Statistics, Tactics, New Roles

[Haider.08]

Interpreting Key Passes

A long-standing peeve I've had is how the game has defined key passes and chances created throughout Football Manager editions. If I recall correctly, FM15 had most creative players registering well over 100 key passes per season, whereas in recent editions players are lucky to reach the 30-40 mark in a season. I don't believe that player AI in the match engine is less inclined to create shots - I believe the game has has simply moved its definitions away from the standard understanding.

In the real world, a key pass is defined as any pass/cross that has a shot registered on the end of it, a shot assist, and pre-shot assists precede the shot assist/key pass. When a goal succeeds theses events, the passes are either registered as assists or pre-assists respectively. Chances created in the real world are the sum of assists and key passes.

At the moment, the game attributes key passes arbitrarily. Shot-assists are seldom registered as key passes or chances. When logged, it seems a key pass is attributed to a pass that progresses play, a vertical pass or a ball that opens up the wing for a fullback to cross, and perhaps (but rarely in my experience) an actual shot-assist. Chances seem more consistent, they're what the game defines as a clear goal scoring opportunity, though how that's defined in-game is at the mercy of the coder's subjectivity.

Both these statistical events are very low frequency, compared to their real-life counterparts. I believe the game designers have done this deliberately to help the user, as parsing quality from quantity might be difficult for those unfamiliar to the basics of football data. It's understandable. Not every quality pass has a goal or even a shot succeeding it, so what is the inherent value to the mere volume of these events? On their own and with small sample sizes, these rudimentary metrics aren't absolute representations of creativity, quality or productivity. But by being selective with the data provided to the user, I now have less information to judge player performance on, and to a greater degree, am left in the dark as to whether my offensive tactics are functioning as intended.

I believe the volume of shot-assists from my offensive players is critical data, as it can be quantified as their productivity and their efficiency, relative to how often and where they receive the ball. The quality of a shot assist could instead be measured by a new metric, rather than being selective and obscuring the amount of data a user has to interpret. A solution could be key passes meeting certain criteria for them to be labelled big chances, or a predictive model to display xA/xG... but I'm aware of SI's reticence towards models like xG.

I propose key passes and chances created reflect their real world definitions, but the game continues to ascribe it's high value 'progressive passes' and 'big chances' under new names. One set for the empirical, raw volume that people are used to, and another to parse the subjective quality for those unfamiliar to the metrics. If it's the case that the devs believe the game already operates on this basis, then I guess there's a long-running bug or player AI needs improving to closer replicate real world decision-making.

New Metrics

Passes can be further broken down into subcategories, like 'entry passes'; final third entries, penalty box entries etc. These events are generally an accurate indicator of how incisively a player can move the ball into a new, attacking phase of play. These metrics already have context/value attached to them, and would be of great importance when analysing how players are combining and progressing the ball. 

I also propose a metric that logs how often/successfully a player carries the ball to the next phase of play (something distinct from a direct dribble). A carry can progress play with similar efficacy to a vertical pass that breaks the lines, alongside being more disruptive to opposition shape. Ball retention, playing through pressure, powerful dribbling and then releasing the ball to a teammate for the next phase of play are the components of a carry. Carries don't necessarily coincide with direct dribbles - I believe carrying the ball 30 yards up field, whether directly taking a man or not, has palpable value. This could perhaps be made into a PPM distinct from 'runs with ball often'.

On the defensive side, Statsbomb have recently gone mainstream with their 'pressure/pressure regains' metric. Pressures are simply applying direct pressure in the vicinity of the opposition ball-holder x seconds after a turnover. Pressure regains are logged when the ball is successfully won back following the pressure; I'm not sure if this is given to the ball-winner (or the player who recovers the loose ball opposed to a challenge) or all players applying pressure in the vicinity. The pressure has the inherent value of delaying opposition progression, and testing their composure, whether or not there's a successful regain registered. In terms of your own players, it shows their positional sense, counter-pressing and work-rate without the ball. If your pressures keep getting bypassed or have low frequency, it can also be used to help determine whether you need an additional number in x area of the pitch for numerical parity/covering space.

Over the course of a season, a team that sees less of the ball may inevitably have a higher frequency of pressures, therefore I suggest including an option to adjust the average of a metric for possession, to see if a team has a tactical commitment to applying pressure, or if it's just a byproduct of having less of the ball. The principle can be applied to tackles and interceptions, or even offensive actions to see how efficient a player/team is.

The Tactical Creator

Linking in with the application of pressure, I'd also like to plug this short request by another member, who suggests being able to control the areas pressure is applied throughout different phases of play, perhaps to set up a pressing trap along the touchline etc.

Additionally, a feature that allows users to directly control and differentiate between their off-ball shape in the defensive block versus when pressing would be useful (separate from situational tactics like counter-pressing/regrouping upon losing possession). For example, It might suit a match to press high with a 4141 and stop the opponent from building, but if bypassed retreat to a 442 for more compactness.

These same principles can apply to when you have the ball. It might be necessary for the 6 to split the CBs and the 8/10s to show for the ball to create numerical superiority vs a high press, but once play has progressed, the 6 might need to situate himself higher up the pitch to compress space and be ready to circulate or recover loose balls. I'd like more control over each phase of play when it comes to the 'numbers game' - achieving numerical superiority in different areas of the pitch, offensive or defensive. 

My vision for the tactical creator is distinct from the current one, with positional duties for phases of play separated from how a player with x role operates with the ball or when attacking; more holistic control over tactics. Changes to shape can currently be half-acheived by using a mixture of roles and duties, but roles with the desired positional tendencies also have characteristics with the ball that could be opposed to the style you're trying to achieve.

I feel that these instructions should override the role/duty you select. I want a player I select with an attack duty to have a high risk-taking mentality, but to also partake in the tactical, holistic side of the game, rather than just stay high/look for space and neglect defensive duty (unless instructed to). Teams in real life have risk-takers but they still abide by and work in the shape and tactics managers set. I feel that the mark x player/position or close down more PIs are insufficient. The need for players to stay high and provide the out-ball or pin the opposition back is still important, and I'd suggest making that a PI separate from their role/duty.

And finally: I've always imagined a wide forward role. This role would have the freedom to dribble inside/outside, would contribute without the ball, have positional freedom relative to the hard-coded wingers and inside forwards. Similar to the wide midfielder role but in the AML/AMR spot. 

I don't have the mental energy to go on. I appreciate any readers, and SI for their product. Thank you.

TL;DR

Key passes to work how they work irl, new metrics for pre-shot assists and pre-assists.

New metrics for progressive plays, like final third entries and when the ball is carried to the next phase of play. Inclusion of pressure/pressure regain metrics.

More control over team shape/numbers when in a defensive block, pressing, in transition, building play, attacking - regardless of role and duty. Restrict roles to on-ball characteristics.

A Wide Forward role for AMR AML, similar to the Wide Midfielder.

 

Edited May 17, 2019 by Haider.08

[kingbart]

fantastic post, seconded.

[autohoratio]

Great suggestions, I wholeheartedly agree.

[Kyle Brown]

Reviewed so locked.