How can a game system be modeled through state, inputs, outputs, and connections to other systems?

A game system can be modeled as a stateful transformation whose inputs change internal variables and produce outputs consumed by other systems. - State describes what persists between updates, such as a city holding 80 food, 40 population, and level 2 walls. - Inputs are events or resources that may change that state, such as spending 20 wood to start a wall upgrade. - Outputs are observable results or signals, such as higher defense, a construction slot becoming occupied, and wood leaving the economy. - Connections appear when one system's output becomes another system's input, such as population increasing food demand while food shortage reduces population growth.

What is the difference between first-order and second-order effects in an interconnected game system?

A first-order effect follows directly from a rule change, while a second-order effect follows through one or more affected systems. - Increasing sword damage from 10 to 12 directly shortens the number of hits needed against a 60-health enemy, which is a first-order effect. - Faster kills reduce the damage players receive per encounter, so healing items lose value, which is a second-order effect. - Lower healing demand may let players spend more currency on upgrades, indirectly accelerating later progression. - The distinction is about causal distance rather than importance, because an indirect economy shift can be larger than the original combat change.

What makes a gameplay outcome emergent rather than explicitly authored?

An outcome is emergent when it arises from interacting general rules without being separately scripted as a specific result. - Fire spreading to adjacent dry tiles and rain making tiles wet can produce an unscripted firebreak when a storm crosses a burning field. - The outcome remains rule-consistent because each local interaction uses the same fire, dryness, and wetness rules. - Emergence requires combinable state and interactions, not mere randomness, because a random reward table does not create a new systemic relationship. - Designers still shape the possibility space by choosing rule scope, such as allowing wind to affect both projectile drift and fire spread.

How do positive and negative feedback loops differ, and what effects do they tend to create?

Positive feedback amplifies an existing advantage or disadvantage, while negative feedback counteracts deviation and pushes the system toward stability. - A territory that grants income used to capture more territory forms positive feedback because control produces the means for additional control. - A trailing racer receiving a small speed bonus forms negative feedback because distance behind increases a force that reduces that distance. - Positive feedback can create escalation and decisive endings, but it can also make early leads difficult to challenge. - Negative feedback can preserve contestability, but excessive correction can make prior choices feel irrelevant, such as fully equalizing all racers every lap.

How is systemic depth different from mechanical complexity?

Systemic depth is the range of meaningful decisions produced by rules, while complexity is the amount of information and procedure required to use those rules. - A movement rule with only push and pull can be deep when terrain, hazards, and unit positions make each use materially different. - Ten nearly identical damage types add complexity if the correct response is always to equip the matching resistance. - Depth grows when choices have contextual tradeoffs, such as a shield blocking damage but consuming the stamina needed to escape. - Complexity can support depth when distinctions change decisions, but extra stats, exceptions, and steps are not depth by themselves.

What does it mean for two mechanics to be orthogonal, and why can orthogonality increase depth?

Two mechanics are orthogonal when they affect meaningfully different decision dimensions instead of functioning as stronger and weaker versions of the same choice. - A dash changes position while a shield changes damage exposure, so each solves a different part of combat state. - Two attacks that differ only as 10 damage versus 12 damage have low orthogonality because one usually replaces the other. - Orthogonal tools can combine into new tactics, such as dashing behind cover before shielding an ally during their retreat. - The mechanics still need shared points of interaction, because completely unrelated subsystems create parallel complexity rather than combinatorial depth.

How should a resource conversion loop be represented and analyzed as a game system?

A resource conversion loop should be represented as sources, transformations, storage, sinks, and the rates connecting them. - A mine producing 5 ore per minute is a source, while a forge converting 2 ore into 1 tool is a transformation. - A warehouse with capacity 20 is storage, so it changes timing by allowing ore production to continue before tools are needed. - Tool durability that removes one tool after ten uses is a sink that prevents the stock from growing without bound. - If one tool upgrades the mine to add 3 ore per minute, the loop reinvests output into throughput and may compound unless time, capacity, or costs constrain it.

What is a degenerate loop or exploit from a systemic design perspective?

A degenerate loop is a repeatable rules-valid interaction that dominates alternatives by producing disproportionate value, often with little cost or counterpressure. - If buying one seed costs 1 coin and selling its guaranteed harvest yields 3 coins immediately, repeated farming creates unbounded currency. - The problem lies in the closed conversion path and its net gain, not in whether the player intended to bypass the expected progression. - A strong strategy is not automatically degenerate, because limited land, elapsed seasons, or competing uses for seeds can preserve meaningful tradeoffs. - An exploit may use an unintended rule interaction, but its systemic significance comes from repeatability, scalability, and how strongly it invalidates other choices.

What information should a dependency map capture for a set of interconnected game systems?

A dependency map should show which systems read, modify, enable, or constrain the states of other systems. - Nodes can represent systems or key state variables, such as stamina, movement, attacks, equipment weight, and regeneration. - Directed edges should name the relationship, such as equipment weight increasing stamina cost rather than merely linking equipment and stamina. - Cycles should be visible, such as combat rewards improving equipment while stronger equipment increases combat reward acquisition. - Critical assumptions can be marked on edges, such as movement requiring stamina to stay above zero, which exposes where one rule change propagates widely.

What is a design invariant in a game system, and how does it guide rule changes?

A design invariant is a property that must remain true across valid states and rule changes because other parts of the design depend on it. - A turn economy may require that every normal unit receives exactly one base action per round before bonuses are applied. - A trading system may require that every resource has at least one sink, preventing permanent accumulation with no decision attached. - An invariant differs from a tuning target, because damage averaging near 20 may change while health never dropping below zero remains structurally required. - Stating invariants narrows safe design space, such as allowing new movement abilities only if every reachable combat tile still has at least one counter-entry route.

How do sustained DPS and burst damage differ, and how would you calculate them for a weapon that deals 120 damage every 0.75 seconds plus an instant 360-damage ability with a 6-second cooldown?

Sustained DPS measures long-run damage rate, while burst measures damage delivered inside a specified short window. - The weapon's sustained DPS is 120 / 0.75 = 160 damage per second. - The ability contributes 360 / 6 = 60 average DPS when used on cooldown, so combined long-run DPS is 220. - Firing the weapon and ability together deals an opening burst of 480 damage, but its burst DPS cannot be stated without defining the time window.

How is time to kill calculated when damage arrives in discrete hits, and what is the TTK against 1,000 health for a weapon dealing 180 damage at 2 shots per second if the first shot lands at time zero?

Discrete-hit TTK depends on the number of lethal hits and the intervals between them. - The required hit count is ceil(1,000 / 180) = 6 because five hits deal only 900 damage. - At 2 shots per second, the interval between shots is 1 / 2 = 0.5 seconds. - Six shots contain five intervals after the time-zero shot, so TTK is (6 - 1) × 0.5 = 2.5 seconds.

What is effective health, and how do mitigation and shield ordering affect it for a character with 1,200 health, 25% damage reduction, and a 300-point shield?

Effective health is the amount of incoming raw damage required to exhaust all protected durability. - With mitigation fraction r, effective health is H / (1 - r), so 1,200 / 0.75 = 1,600 raw damage. - If the shield also receives the 25% reduction, total effective health is (1,200 + 300) / 0.75 = 2,000. - If the shield is consumed before mitigation and takes raw damage, total effective health is 300 + 1,600 = 1,900. - The layer order must therefore be part of the formula rather than assumed from the displayed health total.

What is a damage breakpoint, and where is the three-hit breakpoint against 500 health if damage is an integer and attacks occur every 0.8 seconds?

A breakpoint is a threshold where a small numerical change alters a discrete outcome such as hits required to kill. - Damage of 126 and 130 both require ceil(500 / damage) = 4 hits, so the increase does not change hit count. - The three-hit breakpoint is ceil(500 / 3) = 167 damage because 166 × 3 = 498 and 167 × 3 = 501. - With the first hit at time zero, crossing the breakpoint changes TTK from 3 × 0.8 = 2.4 seconds to 2 × 0.8 = 1.6 seconds.

How do expected value and variance describe a reward that gives 10 coins with 70% probability, 30 coins with 25% probability, and 100 coins with 5% probability?

Expected value describes the long-run average, while variance describes how widely individual outcomes spread around it. - The expected value is 0.70 × 10 + 0.25 × 30 + 0.05 × 100 = 19.5 coins. - The second moment is 0.70 × 10^2 + 0.25 × 30^2 + 0.05 × 100^2 = 795. - The variance is 795 - 19.5^2 = 414.75 coins squared, and the standard deviation is about 20.36 coins. - Two reward tables can share an expected value but feel different because the one with higher variance produces more extreme outcomes.

What is the difference between independent and conditional probability for two attacks with a 20% base critical-hit chance?

Independence means one attack's result does not change the probability distribution of the other attack. - For independent attacks, the probability that both crit is 0.20 × 0.20 = 0.04, or 4%. - The probability of at least one crit is 1 - 0.80^2 = 0.36, or 36%. - In general P(A and B) = P(A) × P(B | A), so if a first crit raises the second attack's crit chance to 50%, the probability of two crits becomes 0.20 × 0.50 = 10%. - Independence is the special case where P(B | A) = P(B), not the case where events cannot happen together.

How does a hard pity rule change loot probability if each attempt has an independent 2% drop chance and attempt 50 guarantees the item after 49 failures?

A hard pity converts the geometric distribution into a capped distribution with guaranteed success by a fixed attempt. - The chance of receiving the item during attempts 1 through 49 is 1 - 0.98^49 ≈ 62.84%. - The chance of reaching the guaranteed attempt is 0.98^49 ≈ 37.16%. - The expected attempts per item are sum from k = 0 to 49 of 0.98^k = (1 - 0.98^50) / 0.02 ≈ 31.79 attempts. - The guaranteed fiftieth attempt makes the probability of obtaining the item by attempt 50 equal to 100%, not merely 50 × 2%.

How do linear, polynomial, exponential, and S-curve formulas produce different progression shapes, using n completed levels and a starting value of 100?

The formula family determines how absolute and relative growth change across progression. - Linear growth such as V(n) = 100 + 20n adds a constant 20 per level and gives V(5) = 200. - Quadratic growth such as V(n) = 100 + 5n^2 has an increasing absolute increment and gives V(5) = 225. - Exponential growth such as V(n) = 100 × 1.10^n adds a constant 10% proportion and gives V(5) ≈ 161.05. - A logistic S-curve such as V(n) = 1,000 / (1 + 9e^(-0.5n)) starts at 100, grows fastest near n = 4.39 where V = 500, and approaches 1,000 without crossing it.

How do diminishing returns differ from a soft cap, and how can each be expressed with a simple formula?

Diminishing returns reduce the marginal gain smoothly, while a soft cap changes the conversion rate after a threshold without stopping growth. - The diminishing-return function B(x) = 100x / (x + 100) gives B(100) = 50 and B(200) ≈ 66.7, so the second 100 input adds only about 16.7. - A soft-cap conversion E(x) = 0.4 × min(x, 50) + 0.1 × max(x - 50, 0) grants 0.4 percentage points per input before 50 and 0.1 afterward. - Under that soft cap, x = 80 grants 0.4 × 50 + 0.1 × 30 = 23 percentage points, so investment still helps but at a lower rate.

How can normalization support asymmetric faction balance if a 100-cost unit from faction A has 120 health and 15 DPS while a 100-cost unit from faction B has 90 health and 20 DPS?

Normalization places unlike stat profiles on common scales without requiring the profiles themselves to become identical. - Per 100 cost, faction A provides 120 health and 15 DPS, while faction B provides 90 health and 20 DPS, preserving a durability versus damage tradeoff. - Relative to means of 105 health and 17.5 DPS, A's normalized vector is approximately (1.143, 0.857) and B's is (0.857, 1.143). - With continuous damage and no other effects, A needs 90 / 15 = 6 seconds to defeat B and B needs 120 / 20 = 6 seconds to defeat A. - A single weighted average can hide meaningful strengths, so normalization should compare explicit dimensions and equivalent costs rather than erase faction asymmetry.

Game Designer interview questions

100 real questions with model answers and explanations for Middle candidates.

See a Game Designer resume example →

Practice with flashcards

Spaced repetition · Hunter Pass

Questions

system-design

A game system can be modeled as a stateful transformation whose inputs change internal variables and produce outputs consumed by other systems.

State describes what persists between updates, such as a city holding 80 food, 40 population, and level 2 walls.
Inputs are events or resources that may change that state, such as spending 20 wood to start a wall upgrade.
Outputs are observable results or signals, such as higher defense, a construction slot becoming occupied, and wood leaving the economy.
Connections appear when one system's output becomes another system's input, such as population increasing food demand while food shortage reduces population growth.

Why interviewers ask this: The interviewer evaluates whether the candidate can describe systems as explicit transformations rather than isolated feature lists.

system-design

A first-order effect follows directly from a rule change, while a second-order effect follows through one or more affected systems.

Increasing sword damage from 10 to 12 directly shortens the number of hits needed against a 60-health enemy, which is a first-order effect.
Faster kills reduce the damage players receive per encounter, so healing items lose value, which is a second-order effect.
Lower healing demand may let players spend more currency on upgrades, indirectly accelerating later progression.
The distinction is about causal distance rather than importance, because an indirect economy shift can be larger than the original combat change.

Why interviewers ask this: The interviewer evaluates whether the candidate traces consequences across system boundaries and separates direct from propagated effects.

gameplay

An outcome is emergent when it arises from interacting general rules without being separately scripted as a specific result.

Fire spreading to adjacent dry tiles and rain making tiles wet can produce an unscripted firebreak when a storm crosses a burning field.
The outcome remains rule-consistent because each local interaction uses the same fire, dryness, and wetness rules.
Emergence requires combinable state and interactions, not mere randomness, because a random reward table does not create a new systemic relationship.
Designers still shape the possibility space by choosing rule scope, such as allowing wind to affect both projectile drift and fire spread.

Why interviewers ask this: The interviewer evaluates whether the candidate distinguishes rule interaction from authored sequences and arbitrary randomness.

feedback-loopsfeedback

Positive feedback amplifies an existing advantage or disadvantage, while negative feedback counteracts deviation and pushes the system toward stability.

A territory that grants income used to capture more territory forms positive feedback because control produces the means for additional control.
A trailing racer receiving a small speed bonus forms negative feedback because distance behind increases a force that reduces that distance.
Positive feedback can create escalation and decisive endings, but it can also make early leads difficult to challenge.
Negative feedback can preserve contestability, but excessive correction can make prior choices feel irrelevant, such as fully equalizing all racers every lap.

Why interviewers ask this: The interviewer evaluates whether the candidate identifies loop direction, systemic behavior, and the tradeoffs of amplification versus stabilization.

system-designmechanicsalgorithms

Systemic depth is the range of meaningful decisions produced by rules, while complexity is the amount of information and procedure required to use those rules.

A movement rule with only push and pull can be deep when terrain, hazards, and unit positions make each use materially different.
Ten nearly identical damage types add complexity if the correct response is always to equip the matching resistance.
Depth grows when choices have contextual tradeoffs, such as a shield blocking damage but consuming the stamina needed to escape.
Complexity can support depth when distinctions change decisions, but extra stats, exceptions, and steps are not depth by themselves.

Why interviewers ask this: The interviewer evaluates whether the candidate judges mechanics by the decisions they generate rather than by rule count.

mechanics

Two mechanics are orthogonal when they affect meaningfully different decision dimensions instead of functioning as stronger and weaker versions of the same choice.

A dash changes position while a shield changes damage exposure, so each solves a different part of combat state.
Two attacks that differ only as 10 damage versus 12 damage have low orthogonality because one usually replaces the other.
Orthogonal tools can combine into new tactics, such as dashing behind cover before shielding an ally during their retreat.
The mechanics still need shared points of interaction, because completely unrelated subsystems create parallel complexity rather than combinatorial depth.

Why interviewers ask this: The interviewer evaluates whether the candidate can distinguish independent decision axes from redundant content and explain useful interaction between them.

conversionsystem-design

A resource conversion loop should be represented as sources, transformations, storage, sinks, and the rates connecting them.

A mine producing 5 ore per minute is a source, while a forge converting 2 ore into 1 tool is a transformation.
A warehouse with capacity 20 is storage, so it changes timing by allowing ore production to continue before tools are needed.
Tool durability that removes one tool after ten uses is a sink that prevents the stock from growing without bound.
If one tool upgrades the mine to add 3 ore per minute, the loop reinvests output into throughput and may compound unless time, capacity, or costs constrain it.

Why interviewers ask this: The interviewer evaluates whether the candidate can reason about resource topology, rates, bottlenecks, and reinvestment.

system-designdesign

A degenerate loop is a repeatable rules-valid interaction that dominates alternatives by producing disproportionate value, often with little cost or counterpressure.

If buying one seed costs 1 coin and selling its guaranteed harvest yields 3 coins immediately, repeated farming creates unbounded currency.
The problem lies in the closed conversion path and its net gain, not in whether the player intended to bypass the expected progression.
A strong strategy is not automatically degenerate, because limited land, elapsed seasons, or competing uses for seeds can preserve meaningful tradeoffs.
An exploit may use an unintended rule interaction, but its systemic significance comes from repeatability, scalability, and how strongly it invalidates other choices.

Why interviewers ask this: The interviewer evaluates whether the candidate defines degeneracy through incentives and loop structure rather than through player intent alone.

system-designdependencies

A dependency map should show which systems read, modify, enable, or constrain the states of other systems.

Nodes can represent systems or key state variables, such as stamina, movement, attacks, equipment weight, and regeneration.
Directed edges should name the relationship, such as equipment weight increasing stamina cost rather than merely linking equipment and stamina.
Cycles should be visible, such as combat rewards improving equipment while stronger equipment increases combat reward acquisition.
Critical assumptions can be marked on edges, such as movement requiring stamina to stay above zero, which exposes where one rule change propagates widely.

Why interviewers ask this: The interviewer evaluates whether the candidate can make causal dependencies explicit enough to reason about propagation and cycles.

system-designdesign

A design invariant is a property that must remain true across valid states and rule changes because other parts of the design depend on it.

A turn economy may require that every normal unit receives exactly one base action per round before bonuses are applied.
A trading system may require that every resource has at least one sink, preventing permanent accumulation with no decision attached.
An invariant differs from a tuning target, because damage averaging near 20 may change while health never dropping below zero remains structurally required.
Stating invariants narrows safe design space, such as allowing new movement abilities only if every reachable combat tile still has at least one counter-entry route.

Why interviewers ask this: The interviewer evaluates whether the candidate can separate structural guarantees from adjustable parameters and use them to preserve system coherence.

Sustained DPS measures long-run damage rate, while burst measures damage delivered inside a specified short window.

The weapon's sustained DPS is 120 / 0.75 = 160 damage per second.
The ability contributes 360 / 6 = 60 average DPS when used on cooldown, so combined long-run DPS is 220.
Firing the weapon and ability together deals an opening burst of 480 damage, but its burst DPS cannot be stated without defining the time window.

Why interviewers ask this: The interviewer evaluates whether the candidate separates immediate damage pressure from time-averaged output and handles cooldown arithmetic correctly.

Discrete-hit TTK depends on the number of lethal hits and the intervals between them.

The required hit count is ceil(1,000 / 180) = 6 because five hits deal only 900 damage.
At 2 shots per second, the interval between shots is 1 / 2 = 0.5 seconds.
Six shots contain five intervals after the time-zero shot, so TTK is (6 - 1) × 0.5 = 2.5 seconds.

Why interviewers ask this: The interviewer evaluates whether the candidate avoids treating discrete attacks as perfectly continuous DPS.

Effective health is the amount of incoming raw damage required to exhaust all protected durability.

With mitigation fraction r, effective health is H / (1 - r), so 1,200 / 0.75 = 1,600 raw damage.
If the shield also receives the 25% reduction, total effective health is (1,200 + 300) / 0.75 = 2,000.
If the shield is consumed before mitigation and takes raw damage, total effective health is 300 + 1,600 = 1,900.
The layer order must therefore be part of the formula rather than assumed from the displayed health total.

Why interviewers ask this: The interviewer evaluates whether the candidate can translate mitigation rules and layer ordering into comparable durability.

A breakpoint is a threshold where a small numerical change alters a discrete outcome such as hits required to kill.

Damage of 126 and 130 both require ceil(500 / damage) = 4 hits, so the increase does not change hit count.
The three-hit breakpoint is ceil(500 / 3) = 167 damage because 166 × 3 = 498 and 167 × 3 = 501.
With the first hit at time zero, crossing the breakpoint changes TTK from 3 × 0.8 = 2.4 seconds to 2 × 0.8 = 1.6 seconds.

Why interviewers ask this: The interviewer evaluates whether the candidate recognizes discontinuities hidden behind smooth stat increases.

dispersionprobabilityrewards

Expected value describes the long-run average, while variance describes how widely individual outcomes spread around it.

The expected value is 0.70 × 10 + 0.25 × 30 + 0.05 × 100 = 19.5 coins.
The second moment is 0.70 × 10^2 + 0.25 × 30^2 + 0.05 × 100^2 = 795.
The variance is 795 - 19.5^2 = 414.75 coins squared, and the standard deviation is about 20.36 coins.
Two reward tables can share an expected value but feel different because the one with higher variance produces more extreme outcomes.

Why interviewers ask this: The interviewer evaluates whether the candidate can quantify both average payout and outcome volatility.

probability

Independence means one attack's result does not change the probability distribution of the other attack.

For independent attacks, the probability that both crit is 0.20 × 0.20 = 0.04, or 4%.
The probability of at least one crit is 1 - 0.80^2 = 0.36, or 36%.
In general P(A and B) = P(A) × P(B | A), so if a first crit raises the second attack's crit chance to 50%, the probability of two crits becomes 0.20 × 0.50 = 10%.
Independence is the special case where P(B | A) = P(B), not the case where events cannot happen together.

Why interviewers ask this: The interviewer evaluates whether the candidate distinguishes independence, mutual exclusivity, and state-dependent odds.

probability

A hard pity converts the geometric distribution into a capped distribution with guaranteed success by a fixed attempt.

The chance of receiving the item during attempts 1 through 49 is 1 - 0.98^49 ≈ 62.84%.
The chance of reaching the guaranteed attempt is 0.98^49 ≈ 37.16%.
The expected attempts per item are sum from k = 0 to 49 of 0.98^k = (1 - 0.98^50) / 0.02 ≈ 31.79 attempts.
The guaranteed fiftieth attempt makes the probability of obtaining the item by attempt 50 equal to 100%, not merely 50 × 2%.

Why interviewers ask this: The interviewer evaluates whether the candidate can use complements and a capped geometric expectation for pity systems.

progression

The formula family determines how absolute and relative growth change across progression.

Linear growth such as V(n) = 100 + 20n adds a constant 20 per level and gives V(5) = 200.
Quadratic growth such as V(n) = 100 + 5n^2 has an increasing absolute increment and gives V(5) = 225.
Exponential growth such as V(n) = 100 × 1.10^n adds a constant 10% proportion and gives V(5) ≈ 161.05.
A logistic S-curve such as V(n) = 1,000 / (1 + 9e^(-0.5n)) starts at 100, grows fastest near n = 4.39 where V = 500, and approaches 1,000 without crossing it.

Why interviewers ask this: The interviewer evaluates whether the candidate links common progression equations to their rates of change and practical numeric behavior.

Diminishing returns reduce the marginal gain smoothly, while a soft cap changes the conversion rate after a threshold without stopping growth.

The diminishing-return function B(x) = 100x / (x + 100) gives B(100) = 50 and B(200) ≈ 66.7, so the second 100 input adds only about 16.7.
A soft-cap conversion E(x) = 0.4 × min(x, 50) + 0.1 × max(x - 50, 0) grants 0.4 percentage points per input before 50 and 0.1 afterward.
Under that soft cap, x = 80 grants 0.4 × 50 + 0.1 × 30 = 23 percentage points, so investment still helps but at a lower rate.

Why interviewers ask this: The interviewer evaluates whether the candidate can distinguish smooth marginal decay from a threshold-based rate change.

normalizationbalancing

Normalization places unlike stat profiles on common scales without requiring the profiles themselves to become identical.

Per 100 cost, faction A provides 120 health and 15 DPS, while faction B provides 90 health and 20 DPS, preserving a durability versus damage tradeoff.
Relative to means of 105 health and 17.5 DPS, A's normalized vector is approximately (1.143, 0.857) and B's is (0.857, 1.143).
With continuous damage and no other effects, A needs 90 / 15 = 6 seconds to defeat B and B needs 120 / 20 = 6 seconds to defeat A.
A single weighted average can hide meaningful strengths, so normalization should compare explicit dimensions and equivalent costs rather than erase faction asymmetry.

Why interviewers ask this: The interviewer evaluates whether the candidate can compare asymmetric power budgets mathematically without equating balance with identical stats.

Locked questions

21
How should a multi-currency economy assign distinct roles to soft currency, premium currency, and activity-specific tokens, and what trade-offs appear when their uses overlap?
economytokensdistinct
22
How do faucet and sink structures determine whether a game economy remains stable across early and late progression, and why is matching only their global totals insufficient?
progressioneconomy
23
How are currency velocity and inflation related in a game economy, and what does each metric reveal that the other does not?
economymonitoring
24
How should exchange rates between currencies or tradable resources be designed to limit arbitrage without eliminating meaningful conversion choices?
conversiondesign
25
How do inventory caps and resource decay change player behavior, and when do they create healthier spending rather than frustration or forced scheduling?
playerjobs
26
How can resource scarcity create meaningful decisions across progression without becoming a universal bottleneck, and how should scarcity differ between common and aspirational resources?
trackingprogression
27
How should vertical and horizontal progression complement each other, and what trade-offs arise when one becomes the dominant source of long-term motivation?
scalingprogression
28
Why does power creep cause content obsolescence, and which progression structures can add desirable rewards while preserving the value of older gear and activities?
rewardsprogression
29
What makes metagame goals effective across different time horizons, and how should they connect to core play without turning into disconnected checklists?
metagame
30
How can seasonal resets and catch-up systems restore competition and accessibility while preserving the achievement and investment of long-term players?
playersystem-designa11y
31
How do session, core, and meta loops work together to support retention without making play feel compulsory?
retentionsessions
32
What is the difference between intrinsic return motivation and scheduled return motivation, and how should a designer balance them?
designbalancing
33
How should a free-to-play game structure its value exchange between free players, paying players, and the game?
f2pplayer
34
Where are the ethical boundaries of monetization in game design, and which tests help define them?
monetizationdesigntesting
35
How do probability disclosure, pity systems, and duplicate protection change the fairness of a gacha system?
probabilitysystem-designmonetization
36
How can monetization coexist with competitive integrity in a multiplayer game?
monetizationpricing
37
What makes a sustainable LiveOps cadence, and how do cadence and rhythm differ?
liveops
38
Why might a game isolate an event economy from its main economy, and what trade-offs does that create?
economy
39
How should seasonal goals be structured to support different player motivations and available time?
motivationplayer
40
What trade-offs determine a healthy content cadence, and how can it avoid player fatigue?
player
41
What are narrative verbs, how do they differ from theme or plot, and how should they shape a game's system verbs? Give a concrete example of alignment and misalignment.
narrativesystem-design
42
What is systemic storytelling, and which design conditions let rule-driven events produce readable stories rather than random noise? Give a concrete example.
system-designdesign
43
How should choice and consequence be represented as game state so decisions remain causal and expressive without becoming a brittle tree of branches? Give a concrete state model.
causal
44
What is ludonarrative consistency, and how can rules, incentives, affordances, and feedback support or contradict a stated character fantasy? Use a concrete example and address the root cause rather than the symptom.
feedbackconsistency
45
Why is difficulty a multidimensional profile rather than a single scalar, and which dimensions should a designer separate? Give concrete examples of independent changes.
designdifficulty
46
How can accessibility assists reduce barriers without erasing the integrity of a mechanic, and when does an assist fundamentally change the tested skill? Give concrete examples.
a11ymechanics
47
How do information, motor, and cognitive accessibility barriers differ in game design, and why does each require a different intervention? Give one concrete example for each barrier.
a11ydesign
48
How should a balance spreadsheet be structured to remain auditable and serve as the design source of truth? Explain with a concrete parameter and derived-output example.
designbalancingspread
49
What must a mechanic specification contain to define player-facing behavior, complete rules, edge cases, and tunable parameters without becoming an implementation document? Use a concrete mechanic example.
playermechanics
50
What is sensitivity analysis in game balance, and how does it help define useful tuning ranges instead of changing values by intuition? Give a concrete example with a breakpoint or interaction.
balancingtuning
51
In the top competitive tier, 68% of players use the Scavenger perk with the Forge because selling reforged loot funds both repairs and further reforges, making most combat builds economically inferior. How would you fix this dominant strategy without flattening both systems?
playersystem-design
52
After a 20% buff to a support hero's shield duration, average duel length rose by 35% and timeout finishes doubled, although the hero's win rate moved only slightly. How would you assess and correct this second-order effect?
resilience
53
Players discovered that an expedition key yields an artifact that can be salvaged into enough dust to craft 1.15 new keys on average, so automated expeditions generate unlimited dust. How would you repair the economy?
playerartifactseconomy
54
Veterans in the top 5% hold 40 times more gold than active midgame players, and player-market prices now put basic upgrade materials out of reach for newcomers. How would you address this inflation?
player
55
New players encounter eight currencies before finishing their first hour, and usability tests show that they cannot explain which currency buys upgrades, cosmetics, or event rewards. How would you simplify the system?
rewardsplayersystem-design
56
Unlocking Chapter 6 requires an upgrade that takes about 18 full replays of Chapter 5, and the largest progression quit spike now occurs after the sixth replay. How would you fix the grind without making progression trivial?
progression
57
Season 4 heroes reduce a legacy raid's average clear time from 12 minutes to 3, and its once-desirable rewards are now ignored. How would you restore the old content's value without invalidating new heroes?
rewards
58
A stat is supposed to softcap at 500, yet 74% of top-ranked builds push it past 800 because post-cap points still multiply several damage bonuses. How would you make the softcap produce real build choices?
59
Only 38% of players clear the Chapter 7 boss on their first attempt, but 82% of those who reach Chapter 8 clear its finale immediately. How would you evaluate and fix this apparent difficulty drop?
playerdecision-makingdifficulty
60
Guild dispatches now provide 70% of weekly upgrade materials while active matches provide 15%, and many players log in only to collect dispatch rewards. How would you stop the metagame from displacing core play?
rewardsplayermetagame
61
A new-player funnel is stable through the tutorial until the material-selection step, where continuation fell from 78% to 44% after crafting choices were expanded. How would you respond?
funnelplayer
62
After daily quests were added, average session length rose from 18 to 25 minutes, but post-session enjoyment fell from 4.1 to 3.3 out of 5. How would you interpret and iterate on this result?
sessionsiteration
63
A hero has an 8% pick rate and a 58% win rate in ranked matches. The team suggests a buff because the hero is rarely chosen. What would you do?
64
Seven of eight playtesters call a boss attack unreadable, while live telemetry shows a stable average of 1.4 deaths at that boss. How would you resolve the conflict?
playtestingtelemetry
65
Onboarding completion is 76% on desktop and 52% for new mobile players, while returning mobile players complete it at 74%. What is your next move?
onboardingplayer
66
In a prototype playtest, 9 of 14 players fail the same puzzle, but the sample is too small for a reliable completion rate. How would you decide whether to change it?
playerprototypingprototypes
67
Day-one retention rose from 31% to 35% after a shop redesign, but a weekend event launched on the same day. Can the redesign be credited, and what would you do next?
retention
68
In a three-minute combat prototype, players dash about 12 times, but only 3 of 10 can explain when the stamina dash is strategically useful. Which events and metrics would you choose for the next iteration?
playerprototypingprototypes
69
An A/B test of a new mission picker raises mission starts from 62% to 68%, but early mission quits rise from 11% to 17%. Would you ship it?
ab-testing
70
After a boss update, completion rises from 64% to 70%, retries rise from 18% to 31%, and 6 of 8 playtesters say victories feel random. How would you prioritize these signals?
playtestingprioritization
71
In a live three-faction PvP mode, the strongest faction wins 56% of matches and the weakest wins 47%, with the spread persisting across ranks and maps. What would you change first, and how would you judge whether the fix worked?
spread
72
A rifle appears in 42% of loadouts but its users win only 47% of matches. Would you buff it, and what evidence would determine your change?
73
A mechanically demanding hero wins 44% of matches for beginners and 55% for expert players. How would you reduce the gap without removing the hero's mastery curve?
playermechanics
74
One side of a competitive map wins 54% of matches across 20,000 games, and the bias remains after controlling for team skill. What targeted map change would you try?
75
A defender's ability shuts down a diver in 72% of equal-skill engagements, and the diver has no practical way to bait or avoid it. How would you restore counterplay without deleting the counter?
engagement
76
Teams that secure the first major objective win 78% of matches, and later fights rarely change the outcome. How would you reduce snowballing without making the opening irrelevant?
77
A proposed damage buff moves a common armored target from four hits to three. The weapon is underperforming, but the new breakpoint sharply lowers time to kill. What would you change instead?
78
A specific three-character composition appears in 18% of matches but wins 61%, while each character is balanced in most other lineups. Where would you intervene?
balancingoop
79
A support character wins 53% in ranked play but 47% in casual queues, mainly because coordinated teams exploit ally-triggered effects. Would you balance the modes separately?
balancingdata-structures
80
A dominant strategy reaches a 57% win rate during a live competitive season, while a major balance patch is due in two weeks. What would you change now, and what would you defer?
balancing
81
New-player D7 retention is 19%, below the 24% target. Design one feature to improve it without relying on streak pressure or fear of missing out.
retentiondesignplayer
82
First-purchase conversion among players who complete Chapter 2 is 2.1%, against a 3% target. Redesign the first paid offer while keeping the decision informed and voluntary.
conversionplayer
83
Only 14% of players inactive for 7 to 30 days return during the weekly LiveOps event. Design an event that gives this segment a durable reason to come back.
playerliveopsdesign
84
After a limited event launched, sessions in the core mode fell by 18%. How would you redesign the event without simply removing its rewards?
rewardssessions
85
Only 22% of active battle pass owners reach the final tier. How would you adjust pacing without making completion automatic or selling the solution to excess grind?
pacing
86
Season completion among players who join after week four is 9%. Design a catch-up system that helps late entrants without devaluing early participation.
system-designdesignjoins
87
A cosmetic gacha banner converts 1.8% of eligible non-payers. Improve conversion while giving players a clear maximum cost and meaningful protection from bad luck.
monetizationconversionplayer
88
An offer containing exclusive combat power increased revenue per payer by 12%, but the advantage cannot be earned through play. What design decision would you make?
design
89
The total soft-currency supply rises 16% after each event because leftover event tokens convert automatically. Redesign the event closeout to respect earned value without feeding inflation.
evmtokenseconomy
90
Participation in the third consecutive LiveOps event is 31% lower than in the first. How would you revise the calendar to reduce fatigue without abandoning regular updates?
liveops
91
Engineering must start a three-state combo mechanic tomorrow, but the current spec only describes the intended feeling and you will not implement any code. What do you hand off so the mechanic can be built without designers answering basic questions throughout the sprint?
designmechanicsagile
92
Three days before art lock, playtests show that elite enemy silhouettes and attack telegraphs disappear against the new arena effects during eight-player fights, while the art team says reducing the effects will weaken the scene. How do you resolve it?
playtestingplayer
93
A product goal asks for 15% longer daily sessions, but a proposed mandatory reward screen after every encounter increases average session time while early tests show more interrupted runs and a higher quit rate after the third screen. What do you recommend?
rewardssessionstesting
94
Production says a five-variant companion command feature cannot fit into the eight days left before content lock and asks for a scope cut today. How do you reduce scope without losing the feature's design purpose?
design
95
Release is 48 hours away and the triage list contains a tutorial soft lock seen in 0.5% of sessions, an economy exploit used by 2% of testers, and a cosmetic alignment issue visible to 20% of players. How do you decide what to fix, cut, or defer?
playersessionseconomy
96
One week before console certification, a color-blindness review finds that red and green floor telegraphs are identical except for hue, but replacing the full effects package would miss the release date. What do you do?
97
A late change lets players cancel a revive into a dodge and grants brief invulnerability, but QA has only two days and asks which edge cases matter most. How do you prepare the mechanic for focused verification?
playermechanics
98
A contract mission feature is blocked because its inventory check, reward grant, and progress tracking depend on three systems with different owners, and one dependency will not be stable before feature lock. How do you unblock a releasable design slice?
system-designdesigndependencies
99
A developer wants to shorten a boss warning to simplify sequencing, an artist wants a larger effect for spectacle, and the product manager wants the encounter shipped unchanged, while playtests show players cannot connect the warning to the lethal attack. How do you resolve the disagreement?
playtestingconflictplayer
100
Six days before release, enemy tuning and economy rewards are still off target, chapters 3 and 7 have difficulty spikes, and only two moderated playtest slots remain. How do you plan the pass?
economydifficultytuning

Questions

1How can a game system be modeled through state, inputs, outputs, and connections to other systems?

2What is the difference between first-order and second-order effects in an interconnected game system?

3What makes a gameplay outcome emergent rather than explicitly authored?

4How do positive and negative feedback loops differ, and what effects do they tend to create?

5How is systemic depth different from mechanical complexity?

6What does it mean for two mechanics to be orthogonal, and why can orthogonality increase depth?

7How should a resource conversion loop be represented and analyzed as a game system?

8What is a degenerate loop or exploit from a systemic design perspective?

9What information should a dependency map capture for a set of interconnected game systems?

10What is a design invariant in a game system, and how does it guide rule changes?

11How do sustained DPS and burst damage differ, and how would you calculate them for a weapon that deals 120 damage every 0.75 seconds plus an instant 360-damage ability with a 6-second cooldown?

12How is time to kill calculated when damage arrives in discrete hits, and what is the TTK against 1,000 health for a weapon dealing 180 damage at 2 shots per second if the first shot lands at time zero?

13What is effective health, and how do mitigation and shield ordering affect it for a character with 1,200 health, 25% damage reduction, and a 300-point shield?

14What is a damage breakpoint, and where is the three-hit breakpoint against 500 health if damage is an integer and attacks occur every 0.8 seconds?

15How do expected value and variance describe a reward that gives 10 coins with 70% probability, 30 coins with 25% probability, and 100 coins with 5% probability?

16What is the difference between independent and conditional probability for two attacks with a 20% base critical-hit chance?

17How does a hard pity rule change loot probability if each attempt has an independent 2% drop chance and attempt 50 guarantees the item after 49 failures?

18How do linear, polynomial, exponential, and S-curve formulas produce different progression shapes, using n completed levels and a starting value of 100?

19How do diminishing returns differ from a soft cap, and how can each be expressed with a simple formula?

20How can normalization support asymmetric faction balance if a 100-cost unit from faction A has 120 health and 15 DPS while a 100-cost unit from faction B has 90 health and 20 DPS?

Locked questions