Considering player defensive performances in the 2-0 Derby win over The Rangers.
(Relatively) Easy Money
Firstly, Gordon had little to do to earn his win bonus. He caught 2 crosses and made one save, but it was a key defensive action.
The Expected Goal model I use calculates the goalkeeper’s xG separately. From Gordon’s perspective, the Morelos header had an xG of 0.382, which is the same as Gordon’s average xG saves per 90 minutes.
As far as defensive actions are concerned for Gordon, that was it!
The high press restricted his distribution with only 11 passes completed and 2 were incomplete. He was also forced into 4 clearances. From restarts he only found a fellow Hoop with 6 out of 11 attempts. The home side effectively pressed the Celtic box. It should be noted, however, that the well-rehearsed long “out” ball to Lustig wide right led to the second goal as the Swede’s winning header found McGregor then Roberts to assist Griffiths.
Defensive Action Success Rate (DASR)
The DASR rating aggregates defensive actions to assess overall defensive effectiveness. See the Glossary for a more precise definition.
This was the Boyata / Simunovic axis first game of the season. And there was rustiness. Both eager to impress, and were guilty of trying to win challenges they could not, usually trying to get around the bowling ball shaped Morelos. But no defensive errors and another clean sheet.
The centre backs should have the highest DASR as they should take the least risks and not have their ratings degraded by being tackled in advanced positions (see Tierney, K.). And the numbers were good. Simunovic leads the season DASR ratings with a season average of 76%. Therefore 90% and 89% is excellent in a single game.
Lustig’s performance must be seen with the additional context of his attacking output which I will cover separately.
Boyata, getting back up to match speed, was the least effective, but generally solid.
The home side’s high press was quite effective and 31 team clearances ensued – virtually guaranteeing possession will be turned over.
Apart from the raw Morelos, it is difficult to see where goals will come from for The Rangers.
As for the others:
Normally Brown has a better DASR than some of the defenders. The loud and insistent calls for Brown to be Man of the Match are surprising given this defensive performance. Apart from the Linfield game away where there was virtually no defending to be done, this is by some distance Brown’s “worst” defensive performance of the season. He managed to win 50% of his challenges against Hearts on the opening day.
He did not have one of the defensive error lapses that have started to creep into his play – usually involving being caught in possession deep and central. And it wasn’t one thing. Breaking down his 10 unsuccessful challenges and fouls, he:
- lost 4 tackles but Celtic kept possession from 3 of them
- lost 4 aerial challenges but Celtic picked up possession from 2 of them
- missed one tackle
- conceded one foul
This is 3 more than any other game this season (7 lost away to RK Rosenborg).
Armstrong’s 50% is on par with his 51% average, but it is unusual for him to be leading the midfield DASR stats. Surprising is Griffiths who rarely wins a challenge. 5 won doubles his challenge successes for the season!
Griffiths successfully tackles the corner flag.
Possessions Won and Lost
Whereas DASR is concerned with winning defensive actions, whether possession is maintained or not (e.g. a winning tackle putting the ball out for a corner). Possessions Won and Lost simply considers the actions concerned with winning or losing possession. See the Glossary again for a full explanation.
Simunovic leads the way again, with 8 possessions won, 3 more than any other player. Boyata can be commended for not losing any possessions from his defensive actions.
Passing and Overall Possession Effectiveness
Another reason Brown was not Man of the Match was that he usually dominates the ball, averaging 90 completed passes per 90 minutes. He was very accurate with his passing, but both the defenders, and Armstrong, dominated the ball.
Usually the defenders would be top of these rating, but it was noticeable that Forrest and McGregor coming on closed the game out for Celtic. They both added the ability to run at the home side, pinning them back, and both are reliable passers of the ball. Forrest did not give up a single pass.
Armstrong is usually more adventurous in his passing, taking more risk, and I will show his worth in this regard when I look at the attacking data. 92% passing success is high for him, he averages 88%. But with Rogic in the side, Armstrong had to provide the midfield solidity alongside Brown.
Further evidence, if needed, that Griffiths game doesn’t really revolve around building play. His value will become clearer when I consider the attacking data in the next piece.
Usage Rate
Finally, the Usage Rate shows the % of total possession each player had.
Defenders are usually more dominant possession wise, and in particular, the Celtic centre backs enjoyed a lot of possession – taking a rest with the ball as Rodgers calls it.
Noticeable again that Armstrong was the dominant midfielder, with Brown failing to hit 10% for only the second time this season.
Conclusion
Simunovic was the tidiest and most effective defender on the day. Lustig was a fraction behind for me. Armstrong performed a more conventional central midfield role admirably, out shining the reliable Brown.
I hope this shows why the Brown for Man of the Match calls are wide of the mark. The stats don’t capture his leadership skills, there is no metric for the effectiveness of his on-pitch exhortations and quality of his encouragement. Nor was his challenge with the home manager recorded as a challenge won! Brown had a very good match, and his Iniesta moment in the first half was magical. But there were better individual performances.
A reminder that an analysis of the team performances in this match is here -> Derby Delight for Dominant Celtic
SFTB says
If Celtic kept possession in 3 of the 4 instnaces where Broony “lost” a challenge, should the metric of “possession retained after a tackle” be considered the more telling metric?
After all, if the “lost” challenge” forces the “winner” of the challenge to lose control of the ball, then the pack of players harrying the opposition winning player have “won” the challenge even if the individual primary tackler did not do so.
celticbynumbers@btinternet.com says
You make a good point. This starts to stray into the area of chaining events together which is a fantastic concept but would take even more hours than it already does! The event based recording of on field actions is unsophisticated for this reason but I take the view that if you build enough data history real meaningful trends emerge. Brown s data was surprising but it did happen!
I do think people are influenced by 1 or 2 stand out events then fit their summary around that hook. Even watching in real time I thought he was a bit subdued. But Armstrong did a lot of the leg work.
I will have another think about your point though to make sure the data point is doing what it should.
A.p.g says
Could you use a number of increasing multipliers based on defensive risk for each incident?
A quick crude example: x90 for event leading to loss of goal, x32 for event leading to missed attempt at goal, x9 for loss of possession, x1 for loss of ball but team ball retention.
This might be a quick way to account for event chaining.