Part 1 laid out what was different between the two managers. Part 2 is about why.
To get there, we ran correlation analysis across the full dataset — looking at how each metric moved in relation to others, separately for Amorim’s matches and Carrick’s matches. Two things stood out as particularly useful. First, cases where a metric showed a significant correlation under both managers but in opposite directions — meaning the same stat played a completely different role in each system. Second, cases where a correlation only showed up under one manager, pointing to something specific about how that side operated.
This isn’t about finding causation. Correlation analysis doesn’t tell you that one thing caused another. What it does tell you is which metrics moved together consistently, and whether that pattern looked the same or different under each manager. When the same metric behaves in opposite ways across two different regimes, that’s a strong signal that it meant something different in each system.
Part. 1
“The Numbers Don’t Lie.”
◀◀◀ Click to read!
[ The Attacking Pattern — How Did They Create? ]
( The Big Chances to Goal (%) story )
- The 21.6 percentage point gap in big chance conversion was the headline number from Part 1. Under Carrick, United converted 50.7% of their big chances. Under Amorim, just 29.1%. If you stopped there, you’d conclude Carrick’s side was simply more clinical. The correlation data tells a more complicated — and more interesting — story.
- Under Amorim, big chances to goal (%) moved in the same direction as xG (r=+0.536), goals scored (r=+0.682), and total duels (r=+0.490). On days when United converted big chances at a higher rate, they were also generating more expected goals, scoring more, and contesting more duels. Conversion rate, volume of chance creation, and overall attacking output were all tied together. Amorim’s side converted well on the days when everything was firing — when the press was working, the duels were being won, and chances were being created in quantity. The 29.1% figure looks low partly because there were simply more big chances in the denominator.
- Under Carrick, the picture was inverted. Big chances to goal (%) correlated negatively with both xG (r=-0.444) and goals scored (r=-0.404). Higher conversion rates came on days when expected goals were lower and fewer goals were scored overall. On top of that, PPDA showed a positive correlation with conversion rate (r=+0.405) — meaning Carrick’s side converted big chances more efficiently on days when they were pressing less. The high conversion rate wasn’t a product of attacking dominance. It was a product of discipline and focus in low-volume, transitional situations — most likely off the back of a defensive block and quick counter.
- The 21.6 percentage point difference between the two managers doesn’t mean one was more clinical than the other. It means they were operating entirely different attacking models. Amorim’s side created more and converted in proportion to that volume. Carrick’s side created less but concentrated their finishing in higher-pressure moments.
( What the duels were actually doing )
- Under Amorim, total duels correlated positively with goal difference (r=+0.477) and big chances to goal (%) (r=+0.490), and negatively with shots off target (r=-0.351). Days with more duels were days when United played better — better output, better conversion, cleaner shooting. The physicality wasn’t random noise; it was load-bearing. Fighting more meant playing better under Amorim.
- Under Carrick, it was the opposite. More duels correlated negatively with opponent danger possession lost (r=-0.381). High duel counts under Carrick weren’t a sign of United imposing themselves — they were a sign of United being pushed around. The same metric, completely different meaning.
( How they moved the ball forward )
- The relationship between duels won and accurate crosses revealed another structural difference.
- Under Amorim, winning more duels correlated negatively with accurate crosses (r=-0.403). When Amorim’s side was winning the physical battle, they weren’t crossing the ball — they were driving forward directly.
- Under Carrick, that correlation flipped to positive (r=+0.421). Winning ground duels led to wide progression and delivery into the box. Two different endpoints from the same starting point.
[ The Pressing Pattern — Intensity vs Efficiency ]
( Amorim’s press and what it produced )
- On days when Amorim’s side pressed harder — lower PPDA — goalkeeper high claims (r=+0.432) and long balls (r=+0.402) both increased. The press was forcing opponents into direct, hurried clearances, with the goalkeeper sweeping up the second balls. That’s a recognisable chain: high press → opponent forced long → aerial contest → goalkeeper claims.
- But there’s an important detail to layer in here. Amorim’s side also played more long balls themselves — 53.05 per game compared to Carrick’s 46.82, a statistically significant gap. This wasn’t just about forcing the opponent to go long. Amorim was using long balls as a direct attacking tool, bypassing the build-up phase and targeting the final third immediately. The higher rate of fouls won in the final third (2.25 vs 1.41) sits alongside this — United were getting into advanced areas and drawing contact, not just recycling possession.
- The pressing also came with a cost. On days when Amorim’s side won more opponent danger possessions, errors leading to shots (r=+0.437) also increased. The high-intensity approach created turnovers at one end and vulnerability at the other. That’s not a flaw unique to Amorim — it’s the inherent trade-off in aggressive pressing football.
( The paradox at the heart of Carrick’s system )
- Here’s the number that should make you stop and think. PPDA was lower — meaning more intense pressing — under Amorim (9.04 vs 11.65). But opponent high turnovers were higher under Carrick (5.06 vs 3.70 per game).
- Amorim pressed harder and still generated fewer total opponent turnovers. Carrick pressed less aggressively and generated more. That’s not a contradiction — it’s a distinction between pressing intensity and pressing organisation. Amorim’s press was front-loaded, concentrated in dangerous areas, and produced more opponent danger possession lost (24.3 vs 18.94). Carrick’s press was more structured in the mid-block, less aggressive in terms of PPDA, but more effective at producing turnovers across the whole pitch.
- The consequences of this in Carrick‘s system were significant. When the press worked — when opponent danger possession lost was high — goal difference (r=+0.527) and pass accuracy (r=+0.499) both rose together. One metric was pulling the whole game quality upward. When it didn’t work, the data showed a different pattern: long balls (r=-0.588), interceptions (r=-0.555), and aerial duels (r=-0.452) all increased as opponent danger possession lost dropped. Pressing failure led directly to a more reactive, deeper defensive posture.
( Pressing and fouling in the final third )
- One correlation that only appeared under Carrick: opponent build-up (%) and fouls won in the final third moved in opposite directions (r=-0.480). On days when United were pressing well and limiting opponent build-up, they were also winning more fouls in advanced areas. Under Carrick, fouls won in the final third weren’t a product of desperate play — they were a byproduct of the press functioning. Under Amorim, this relationship didn’t exist at all.
[ The Defensive Pattern — How Did They Defend? ]
( Aerial duels and what they signalled )
- Under Amorim, aerial duels and ground duels moved in opposite directions (r=-0.459). High-aerial-duel games were low-ground-duel games, and vice versa. The two modes of physical engagement were separated, suggesting the team shifted between aerial-contest-heavy and ground-battle-heavy approaches depending on the opponent or game state.
- Under Carrick, aerial duels correlated negatively with opponent danger possession lost (r=-0.452). More aerial duels meant fewer opponent turnovers in dangerous areas — meaning aerial contest volume was a signal of defensive stress, not tactical intent. When opponents were winning the press battle and getting into advanced areas, Carrick’s side was dealing with it aerially. It wasn’t a choice; it was a response.
( What happened after interceptions )
- Interceptions were higher under Carrick (9.53 vs 7.85 per game). But the direction of play after those interceptions was completely different under the two managers.
- Under Amorim, more interceptions correlated with fewer shots inside the box (r=-0.487). Winning the ball back didn’t immediately translate into attacking threat. The interception was a defensive act that ended the danger rather than starting something going the other way.
- Under Carrick, the correlation flipped: more interceptions correlated with more shots inside the box (r=+0.463). Winning the ball in midfield led directly to box entries. The transition from defensive action to attacking threat was immediate.
- This is probably the single clearest illustration of how differently the two systems were wired — the same action, interception, feeding into completely opposite outcomes.
( Blocking and how it was done )
- Blocked opponent shots were higher under Carrick (4.41 vs 2.95 per game). Under Amorim, days with more blocks also saw more opponent turnovers (r=+0.436) and more interceptions (r=+0.386) — the blocks were part of a broader picture of the defensive unit functioning well collectively.
- Under Carrick, more blocks correlated with fewer ground duels (r=-0.468). The blocking was happening through positioning and shape rather than physical engagement. Carrick’s defenders were getting into the right places to put bodies in the way, not winning individual ground battles to get there.
[ What It’s Telling Us ]
- The correlation data draws two fairly distinct tactical profiles.
- Amorim built a system around intensity. The press was front-loaded and aggressive. The attacking play bypassed build-up in favour of direct progression through long balls and ground duels. Duels won meant United were in control. Interceptions happened in defensive moments. Everything ran at a high physical pitch, and the output — chances created, goals scored — reflected that volume.
- Carrick built a system around organisation and transition. The press was less intense but more structured, generating more total turnovers across the pitch. The team defended with shape rather than aggression, and when they won the ball back, they moved quickly. Interceptions were the start of attacks, not the end of them. The output was smaller in volume but higher in efficiency.
- What makes this particularly interesting is how many of the same metrics — duels, interceptions, long balls, aerial contests — appear in both profiles but mean entirely different things. Under Amorim, a high-duel game was a good game. Under Carrick, it was a bad one. Under Amorim, interceptions were defensive. Under Carrick, they were offensive. The language of the statistics looks the same on the surface. The underlying meaning was almost entirely different.
- Part 3 pulls this together and asks what it all means for United.
( attacking pattern )
| Big Chances to Goal (%) | Correlation (Amorim) | Correlation (Carrick) |
| xG | + 0.536 | – 0.444 |
| Scored | + 0.682 | – 0.404 |
| Total Duels | + 0.490 | . |
| Opponent High Turnovers | – 0.440 | . |
| PPDA | . | + 0.405 |
| Total Duels | Correlation (Amorim) | Correlation (Carrick) |
| Big Chances to Goal (%) | + 0.490 | . |
| Goal Difference | + 0.477 | . |
| Shots off Target | – 0.351 | . |
| Opponent Danger Possession Lost | . | – 0.381 |
| Duels Won | Correlation (Amorim) | Correlation (Carrick) |
| Accurate Crosses | – 0.403 | + 0.421 |
( pressing pattern )
| PPDA | Correlation (Amorim) | Correlation (Carrick) |
| High Claims | + 0.432 | . |
| Long Balls | + 0.402 | . |
| Big Chances to Goal (%) | . | + 0.405 |
| Opponent Danger Possession Lost | Correlation (Amorim) | Correlation (Carrick) |
| Errors Leading to Shot | + 0.437 | . |
| Long Balls | . | – 0.588 |
| Interceptions | . | – 0.555 |
| Aerial Duels | . | – 0.452 |
| Goal Difference | . | + 0.527 |
| Accurate Passes (%) | . | + 0.499 |
| Opponent Build-up (%) | Correlation (Amorim) | Correlation (Carrick) |
| Fouled in Final Third | . | – 0.480 |
( defensive pattern )
| Aerial Duels | Correlation (Amorim) | Correlation (Carrick) |
| Ground Duels | – 0.459 | . |
| Aerial Duels Won (%) | + 0.334 | . |
| Opponent Danger Possession Lost | . | – 0.452 |
| Interceptions | Correlation (Amorim) | Correlation (Carrick) |
| Shots Inside Box | – 0.487 | + 0.463 |
| Blocked Opponent Shots | Correlation (Amorim) | Correlation (Carrick) |
| Opponent High Turnovers | + 0.436 | . |
| Interceptions | + 0.386 | . |
| Ground Duels | . | – 0.468 |
Raw per-match data was collected from Sofascore and MARKSTATS.
Aggregate metrics — including per-game averages, cumulative totals, and derived indicators such as Accurate Passes (%) and Total Duels — were constructed independently by the author.
All statistical analysis, including Welch’s t-test and Pearson correlation, was conducted by the author using this dataset.
Leave a Reply