There is a quiet paradox at the heart of defensive statistics: the player at the top of the tackles chart is frequently not the best defender on the pitch but one of the worst-positioned. A tackle is a recovery from a situation that, ideally, should never have arisen. The defender who reads the danger early, steps across, and forces the attacker to play backwards registers nothing at all: no tackle, no interception, no entry in the box score. Raw defensive counts reward the cure and ignore the prevention, and that single fact is enough to make most casual readings of them wrong.
Tackles and interceptions vs possession: 760 Premier League team-matches, 2025-26
We tested that idea on a full season. ESPN's box score for every 2025-26 Premier League match carries, for each side, tackles attempted, tackles won, interceptions and a possession percentage (the two sides add to 100). That gives 760 team-matches (380 matches, two clubs each). We added tackles attempted and interceptions into one count of defensive actions; clearances and blocks are left out. The average side made 24.9 a match: 16.7 tackles and 8.3 interceptions.
Less of the ball does mean more defensive actions, but less than you might think. Across the 760 team-matches the correlation with possession is r = -0.30, and the fitted line falls by 1.8 actions per 10 points of possession: about 27.7 at 35% of the ball, 22.2 at 65%. The side with less of the ball made more tackles and interceptions in 250 of the 380 matches (66%); the counts were level in 15. Possession explains only about 9% of the match-to-match variation. The rest is the opponent, the score and the style of play.
The usual adjustment overcorrects. The standard possession adjustment (the one on our possession-adjusted stats page) asks: how many actions would this side have made if the opponent had held the ball exactly half the time? It multiplies the count by 50 and divides by the opponent's possession. That assumes actions rise in proportion to the opponent's share of the ball, which here means 5.0 fewer per 10 points of possession, against the 1.8 the matches actually show. So the adjusted rate swings the other way: it correlates r = +0.50 with possession at match level. Raw counts flatter the sides without the ball; the adjusted rate flatters the sides with it.
Neither number tells you who won. We split the 760 team-matches into thirds on each measure and compared results. The busiest third by raw count won 38.5% of its matches, the quietest third 36.4%. On the adjusted rate the gap is just as small, and the same holds for a third measure, actions above or below the fitted line for that much possession. Correlations with points are 0.01 (raw), 0.00 (adjusted) and 0.01 (above the line). The raw count does not track goals conceded either (r = 0.03).
| Measure | Lowest third | Middle third | Highest third |
|---|---|---|---|
| Tackles + interceptions | 36.4% / 1.36 | 34.3% / 1.33 | 38.5% / 1.41 |
| Possession-adjusted rate | 34.8% / 1.33 | 38.6% / 1.40 | 35.6% / 1.36 |
| Above or below the trend line | 36.0% / 1.36 | 36.2% / 1.36 | 36.8% / 1.37 |
| Possession (for comparison) | 35.8% / 1.38 | 39.1% / 1.38 | 34.0% / 1.32 |
Over a season, the raw count mostly ranks who lacked the ball. Across the 20 clubs, raw actions per game correlate r = -0.57 with average possession and -0.52 with points: the busiest defences belonged to sides that had less of the ball and finished lower. The adjusted rate turns that round (+0.63 with possession, +0.37 with points), mostly because possession and points move together (r = 0.74). The biggest movers when the count is adjusted: Manchester City from 17th to 3rd, Chelsea 12th to 1st, Liverpool 20th to 10th, West Ham 5th to 16th and Wolves 2nd to 12th.
| Club | Possession | Raw per game (rank) | Adjusted (rank) | Move | Points |
|---|---|---|---|---|---|
| Tottenham Hotspur | 50.4% | 27.9 (1) | 28.1 (4) | -3 | 41 |
| Wolverhampton Wanderers | 42.7% | 27.6 (2) | 24.1 (12) | -10 | 20 |
| AFC Bournemouth | 50.1% | 26.8 (3) | 26.9 (6) | -3 | 57 |
| Crystal Palace | 45.9% | 26.8 (4) | 24.8 (9) | -5 | 45 |
| Manchester United | 51.8% | 26.8 (5) | 27.8 (5) | 0 | 71 |
| West Ham United | 42.5% | 26.8 (5) | 23.3 (16) | -11 | 39 |
| Everton | 43.7% | 26.7 (7) | 23.7 (13) | -6 | 49 |
| Leeds United | 45.8% | 26.3 (8) | 24.3 (11) | -3 | 47 |
| Brighton & Hove Albion | 54.1% | 26.3 (9) | 28.6 (2) | +7 | 53 |
| Burnley | 42.3% | 25.8 (10) | 22.4 (18) | -8 | 22 |
| Fulham | 51.7% | 25.7 (11) | 26.6 (7) | +4 | 52 |
| Chelsea | 57.7% | 25.3 (12) | 29.9 (1) | +11 | 52 |
| Nottingham Forest | 46.8% | 24.9 (13) | 23.4 (15) | -2 | 44 |
| Brentford | 47.6% | 23.3 (14) | 22.2 (19) | -5 | 53 |
| Sunderland | 44.9% | 23.3 (15) | 21.1 (20) | -5 | 54 |
| Arsenal | 56.2% | 22.9 (16) | 26.1 (8) | +8 | 85 |
| Manchester City | 60.5% | 22.3 (17) | 28.3 (3) | +14 | 78 |
| Newcastle United | 52.6% | 21.8 (18) | 22.9 (17) | +1 | 49 |
| Aston Villa | 53.6% | 21.7 (19) | 23.4 (14) | +5 | 65 |
| Liverpool | 59.3% | 19.9 (20) | 24.5 (10) | +10 | 60 |
What this means. Raw counts do lean toward measuring how little of the ball a side had, clearly over a season and weakly in a single match. Scaling them to 50% opponent possession does not turn them into a measure of good defending; it overshoots and starts rewarding possession instead. In this season neither version, nor the count relative to the fitted line, separated the sides that won from the sides that lost. Tackles and interceptions describe how a match was played, not how well a team defended.
Limits. One season, one league and one provider: ESPN's definitions of a tackle and an interception, and its possession figure, which we did not check against another source. Team totals hide which players made the actions and where on the pitch. Match stats are also shaped by the score: a side that leads early often drops deeper and defends more, which blurs any link between defensive volume and results.
What a tackle count actually measures
Start with what has to happen for a tackle to be recorded. An opponent must have the ball, in your zone, in a position dangerous enough that someone has to physically challenge for it. That is already a loaded set of conditions. A defender who is constantly making tackles is, by definition, a defender whose team keeps allowing opponents to carry the ball into challenging positions. Sometimes that is heroism. Just as often it is a symptom: of a high line being run in behind, of a midfield that does not screen, of a full-back repeatedly isolated against a winger.
The same logic applies to interceptions and clearances, with a twist. An interception requires the opponent to attempt a pass into the space you are occupying; a clearance requires the ball to reach your defensive third in the first place. A team that defends deep, invites pressure, and lives on the edge of its own box will rack up interceptions and clearances by the dozen, not because its defenders are exceptional readers of the game but because the ball keeps arriving in their penalty area. Volume here is a story about territory conceded, not quality applied.
None of this means the counts are worthless. It means they are descriptive, not evaluative. They tell you what a defender did and roughly where, but the leap from "made many tackles" to "defends well" is exactly the kind of inference the number cannot support on its own.
The possession problem, again
Layered on top of positioning is a structural bias that affects every defensive metric equally: the amount of the ball your team has. Defensive actions can only happen when the opposition is in possession. A side that controls sixty-five percent of the ball simply hands its defenders far fewer opportunities to tackle, intercept, or block than a side pinned back with thirty-five percent. The low-possession team will usually post bigger raw defensive numbers (it did in 250 of the 380 Premier League matches of 2025-26), and little of that gap is about defensive ability.
This is the most visible confound in the category, and the standard fix is possession-adjustment: scaling each team's raw count to a neutral fifty-fifty baseline so that two sides with wildly different styles can be compared on the same footing. We work through the formula and a full example in possession-adjusted stats. Before you compare any two teams' defensive volumes, ask how much each side actually had to defend. Unadjusted, a season tackle leaderboard leans toward a ranking of who saw the least of the ball (r = -0.57 across the 20 Premier League clubs of 2025-26). But the 2025-26 numbers above show the proportional fix overshooting: counts fell far more slowly than the opponent's share of the ball, so the adjusted figures end up favouring the sides with the most possession. Read the adjusted column as a correction, not as a verdict.
Success rate beats attempts
If the count of attempts is a weak signal, the rate of success inside those attempts is a more informative one for judging a player. Tackle success rate, tackles won as a share of tackles attempted, tells you something the raw tally cannot: when this defender does commit to a challenge, how often does it work? A player making six tackles a game at a fifty percent success rate is diving in, getting beaten half the time, and leaving space behind. A player making three tackles a game at eighty percent is picking moments and winning them.
The deeper point is that a low attempt count paired with a high success rate is often the marker of a genuinely good defender. Such a player is not avoiding duels out of timidity; they are positioning so well that the duels they do enter are ones they have already half-won before contact. The metric to be suspicious of is a high volume of attempts, whatever the success rate: it usually means the defender is reacting rather than anticipating.
One caution from the team-level data: across the 20 Premier League clubs of 2025-26, a club's share of tackles won had no link to its points (r = -0.03), so success rate is a lens on individual defenders, not a shortcut to judging a team's defence.
A tale of two centre-backs: Souttar and Rodri at the 2022 World Cup
To see how the raw count can mislead for one player, take two real centre-backs from the 2022 World Cup, measured from StatsBomb Open Data (StatsBomb Public Data User Agreement): Australia’s Harry Souttar and Spain’s Rodri (Rodrigo Hernández, usually a defensive midfielder, who played the tournament at centre-back). Both played all four of their side’s matches. Every figure is counted only while the player was on the pitch.
| Metric | Souttar | Rodri |
|---|---|---|
| Minutes | 387 | 414 |
| Tackles + interceptions per 90 (rank) | 2.33 (14) | 1.74 (21) |
| Per 100 opponent passes (rank) | 0.36 (21) | 0.65 (6) |
| Tackles won | 2 of 7 (29%) | 3 of 4 (75%) |
| Team share of the passes | 38% | 76% |
| Opponent shots per 90 | 15.1 | 5.0 |
Read only the per-90 row and Souttar is the busier defender: 2.33 actions per 90 to 1.74. Read the rest and the picture changes. Australia had 38% of the passes while Souttar was on the pitch and faced 15.1 shots per 90; Spain had 76% and faced 5.0. Souttar had far more opposition possession to defend against, and he won 2 of his 7 tackles. Scale both counts by the opponent’s passes and the order flips: Rodri made 0.65 tackles and interceptions per 100 opponent passes, sixth of the 27 centre-backs with 360 or more minutes, against Souttar’s 0.36, 21st. Spain rarely let the ball near Rodri, and when it came, he engaged more often per opponent pass than most. The samples are small, seven tackles and four, so the success rates are a description of four matches, not a judgement of either player.
The wider group shows the same two pulls as the Premier League figures above. Across the 27 centre-backs, tackles and interceptions per 90 correlate r = -0.17 with the team’s share of the passes while they played, and the count per 100 opponent passes correlates r = +0.34: the raw rate leans toward the sides without the ball, the adjusted one toward the sides with it. Busier centre-backs did not win fewer of their tackles (r = +0.15 between the per-90 count and the tackle success rate). And the count also reflects the job a player is given. France’s Dayot Upamecano made the most of any centre-back, 4.52 per 90, while his partner Raphaël Varane made 1.16, 26th of 27, in the same team, with France at 51% and 47% of the passes in their minutes. Two partners in the same back line, facing the same opponents, differed almost fourfold: the count reflects a role as much as a level of defending.
Defensive actions are a location story
The most useful thing about defensive-action data is not how many but where. A tackle won in the opponent's third is a completely different event from a clearance hacked off your own goal line, even though a naive count treats them as one apiece. Mapping where a team wins the ball tells you what kind of defending it is actually doing: a cluster of recoveries high up the pitch signals an aggressive press; a band of actions hugging the edge of the penalty area signals a deep, reactive block.
This is precisely the intuition that pressing metrics formalise. PPDA, passes allowed per defensive action, restricts itself to defensive actions in the opponent's build-up zone for exactly this reason: it is trying to isolate the high, ball-hunting behaviour and strip out the deep-block volume that would otherwise drown it. We unpack the construction in PPDA explained. The lesson carries straight back to individual defenders: the value of a defensive action depends overwhelmingly on the pitch location and game context it happened in, and a leaderboard that collapses location into a single total throws away the most informative dimension it had.
How to read defensive numbers honestly
None of this is an argument for ignoring defensive stats. It is an argument for reading them in the right order. Start by asking how much of the ball the team had, and prefer possession-adjusted figures over raw ones, while checking that the adjustment has not simply swapped one bias for the other. Then look at success rate before volume, because rate survives the positioning problem that volume does not. Then ask where the actions happened, because a recovery in the final third and a clearance off the line are not the same currency. Only after all three filters does a raw count start to mean what a casual reader assumes it means at first glance.
Done in that order, the numbers genuinely help. They can flag a full-back being targeted, a midfield that has stopped screening, a centre-back whose engagement has quietly dropped after a transfer. Skipped, they will do the opposite: crowning the most over-worked, worst-protected defender in the league as its best, simply because the ball kept finding them and they kept having to react. The tackle is the most visible defensive act and the least informative in isolation. The defender who never has to make one has usually already done the harder work.
Reproduce it
Every figure in the 2025-26 section comes from our analysis script. It reads data_layer/epl_2025_26_boxscores.json, ESPN's public Premier League box scores for all 380 matches of 2025-26, pulled on 7 September 2026, through our archive loader, which turns ESPN's display strings into numbers. It fits the trend line, computes the adjusted rates, splits the team-matches into thirds, builds the club table, writes every number to data/stats/defensive-actions-tackles-interceptions-espn.json and draws the chart. No network, nothing hand-entered.
The centre-back comparison comes from a second analysis script, which reads StatsBomb Open Data for the 64 matches of the 2022 World Cup (the match list), retrieved June 4, 2026, under the StatsBomb Public Data User Agreement. It works out each player’s minutes from the event clock (starting line-ups, substitutions and red cards, stoppage time included), counts tackles (StatsBomb tackle duels) and interceptions, and counts both teams’ passes and the opponent’s shots while he was on the pitch. A centre-back is a player whose most frequent recorded position is one of StatsBomb’s three centre-back roles. The figures are in data/stats/defensive-actions-tackles-interceptions-wc2022.json.
The data behind this piece
- 2025-26 section (added November 5, 2026): all 380 Premier League matches of 2025-26 with ESPN's per-team box scores, /data_layer/epl_2025_26_boxscores.json. ESPN public scoreboard and summary feeds (soccer/eng.1), pulled 7 September 2026.
- FBref: tackles, interceptions, clearances and tackle-success splits (via Opta) for the major competitions.
- StatsBomb: documentation on defensive event definitions and the difference between contesting and recovering possession.
- Centre-back section (added November 7, 2026): StatsBomb Open Data, the 2022 World Cup event files for all 64 matches (StatsBomb Public Data User Agreement), retrieved June 4, 2026.
- StatsBomb open data: event-level data with locations, for mapping where defensive actions actually happen.
- Understat: season-level data for the major European leagues; further reading only, no figure on this page comes from it.
Published by SoccerAnalytics from public data and the sources named in this piece. How this site is made.