Tycoon City: New York

Simulation Systems·2006·3 min read

City simulation systems for Deep Red's Manhattan-scale business tycoon — economy, rival behaviour and the feedback that lets a player read cause and effect.

Overview

Tycoon City: New York, published by Atari in 2006, puts the player in charge of building businesses across Manhattan. Development at Deep Red built on the shared studio engine and the simulation lineage established by Monopoly Tycoon. My work was on the simulation systems: the economic model underneath the city and the behaviour that makes it legible to a player.

Problem

A city tycoon simulation has to be deep enough to reward strategy and transparent enough that a player can tell why something happened. Those pull against each other. A model rich enough to be interesting produces outcomes with many contributing causes, and a player who cannot attribute an outcome to a decision stops believing the simulation is fair — at which point the depth is worthless.

Constraints

  • Shared studio engine and C++98 codebase — extend rather than replace
  • Manhattan-scale map simulated in real time on 2006 consumer hardware
  • Simulation depth constrained by what could be explained to a player through the UI
  • Building on established Deep Red tycoon systems rather than starting fresh

Approach

Treat readability as a simulation requirement rather than a presentation concern. Model the economy so that the factors driving an outcome are few enough to surface directly, and tune rival behaviour to be comprehensible before making it clever — a rival whose actions a player can anticipate creates better competition than one that plays optimally and opaquely.

Key Decisions

Constrain simulation depth to what the interface can explain

Reasoning:

Every additional input to a business's performance makes the model richer and the player's mental model worse. Capping the contributing factors at a number the UI could actually surface kept the economy strategic rather than arbitrary.

Alternatives considered:
  • Maximally detailed economic model with summary reporting
  • Simplified economy with fewer strategic levers

Comprehensible rival behaviour over optimal rival behaviour

Reasoning:

Rivals exist to pressure the player, not to win. Behaviour a player can read and plan against produces a better game than a stronger opponent whose moves look random — and it is dramatically cheaper to tune and debug.

Alternatives considered:
  • Utility-maximising AI with full economic visibility
  • Scripted rival behaviour per scenario

Extend the studio's shared simulation systems

Reasoning:

Deep Red's model reused core technology across titles, so simulation work benefited more than one game. Building on the systems proven in Monopoly Tycoon meant starting from a codebase whose failure modes were already understood.

Alternatives considered:
  • Title-specific simulation written from scratch
  • Licensed middleware

Tech Stack

  • C++98
  • Direct3D
  • Lua
  • Proprietary engine

Result & Impact

  • Atari, 2006
    Published by

Shipped as a Manhattan-scale real-time business simulation, extending the tycoon systems Deep Red had established and carrying them forward into a larger and more detailed city than the studio had previously simulated.

Learnings

  • Simulation depth a player cannot attribute to their own decisions reads as randomness, however sophisticated the model underneath
  • AI that is predictable in the right way makes a better opponent than AI that is strong in the wrong one
  • Shared core technology pays off across titles, but only where the second game's requirements were anticipated in the first
All projects