The CPU chooses to cross with him far more often (+30 points), and while his team has the ball he is pulled in towards the middle much less.
Who does this affect?
Only players the CPU is moving: your AI teammates and the opponent's players. The style changes where the CPU puts him and what the CPU decides for him; while you steer him yourself, your input decides. Found on: Compatible: LWF/RWF, LMF/RMF. In the mobile card data: 329 cards (side midfielders 242, wingers 85, RB 2 - no positioning effect at RB).
How it works
When the CPU has the ball at his feet and a cross is possible, its chance of choosing the cross is 10 + 50 x (Lofted Pass from 60 to 90) points, and a Cross Specialist adds 30 (this part works in any slot). The CPU's cross option scores the receivers by height, jump and strength and aims at the goal area. While his team has the ball, his tuck-in factor is capped at 0.1, so he is pulled at most 5 % of his distance towards the middle (other wide players up to 25 %), and the late squeeze towards the middle near the box starts for him only when play is 2 m short of the penalty-box line (others: when the ball is 8 m short). In one CPU support calculation his spot is on the touchline, like a Prolific Winger. Limit: these CPU decisions give players numbered roles that have no names in the code. Turning each role into a named move (for example a run in behind) needs the whole CPU role system mapped, which this audit could not complete, and live measurement was not possible (no device connection). They are therefore described only by what the code checks.
When it applies
CPU cross choice (ball at his feet, CPU-controlled): chance = 10 + 50 x (Lofted Pass 60 to 90) + 30 for Cross Specialist; the +30 is given once even if he also has the AI playing style flag that gives the same +30 (AI-mask bit 3, Early Crosser); +40 more with the Pinpoint Crossing skill. Then x0.6 or x0.8 in some situations, x1.2 in one mode, capped at 100; compared with a random number 0-99.
In possession: tuck-in factor at most 0.1 (pull <= 5 % of his distance from the middle).
Late side squeeze (60 % x factor): for others the factor rises while the ball goes from 8 m short of the box line to 3 m inside it; for him while the shape's reference point goes from 2 m short to 3 m inside.
One CPU support calculation puts his spot on the touchline (34.06 m from the middle).
Positioning parts only in a winger or side-midfield slot; the cross chance in any slot.
Without it
A wide player without the style has 30 points less cross chance, a tuck-in factor up to 0.5, and starts the late squeeze earlier.
Which stats it reads
Lofted Passsets how strong
Adds 0 to 50 points to the CPU's cross chance: nothing at 60 or below, the full 50 at 90 and above.
In numbers
CPU cross chancePoints out of 100 before situation factors (x0.6/x0.8/x1.2, cap 100). Without the style vs Cross Specialist; no Pinpoint Crossing. CPU-controlled only.
Lofted Pass → · points
Chart values
60
10
40
90
60
90
Common questions
Does Cross Specialist improve his crosses?
Short answerNo
Cross kick and accuracy code does not read the style. It changes how often the CPU chooses to cross and where he stands.
Does it stack with Early Crosser?
Short answerNo
The code adds +30 for the style, or else +30 for the Early Crosser flag, never both.
Does the style make me cross better when I control him?
Short answerNo
The cross chance is a CPU decision; your crosses are your button presses.
Does the style still work if I play him out of position?
Short answerPartly
No for the positioning parts: they check his formation slot every time and do nothing outside winger or side-midfield slots. The CPU cross chance ignores the slot.
The maths
ConditionPositioning parts: only in LWF/RWF or LMF/RMF slots. The cross chance ignores the slot.
BendLofted Pass counts from 60 to 90.
Step+30 cross chance (not stacked with the Early Crosser +30).
LimitTuck-in factor at most 0.1.
DetailLimit: these CPU decisions give players numbered roles that have no names in the code. Turning each role into a named move (for example a run in behind) needs the whole CPU role system mapped, which this audit could not complete, and live measurement was not possible (no device connection). They are therefore described only by what the code checks.
DetailNot used by this style: Curl and Low Pass are not read by the cross-choice code; Crossing accuracy itself is not changed by the style.