A match is not played by ratings alone. It is played by people responding to pressure, mistakes, momentum and each other. The Emotional Engine makes those responses part of the simulation.
Imagine two players with identical technical numbers. The match turns hostile. One keeps his shape and waits for the next chance. The other starts chasing the game, forces an ambitious pass, arrives late into a tackle and sees yellow. Until now, most simulations could only call that randomness. Legends Manager can explain why it happened.
The system does not make great players forget how to play. It changes the margins: how boldly they choose, how reliably they execute and how close they move towards a foul or a card. Every reaction begins with the player's own profile, responds to what actually happens on the pitch and gradually settles back towards his personal baseline.
From personality to consequence
The model keeps three things separate: who the player is, what the match does to him and what that state is allowed to change.
The player arrives
A fixed, research-led disposition meets current morale and the emotional residue of recent matches.
The match acts
Goals, errors, cards, penalties and substitutions move activation, tension, frustration and belief.
Margins change
Execution, risk and discipline shift within strict limits. Ratings remain the foundation.
Four states tell the story
Every called-up player carries a live emotional state, whether he is starting, sitting on the bench or has already been withdrawn. Three values can survive from one match to the next. One exists only for the ninety minutes.
From flat and disengaged to fully switched on. Coming off the bench can create an immediate lift.
From unusually calm to overloaded. A dismissal or a decisive late moment can spread it across a team.
From settled to irritable. It lingers longer than nerves and has the strongest carry-over after full time.
Confidence in the next action. It lives only inside the match, then disappears to avoid counting form twice.
The event is shared. The response is personal.
A conceded goal reaches everyone, but not with the same force. The player responsible absorbs more than the teammate watching from the bench. A highly reactive player moves further than a composed one. A substitution made for tactical reasons can sting; one forced by injury does not.
This is the essential difference between scripted drama and a working emotional system. Legends Manager does not assign a mood because a story would look better. It follows a traceable chain from event, to exposure, to personality, to consequence.
Technical reliability moves only slightly. Emotion may disturb a pass; it cannot erase a career's worth of quality.
Belief and activation can make the ambitious option more tempting than the safe one.
Frustration and tension can meaningfully change foul and card propensity, on top of the player's fair-play baseline.
Not a black box. Not a personality verdict.
The first ninety-one profiles were built by hand from biographies, contemporary reporting, teammate and manager accounts, documented incidents and career context. Each value has a written justification. The rest of the roster follows a constrained extraction process that is allowed to return “no signal” instead of inventing one.
The scores never appear on player pages. They are not diagnoses, claims about anyone's inner life or a substitute for real psychology. They are editorial parameters describing football behaviour inside a simulation—and they remain open to review against the evidence that produced them.
One match creates pressure. A World Cup lets it accumulate.
The All-Time World Cup turns the Emotional Engine into a campaign: tension survives, frustration follows a player into the next fixture and every knockout decision carries the weight of what came before.
Discover the All-Time World CupHow the engine is built
The overview ends here. What follows is the architecture, calibration logic, runtime contract and deliberately limited scope of the system.
Two levels, kept strictly separate
The architecture distinguishes disposition from state. Disposition is who a player is inside the model. State is how he feels now. Collapsing the two makes personality drift match after match until every player converges on the same average temperament.
Level 1 holds a static psychological profile: twelve axes derived from biographical and career evidence, never modified by simulation. Level 2 holds a small set of affective variables that respond continuously to match events and decay towards the player's own anchor rather than a neutral zero.
Level 1: the psychological profile
Six trait axes are close in spirit to the established five-factor framework, with discipline added: openness, conscientiousness, extraversion, agreeableness, emotional stability and fair play. Six further axes describe pressure processing: attribution of events, attachment to origins and club, emotional reactivity, rupture threshold and whether a player holds grudges.
Ninety-one profiles at the top of the roster were built by hand from long-form research documents. Each axis value was argued for and recorded with its evidence, creating calibration ground truth. The remaining roster is populated through structured extraction from curated biographies. The pipeline must ground each value in the source, report confidence and return no value where the text offers no signal. Sensitive fields pass through human review.
The profile contributes one runtime number per axis: a reactivity coefficient bounded between 0.75 and 1.25, defaulting to exactly 1.0 when the biography is silent. Reactivity does not move a player's starting state. It changes how far he moves when an event reaches him.
Level 2: persistent context and kickoff anchor
Activation, tension and frustration persist between matches on a scale from −1 to +1. They sit alongside, and never replace, the existing managerial state: team morale, role satisfaction, confidence and trust in the manager, which move slowly on a 20–100 scale. A conceded goal can raise tension without rewriting a player's relationship with his manager.
The context axes are versioned inside the existing morale record, so there is one owner for emotional state rather than a second parallel store. At kickoff, starters and substitutes receive an anchor derived from persistent context and an asymmetric normalization of morale centred on 72. Confidence and general morale combine into a belief baseline. The result is scaled to three quarters of its value and clamped within ±0.65, preventing a player from beginning at an emotional extreme.
Belief is added as a fourth, match-only axis. It is seeded at kickoff, evolves independently and is discarded at full time.
During the match: decay and events
Every called-up player carries an ephemeral record containing the anchor, four current values, last update minute, playing status and a compact history of peaks and troughs. Between events, state relaxes along a closed-form exponential curve rather than a per-tick decrement:
Because the value is computed directly from elapsed time, the same match produces the same emotional path regardless of how the runtime segments its minutes. Half-lives are calibrated by axis: belief roughly 18 minutes, activation 12, tension 10 and frustration 20. Anger outlasts nerves.
Events apply bounded deltas, scaled by reactivity and exposure. Players on the pitch take the full collective effect, substitutes roughly a third and withdrawn players half. Personal events always land at full weight. Scoring raises belief and activation while reducing tension and frustration; a fault on a conceded goal reverses that pattern. Missed and saved penalties are the largest negative and positive events. Cards mainly raise tension and frustration, dismissals affect both the player and the remaining team, and tactical substitutions are distinguished from injuries.
Exactly three runtime effects
| Multiplier | Range | Applied to | Design intent |
|---|---|---|---|
| Execution | 0.94–1.05 | Technical reliability | Emotion nudges quality; it never overrides ratings. |
| Risk | 0.90–1.12 | Ambitious vs safe choice | State changes appetite for the difficult option. |
| Discipline | 0.80–1.30 | Foul and card propensity | Agitation affects behaviour more than raw technique. |
The asymmetry is deliberate. A bad mood is more likely to cost a yellow card than the ability to pass. Discipline layers on top of static fair play rather than replacing it, and passes through concurrent risk and individual weighting gates so one agitated player cannot distort an entire match.
No multiplier feeds back into team momentum. Closing that circuit would create an avalanche—momentum raises frustration, frustration raises fouls, fouls shift momentum again. The engine refuses that loop on purpose.
After the whistle
Belief is discarded because long-term confidence is already represented by ratings and results. Carrying it forward would count the same information twice.
For activation, tension and frustration, the engine takes the difference between final value and kickoff anchor, clamps it within ±0.40 and scales it into persistent context: roughly 30% for activation, 40% for tension and 45% for frustration. Reactivity is not applied again; personality already shaped the events as they landed.
Determinism and observability
The live path is seeded and deterministic. Identical inputs and seed produce the same match and the same emotional trajectories. This is essential for calibration: a match can be replayed and inspected rather than explained away as random variation.
The engine emits a structured emotional summary with peaks, troughs and responsible events. During development, that trace was the only practical way to observe a layer designed to begin with neutral effects and earn its influence through calibration.
What it does not do
It does not model real psychology, publish psychological scores, diagnose players, override ratings, simulate dressing-room politics or modify the managerial layer during a match. Relationship, grudge and rupture parameters anticipate future design work but remain under human review instead of acting automatically.
The work behind it
The base is human: months of reading, cross-checking player research, debating individual axes, discarding earlier designs and calibrating expected behaviour. The half-life order, asymmetric morale normalization, restrained individual effects and open momentum circuit are judgement calls made from a long relationship with football.
AI also did real work. Multiple systems audited the same implementation independently, finding defects in personality lookup, goalkeeper handling, event identifiers, specification units and runtime fallbacks. It helped draft the constrained extraction pipeline that scales beyond the hand-built profiles. Design, editorial judgement and calibration stayed with a person; verification, cross-audit and mechanical implementation were delegated. The engine works because both halves were treated seriously.