearly very important.
Additional read 1:
From John Harrison,
( https://twitter.com/Jhdharrison1/status/1212021073430097920?s=20
https://twitter.com/Jhdharrison1/status/1224749032939163649?s=20)
This is #Alisson’s #PremierLeague shotmap.
#Alisson has only faced 25 shots & only 6 have been within 12 yard of goal!
Contributing to #Alisson having the highest ExSave% in the #PL of 82% (ie he has faced the easiest shots)
#Liverpool’s defence has been epic!
Half of the shots on target they have allowed have PostShotExG values less than 0.08!
Meaning they are slow, weak, central shots that rarely go in, likely due to #Liverpool’s great shape & the pressure they put on opposition strikers.
Updated #PremierLeague shot stopper rankings!
This figure highlights how good #Liverpool are.
#LFC’s defence is so good #Alisson faces the easiest shots in the PL by far but even if teams get good shots off #Alisson saves far more of them than the average GK!
#GKStats #GKUnion
Additional read 2:
From Statsbomb’s James Yorke (@jair1970)
( https://twitter.com/jair1970/status/1224847849017565189?s=20)
Liverpool’s defensive record is provoking a lot of debate. A small contribution; look at the zone highlighted, basically Liverpool’s RCB zone. Around 40-45 shots, TWO on target. An Alderweireld header (saved) and a Richarlison goal. That’s ALL SEASON!
Now if this was planned, why isn’t it the same on the other side? A-ha! It’s not even about the goals/not goals here, but as a window to HOW Liverpool have exceeded expectation, it sure helps if the opponent can’t hit a barn door from any location right of their keeper.
To summarize:
1/3 To summarise #LFC ‘s defence. They are giving up 26 xG, but only 13 xG2 by the time these attempts reach or don’t reach the keeper. No such effect in 2018/19. Alisson’s been saving more xG2 than expected, Adrian not so.
2/3 The effect is most pronounced when #LFC are leading by one. It disappears at the extremes (small sample) & in the UCL (small sample).
3/3 These performance levels have greatly driven their 22 point lead…but we’ve seen this before. Specifically with Burnley, where it didn’t persist & they dropped a quarter of their points in the next season.
The tweet link of the original tweet:
https://twitter.com/MarkTaylor0/status/1224609151785095168?s=19
