Source: Manually tracked

Weight: 100%

This is where I bash the player for any time he was caught sleeping; specifically whenever he caused a goal or a major chance via poor positioning, loose marking or lacking gap control (not applying pressure on a shooter/passer soon or urgently enough); a special co-efficient is applied on top of this to account for goal-causing errors. To be clear, I am no seasoned coach or even a tactics experts, so I might be off on a lot of these situation, but across the full season, I'd argue it evens out.

Source: Wyscout.hudl.com

Weight: 100%

Pretty straightforward; with an added co-efficient applied to bring out penalty fouls.

Source: Driblab

Weight: 75%

For years, I have stubbornly ignored Wyscout's "possession-adjusted tackles" metric since it only looked at one side of the coin. Enter Driblab, who bring us this metric, comparing the amount of tackles to the number of times the player was dribbled past (Virgil Van Dijk would never!).

Source: Driblab

Weight: 100%

Per Driblab's own glossary, this communicates the "sum of all bad actions committed (pass unsuccessful, take on unsuccessful, dispossessed, offside pass, ball touch unsuccessful) in own half".

Source: Wyscout.hudl.com

Weight: 100%

To filter out noise that doesn't matter, I'm concerned purely with ground duel success rate in here.

Source: Manually tracked

Weight: 100%

I assign up to 4 goal-allowing slots to any goal conceded across the season. There is one slot reserved for a grave fuck-up and three extra ones for more run-of-the-mill mistakes like losing a key aerial/ground duel, getting peeled off by a ball carrier, etc. This metric then speaks to how big a percentage of goals the defensive midfielder "caused" — only accounting for goals against he's on for.

Source: Manually tracked

Weight: 75%

See "Blame share for goals conceded" and extend it to missed opportunities. I may only refer to it at extremes, which were further off each other in 25/26 than a year ago, by the way (3.7% vs 18.8%).

Source: Wyscout.hudl.com / manually tracked (for broken build-up leading to danger)

Weight: 75%

This is an interesting metric in that it accounts for both misplaced passes and lost aerial or ground duels. The follow-up shot needs to occur inside the next 20 seconds per Wyscout rules, otherwise it’s rendered a “harmless” loss. To extra punish those who lose possession more often (when the ball is under their actual control), I apply an additional co-efficient accounting for any major chance/goal caused via a squandered pass or getting dispossessed while building from the back.

Source: Manually tracked

Weight: 50%

Instead of relying on Wyscout's clearances (sometimes bizarrely tagged deep into the attacking half) or vague "successful defensive actions", I rely on my own set of eyes and track truly vital defensive actions by myself. For this metric, I combine one's balance in damage control (vital clearances or cases of backtracking versus those missed ones) with the number of vital blocks produced.

Source: Manually tracked

Weight: 50%

I have notes on all tangible danger occurring in each game, detailing culprits of all goals and chances allowed. For goals, these can be up to 4; for chances a maximum of 2. Similarly, any prevented threat gets assigned the "saviour". Once you put all starts together, how often does a defensive midfielder appear on the right side of the history? Given that there are more slots to go around for culprits rather than saviours, while a DM also sees less opportunity to act as a saviour than a CB, there were only 3 net-positive DMs in 25/26, but there is — of course — a garden variety of net-negative.

Source: Manually tracked

As previously noted, I assign up to 4 goal-allowing slots to any goal conceded across the season. This index then combines the percentage of goals the player "caused" — only accounting for goals against he is on the pitch for, to be fair — and the goals conceded per 90 mins rate with him around, as well.

Source: Driblab

Weight: 100%

Previously, I considered success in any aerial duel inside own half. Now it's purely defensive duels, but anywhere. This is allowed by the addition of Driblab database that allows for this distinction.

Source: Wyscout.hudl.com

Weight: 100%

As opposed to elsewhere, I am not too concerned with the success rate here; I simply want to see how many races — shoves and pulls — for a free ball in no one's possession the player in question successfully navigates per game. That nicely speaks to both positional sense and agility.

Source: Wyscout.hudl.com

Weight: 100%

Interceptions alone don’t tell the full story — if your team dominates possession, you barely get the chance to make any. Wyscout fixes that by normalizing for context: they assume every team spends an equal 30 minutes per game out of possession (based on a 60-minute average of actual ball-in-play time). That way, players from possession-heavy teams aren’t punished for simply doing their job — which is how guys like Mannsverk or Baráth still manage to land in the 90th+ percentile.

Source: Wyscout.hudl.com / manually tracked (forced high turnovers leading to tangible danger)

Weight: 100%

A standard feature: I take balls recovered in the final third (Wyscout) and add a weighted cherry on top in the form of forced turnovers leading to goals/chances. This add a bit of a purposeful touch.

Source: Driblab

Weight: 50%

While not the most important aspect per weighting, Driblab offers a nice selection of the vital off-ball movement metrics that belongs, in some form, on the pizza chart. This one specifically speaks to any time the player moves to "align with the ball carrier, offering a nearby passing option in attack".

Source: Driblab

Weight: 100%

Previously, this was specifically "progressive" passes allowed to go past, leveraging on the Wyscout metric. For now, we are repainting the metric using the Driblab database, looking at all passes.

Source: Driblab

Weight: 100%

Passing accuracy is an outlived concept, a largely irrelevant stat only data dinosaurs tend to refer to these days. But when you look at pass completion under strong pressure (ie. within 0-2 meters of the nearest opponent), it's a different discipline entirely — telling a story of one's awareness.

Source: Manually tracked

Weight: 50%

While for the deep defending section, I pull out "damage control" related back-stopping actions from my notes, here I focus on one's balance in reading and anticipating dangerous plays. That means any crucial interception counts as a plus, while any botched interception attempt along with a mistimed jump or a case of wild miscommunication counts as a minus. The result is again normalized per 90 mins, because being, say, a minus-2 obviously doesn't mean the same across 10 and 30 starts.

Source: Driblab

Weight: 50%

Percentage of successful and total on-ball pressures (within 2-4 meters of the nearest opponent).

Source: Driblab

Weight: 25%

Runs where the player moves around a teammate, taking inner route (closer to center of the pitch).

Source: Wyscout.hudl.com

Weight: 100%

This metric, normalized per 90 mins, measures how many meters a player advances the ball through successful carries, excluding passes. It calculates the difference in distance to the opponent's goal between the start and end point of a run or dribble, counting only forward movements.

Source: Wyscout.hudl.com

Weight: 100%

This metric quantifies how much ground a player advances the ball through their passing, measuring the forward distance (toward the opponent’s goal) between the start and end points of each completed pass. It doesn't account for any lateral/backward movement, as per common sense.

Source: Wyscout.hudl.com

Weight: 75%

Sorting through tags, I filter out accurate long passes that stretch play more or less diagonally, hitting the channels, and take away any out-of-the-bounds stretch passes from the total. This metric is also normalized per 90 mins since being plus-4 means something else across 10 and 30 starts.

Source: Driblab

Weight: 75%

Driblab is much more playful with xG than Wyscout. Here I am leveraging on their metric that takes into account the player's contribution in the build-up where both the shooter and the one who set him up are excluded. Unsurprisingly, midfielders of Slavia and Sparta rank the highest through the sheer amount of xG generated, but there are still intriguing appearances up and down the ladder.

Source: Wyscout.hudl.com

Weight: 100%

This metric, normalized per 90 mins, measures how many meters a player advances the ball through both carries and passes, divided by the number of times he registers a duel or possession loss.

Source: Driblab

Weight: 50%

Switching from Wyscout to Driblab, I'm still interested in passes that penetrate the defensive line.

Source: Driblab

Weight: 50%

Here I compare two separate Driblab values: the traditional pass accuracy against the expected pass completion. You basically want to be anywhere above break even, which would suggest you are doing well with the passes you pick (only 7 of 42 DMs managed that). Here's more on the logic.

Source: Wyscout.hudl.com

Weight: 100%

While with centre backs, this stat mostly boils down to "who's the biggest set piece alfa", with defensive midfielders it is often more about the ability to close down the opponent's penalty box.

Source: Driblab

Weight: 100%

Per Driblab's own glossary, "Expected Threat (xT) is a metric that measures the threat a player creates to the opponent by carrying or passing the ball to areas where the possession's probability of ending in a goal is higher". At the same time, Driblab also tracks how the player allows dangerous areas to be penetrated at the other end, which is exactly where the "net" aspect comes in play.

Source: Wyscout.hudl.com / manually tracked (for danger created via touch/feel for the game)

Weight: 100%

The number of times, normalized per 90 mins, a player enters the opposition’s penalty area in open play — via a carry, a cross, or a pass. It discounts set pieces for fair comparison. In addition, a co-efficient benefitting prolific chance/goal creators via superior touch and feel for the game is applied.

Source: Driblab

Weight: 100%

Switching from Wyscout to Driblab to isolate open play. Mind that xA value is only assigned to a pass/cross that directly precedes a shot; any sort of a partial touch from the opponent along the way, or even a longer dribble/run with the ball from the finisher, normally leads to no xA.

Source: Wyscout.hudl.com / Manually tracked (for danger-creating actions via a creative pass)

Weight: 50%

As I dish out chance/goal-creating actions, I always assign tags to it — with "pass" standing tall, quite understandably, as the most frequent means of generating threat. This is naturally not just any pass, but a creative pass that showcases at least some vision; be it peripheral or finding a hole in defence vertically. The number of these danger-creating actions, normalized per 90 mins, is then enhanced by the number of accurate smart passes ("a creative and penetrative pass that attempts to break the opposition's defensive lines to gain a significant advantage in attack" per Wyscout definition).

Source: Manually tracked

Weight: 100%

My own alternative to Wyscout's flawed xG/A, where I don't care for unfinished business. As long as the player delivers the ball to a dangerous position, onto his teammate's feet, he gets credited regardless of the outcome. Sometimes, if the teammate doesn't time his slide properly or wildly miscontrols the ball, the "on his teammate's feet" caveat doesn't apply, either. And with the exception of goals, the player can also be credited for creating danger for himself. As opposed to xA, I also disregard set piece delivery to make sure the comparison is fair and indeed at all feasible.

Source: Manually tracked

Weight: 50%

Via all goals and only meaningful assists (ie. not just balls handed over to the lucky/sharp shooter at the edge of the box). For 2025/26, I am also crediting any indirect goal contribution (at a lesser weight), while setting back the player for any goal-causing (grave) error which also carries its own value. Here's full explanation of the innovative model, initially introduced by the tremendous CSfotbal. And here's the final leaderboard for 2025/26 (featuring not too many def midfielders, duh).