My Expectations as a Fan of Boxing for an AI Programmer Building a Truly Realistic Boxing Videogame (Unreal Engine)

A realistic boxing videogame demands an AI system that simulates boxer identity, not just opponent difficulty. The expectation is an AI architecture capable of producing thousands of distinct boxing behaviors through layered sliders, traits, tendencies, mannerisms, and contextual decision-making—without hardcoding outcomes or relying on artificial boosts.

This is not about spectacle. It is about authenticity, variability, and systemic depth.

1) Foundational Principle: Boxing AI Is a Behavioral Ecosystem

The AI must be designed as a multi-layered behavioral ecosystem, where:

Capabilities define what is physically and mentally possible
Tendencies define what is preferred
Traits define what overrides normal behavior
Mannerisms define how behavior is expressed
Psychology defines why behavior changes
Context defines when behavior shifts

Every action taken by the AI should be explainable through these layers.

2) Capability Sliders (What the Boxer Is Capable Of)

Capabilities are not “ratings.” They are constraints and consistency modifiers.

A) Offensive Capabilities

Jab speed
Jab accuracy
Jab recovery
Jab authority (ability to disrupt rhythm)
Straight punch mechanics
Lead hook mechanics
Rear hook mechanics
Uppercut timing precision
Punch chaining fluidity
Punch commitment control (ability to bail mid-action)
Punch retraction speed
Punching balance retention
Power transfer efficiency
Hand speed under fatigue
Punch accuracy decay rate

B) Defensive Capabilities

Static guard integrity
Dynamic guard adjustment speed
Parry window size
Slip window precision
Roll execution reliability
Pull-counter balance retention
Recovery defense (defense while hurt)
Counter-defense transition speed
Guard recovery after impact
Defensive mistake rate under pressure

C) Footwork & Movement Capabilities

Forward pressure balance
Backward movement control
Lateral movement efficiency
Pivot sharpness
Angle exit reliability
Stance stability under fire
Cut-off geometry awareness
Rope awareness
Corner escape capability
Momentum control (stop/start movement)
Footwork degradation under fatigue

D) Athletic & Physical Capabilities

Aerobic stamina
Anaerobic burst capacity
Fatigue recovery rate
Balance under contact
Chin durability
Body durability
Leg durability
Torque generation
Injury resistance
Injury compensation ability

E) Cognitive & Ring IQ Capabilities

Read accuracy
Read speed
Pattern recognition
Deception recognition
Counter timing precision
Gameplan retention
Adjustment speed
Feint interpretation skill
Risk calculation accuracy

Capabilities answer:
“Can this boxer do this, and how well?”

3) Tendency Sliders (What the Boxer Prefers to Do)

This layer must be extensive, granular, and directly wired into decision weighting.

A) Engagement Tendencies

Pressure frequency
Reset frequency
Exchange willingness
Initiation bias
Clinch seeking frequency
Clinch avoidance
Late-round urgency
Early-round caution
Come-forward persistence
Retreat tolerance

B) Shot Selection Tendencies

Jab-first preference
Double-jab preference
Jab-to-body preference
Jab-to-head preference
Lead hook usage
Rear straight usage
Rear hook usage
Uppercut frequency
Body shot prioritization
Head-hunting bias
Single-shot preference
Combination preference
Combination length bias

C) Defensive Habit Tendencies

High guard reliance
Shell usage
Long guard usage
Slip-left vs slip-right bias
Roll preference
Pull-counter preference
Parry-first behavior
Catch-and-shoot preference
Defense-to-offense immediacy

D) Footwork Habit Tendencies

Circle-left bias
Circle-right bias
Pivot frequency
Step-out vs step-back preference
Angle-after-punch behavior
Rope escape preference
Rope trap avoidance
Ring center priority
Cut-off commitment

E) Psychological Tendencies

Patience vs impatience
Risk tolerance
Revenge behavior after being hit
Emotional volatility
Confidence gain rate
Confidence loss rate
Showboating likelihood
Discipline under pressure

F) Finish Tendencies

Swarm instinct
Sniper instinct
Trap-setting instinct
Body-first finishing
Head-first finishing
Finish patience
Overcommit risk when opponent is hurt

Tendencies answer:
“Given multiple valid options, which does this boxer lean toward?”

4) Trait System (Rule-Based Overrides)

Traits must act as conditional modifiers or logic overrides, not flavor text.

Examples

Dangerous when hurt
Slow starter / fast starter
Momentum fighter
Body investment specialist
Clinch disruptor
Late-round closer
Veteran round thief
Glass hands
Iron chin
Crowd-responsive
Emotionally fragile
Ice-cold under pressure

Traits can:

Temporarily override tendencies
Expand or shrink decision windows
Alter risk calculations
Trigger unique behavioral states

5) Mannerism Sliders (How the Boxer Expresses Behavior)

Mannerisms provide human texture.

Mannerisms must scale with:

Fatigue
Damage
Confidence
Fight context

6) Psychology & Internal State Modeling

AI must maintain internal states such as:

Confidence
Composure
Frustration
Momentum perception
Perceived opponent danger
Urgency awareness (round, score)

These states directly modify:

Reaction speed
Risk tolerance
Shot commitment
Defensive caution

7) Ringcraft & Spatial Intelligence

The AI must treat the ring as a tactical environment:

Center control goals
Exit lane evaluation
Rope danger scoring
Corner risk evaluation
Trap construction logic
Escape prioritization

Movement decisions must be intentional, not reactive chasing.

8) Opportunity-Based Decision Making

Punches are chosen based on:

Guard gaps
Weight transfer moments
Rhythm breaks
Post-punch vulnerability
Counter exposure risk

AI must be capable of:

Feint-aborts
Half-commits
Combo truncation
Opportunistic counters

9) Fatigue, Damage, and Injury Integration

Fatigue and damage must:

Reduce available actions
Change preferred tactics
Increase mistake probability
Alter mannerisms and posture

A tired boxer must look, move, and think tired.

10) Unreal Engine Implementation Expectations

The AI programmer must demonstrate mastery of:

DataAssets / DataTables for all sliders
Behavior Trees or StateTree for high-level logic
Animation Montages and Motion Warping
Clean AI ↔ animation ↔ hit reaction loops
Network-safe execution (if applicable)

Designers must be able to tune behavior without touching code.

11) Debugging & Tooling (Non-Negotiable)

Required tools include:

Live decision overlays
Slider visualization
Ring heatmaps
Behavior logs
Adaptation tracking
Replay consistency controls

12) Adaptation & Anti-Exploit Logic

AI must detect and counter:

Repeated attack patterns
Excessive retreating
Guard-only defense
Predictable combos

Adaptation should be gradual and believable, not instant omniscience.

13) Expected Deliverables from the AI Programmer

A qualified AI programmer should deliver:

A deep, slider-driven identity system
Extensive tendency libraries
Trait-based behavior overrides
Mannerism and psychology layers
Ringcraft intelligence
Opportunity-based combat logic
Robust debug and tuning tools

Final Expectation

A realistic boxing videogame does not need smarter AI.
It needs deeper AI—AI that reflects the complexity of boxing itself.

The goal is not to beat the player unfairly.
The goal is to create fights that feel earned, varied, human, and endlessly replayable.

If Poe were hiring an AI Programmer with Unreal Engine Experience

Senior AI Programmer – Boxing Simulation (Unreal Engine)

Project Type: Realistic / Simulation-Driven Boxing Videogame
Focus: Offline + Online Parity, Systemic Depth, Long-Term Expandability

Role Overview

This role exists to build the core intelligence of the boxing experience.

The Senior AI Programmer will design and implement a deep, identity-driven boxing AI system that produces believable, stylistically distinct boxers through extensive sliders, traits, tendencies, mannerisms, psychology, and contextual decision-making—not scripted behaviors or artificial difficulty boosts.

This is not a “combat AI” role in the traditional sense.
This is a boxing intelligence role.

Vision & Non-Negotiables

The AI must:

Represent boxing as it is actually practiced
Produce thousands of distinct boxer identities
Be designer-driven, not programmer-locked
Scale across eras, styles, and difficulty levels
Remain transparent, debuggable, and tunable
Never rely on hidden cheats or input reading

Difficulty must emerge from:

Better reads
Smarter decisions
Fewer mistakes
Greater discipline
Improved adaptation

—not from inflated stats or unfair reactions.

Core AI Philosophy

Boxing AI is treated as a behavioral ecosystem, not a single brain.

The system must clearly separate:

Capabilities – what a boxer can do
Tendencies – what a boxer prefers to do
Traits – what overrides normal behavior
Mannerisms – how behavior is expressed
Psychology – why behavior changes
Context – when behavior shifts

Every AI decision should be explainable through these layers.

Primary Responsibilities

1. AI Architecture & Systems Design

Design a layered AI architecture including:
- Perception
- Intent & round planning
- Tactical evaluation
- Opportunity & risk scoring
- Action selection
- Execution & animation handoff
- Adaptation & learning
Ensure the system supports emergent behavior, not scripted outcomes.

2. Capability Slider Framework (Extensive)

Build a large-scale capability system covering:

Offensive Capabilities

Jab speed, accuracy, recovery, authority
Punch mechanics by type (straight, hook, uppercut)
Combination fluidity
Balance retention
Punch retraction speed
Power transfer efficiency
Accuracy decay under fatigue

Defensive Capabilities

Guard integrity (static & dynamic)
Parry window precision
Slip and roll reliability
Recovery defense
Counter-defense transition speed
Defensive degradation under pressure

Footwork & Movement

Forward/backward control
Lateral efficiency
Pivot sharpness
Angle exit reliability
Cut-off geometry awareness
Rope and corner awareness
Movement degradation under fatigue

Athletic & Physical

Aerobic stamina
Anaerobic burst
Fatigue recovery
Balance under contact
Chin, body, and leg durability
Injury resistance and compensation

Cognitive / Ring IQ

Read accuracy and speed
Pattern recognition
Deception recognition
Counter timing
Adjustment speed
Risk calculation accuracy

Capabilities define possibility and consistency, not personality.

3. Tendency Slider System (Very Deep)

Implement a plethora of tendency sliders that meaningfully alter behavior:

Engagement

Pressure frequency
Reset frequency
Exchange willingness
Clinch seeking / avoidance
Early-round caution
Late-round urgency

Shot Selection

Jab-first bias
Double-jab usage
Body vs head focus
Lead vs rear preference
Uppercut frequency
Combo length preference
Single-shot vs volume bias

Defensive Habits

High guard reliance
Shell usage
Long guard usage
Slip direction bias
Roll preference
Pull-counter usage
Catch-and-shoot tendency

Footwork Habits

Circle direction bias
Pivot frequency
Step-out vs step-back
Rope escape behavior
Ring center priority
Cut-off commitment

Psychological Tendencies

Patience vs impatience
Risk tolerance
Revenge behavior after being hit
Emotional volatility
Confidence gain/loss rate
Discipline under pressure

Finishing Instincts

Swarm vs snipe vs trap
Body-first finishing
Head-first finishing
Overcommit risk when opponent is hurt

Tendencies must directly affect:

Action weighting
Distance selection
Timing windows
Risk evaluation
Round strategy

4. Trait System (Behavioral Overrides)

Design a trait system that can temporarily override normal logic, such as:

Dangerous when hurt
Slow starter / fast starter
Momentum-based fighter
Body investment specialist
Clinch disruptor
Veteran round manager
Emotionally fragile or ice-cold
Crowd-responsive

Traits must create recognizable moments, not passive bonuses.

5. Mannerisms & Expression

Implement systems that add human texture, including:

Rhythm and cadence
Guard posture
Feint language
Breathing behavior
Reset animations
Corner behavior
Ref interactions

Mannerisms must respond dynamically to:

Fatigue
Damage
Confidence
Fight context

6. Psychology & Internal State Modeling

Maintain internal AI states such as:

Confidence
Composure
Frustration
Momentum perception
Urgency (round, score, fight state)

These states must influence:

Reaction speed
Risk tolerance
Shot commitment
Defensive caution
Adaptation speed

7. Ringcraft & Spatial Intelligence

Implement true ring awareness:

Center control goals
Rope danger scoring
Corner risk evaluation
Exit lane analysis
Cut-off geometry
Trap construction and escape logic

Movement must be intentional, not chase-based.

8. Opportunity-Based Combat Logic

All offense must be driven by:

Guard gaps
Rhythm breaks
Weight transfer moments
Post-punch vulnerability
Counter exposure risk

AI must support:

Feint-aborts
Half-commits
Combo truncation
Opportunistic counters

9. Unreal Engine Integration

Required expertise includes:

DataAssets / DataTables for all sliders
Behavior Trees or StateTree
Animation Montages & Motion Warping
Clean AI → animation → hit → response pipelines
Network-safe logic (if applicable)

Designers must be able to tune everything without touching code.

10. Tooling & Debugging (Mandatory)

You will build:

Live AI decision overlays
Slider visualizers
Behavior logs (“why this happened”)
Ring heatmaps
Adaptation tracking tools
Replay variability controls

Explainability is not optional.

Required Qualifications

Strong Unreal Engine AI experience
Proven work on systemic, non-scripted AI
Deep understanding of boxing fundamentals
Experience building designer-facing tools
Strong debugging and profiling skills
Systems-thinking mindset

Preferred Qualifications

Sports, fighting, or simulation game experience
Large-scale slider systems
Gameplay Ability System familiarity
Offline + online system awareness
Passion for realism and long-term depth

Evaluation Criteria

Candidates will be evaluated on their ability to:

Create believable boxer identities
Explain AI decisions clearly
Avoid shortcuts and cheats
Support designers
Build scalable systems
Respect boxing as a craft, not an arcade abstraction

Example Technical Evaluation

Candidates may be asked to:

Design a boxer identity system using sliders, traits, and tendencies
Demonstrate opportunity-based punch selection
Show how adaptation occurs across rounds
Provide debug output explaining decisions
Discuss how realism is preserved at higher difficulty

Final Statement

This role exists to ensure players are not fighting AI—
they are fighting boxers.

The measure of success is not win rate.
It is whether players believe the opponent thinks, reacts, adapts, and behaves like a real human boxer.

Wednesday, January 28, 2026