Why Celtic's 'new normal' could leave Carl Starfelt exposed in defence

For example, some continue to reference defending as a weakness this season so far, yet Celtic have conceded the fewest amount of the following: goals, non-penalty xG, xG from open play, and even xG from set-pieces. Not too shabby for a team with supposed defending issues.

One of the narratives amongst the support seems to have been the growing defensive partnership between Cameron Carter-Vickers and Carl Starfelt. In particular, Starfelt had gotten off to a rough start but performances settled as autumn arrived, which preceded the reemergence of some poor games towards the end of the year.

While the defensive record has been very good in the league, and the defensive partnership did enjoy a good stretch in September up until Starfelt’s subsequent absence, there is concern about the future.

Callum McGregor stated in a press conference this past autumn that Celtic had made some tactical adjustments. It is mere speculation, but one adjustment may have been how aggressive the centre-back pairing was playing relative to having taken more advanced positions.

Over his first five league games, Starfelt’s average defensive distance, which StatsBomb does on a scale between 0-100, was 29.78. In the subsequent six games prior to his injury, and his first upon return versus Hearts on December 2, that distance fell significantly to 24.6.

Here is a scatter plot with Defensive Distance compared with Defensive Action OBV (all graphs are showing StatsBomb data):

We have added a linear regression line, which obviously trends down to the right. Season to date, his average defensive distance in all league games worked out to 28.10. If we look at just the games where it was higher than that average, his defensive action OBV was negative in six of seven of those games, versus just four out of 10 negatives in games below the 28.10 average.

How have his colleagues faired when looking at the same metrics? Here were Carter-Vickers and Stephen Welsh:

In fairness to all these players, the sample sizes are not large and the relative statistical relationships between the two variables differ for each. Obviously, Welsh’s game data is very limited, but he did show a similar decline in defensive performance output as his average distance increased. Carter-Vickers' data stands out as we can see his performances have been quite good even as the average distance increased.

READ MORE: Who should start at left-back for Celtic? Greg Taylor and Josip Juranovic analysed

Of course, defending is a vital part of centre-back play, but at Celtic in Ange Postecoglou’s system, they are also on the ball a lot.

Here were Starfelt’s passing OBV and passing distance:

This makes some intuitive sense, as OBV is intended to measure the amount each pass increases the odds of a team scoring a goal, so this relationship really should have an upward slope for players. Starfelt’s passing has not offered that much value relative to how much he passes, averaging 0.08 passing OBV so far on 87.2 passes from open play.

Here were Carter-Vickers and Welsh:

We can see they similarly have upward regression lines, but the slopes are much steeper than Starfelt’s. So far this season, Carter-Vickers has averaged 0.10 OBV on 83.63 passes and Welsh all the way up at 0.21 on 90.87 passes.

However, are all these just statistical noise? Let’s take a look at some other recent centre-backs who may not have been considered ‘ball-playing’ in a high line. Remember Shane Duffy last season?

His scatter plots were a bit all over the place, though both regression lines trend downward which was particularly noteworthy relative to passing. His average defensive line was 26.07 in games he started.

Here was Dedryck Boyata in the 2018-2019 season:

Boyata’s defending plot showed he was not deployed as advanced, and that was reflected in his season average of just 22.90. Also noteworthy was that he had just four out of 19 games with a negative defensive OBV. Even good old Boyata, famed for not being a very good passer, had a regression line slope similar to Starfelt, but with an average passing OBV of 0.09 on 79.03 passes.

With Celtic having done a lot of exciting transfer business already for the January window, and with the players seemingly recruited with playing ‘Ange-ball’ a priority, Monday’s game against Hibernian may be a sort of preview of what may be in store.

Starfelt’s average defensive distance was all the way up at 41.18, with a defensive OBV of -0.01 and a passing OBV of just 0.05.

Best guess? The degree of high line versus Hibs is more likely to be the ‘new normal' - so here’s hoping that all of this is in fact just statistical noise.