Chapter 32 Quiz

Thirteen questions. Mix of multiple choice, short answer, and short essay. Answers and explanations follow.

Q1. The chapter argues that "balance" does not mean "every option is equally good." What does it mean instead? (Short answer.)

Q2. Which of the following best describes meaningful choice in a balanced game?

A. Every option produces the same outcome on average. B. Every option is the best choice in some situation. C. Every option is mathematically identical when stat-totals are summed. D. Every option has the same cost.

Q3. A unit deals 80 damage per hit and attacks every 1.6 seconds. Compute its DPS.

Q4. A defender has 200 HP and 25% damage reduction. Compute its effective HP against damage that is reduced by the defender's reduction stat.

Q5. In a 1v1 between an attacker dealing 50 DPS and a defender with the EHP you computed in Q4, what is the time to kill?

Q6. Why does the chapter consider the spreadsheet "external memory" rather than just a record-keeping tool? (Short answer.)

Q7. League of Legends is described as having a per-elo balance problem. Explain what this means and why the same kit can be overpowered at one rank and underpowered at another. (Short essay, 3-5 sentences.)

Q8. Which of the following is not one of the four parts of a well-written patch note, according to the chapter?

A. What changed. B. Why it changed. C. What you expect. D. Who complained about it.

Q9. The chapter warns about the "over-nerf cycle." Describe what this is and propose one strategy to avoid it. (Short answer.)

Q10. Why does asymmetric balance (e.g., Dead by Daylight's killer-vs-survivors) require more data per balance decision than symmetric balance? (Short answer.)

Q11. A 30% chance of a critical hit at 2.5× damage produces what expected damage multiplier compared to no crits? Show your work.

Q12. In your progressive project, you have hard-coded enemy HP values in your Enemy.gd script. The chapter recommends moving these values to a CSV. Name two concrete benefits of doing this.

Q13. The chapter ends by noting that next chapter (Ch 33) covers ethics, including "the dark version of the skills you just learned." In one or two sentences, what is the difference between balancing for skill expression and balancing for monetization extraction?

Q14. True or false: a unit that the simulator says is balanced but that players consistently report as feeling terrible is a unit that should be left alone, because the data is more reliable than feelings.

Q15. The chapter argues that the proper response to a single playtester complaining the boss is too hard is not to halve the boss's HP. Why? What is the proper response?

Answers and Explanations

A1. Balance means every option is viable, interesting, and situationally optimal — it produces meaningful choice, not symmetric choice. Equal-power options would mean the choice is meaningless; the goal is that each option is the best one to pick in some context, with counters and tradeoffs that reward player knowledge and judgment.

The trap here is the player-facing definition of "fair," which most players read as "equal." Designers have to internalize the more rigorous definition.

A2. B. Every option is the best choice in some situation. This is the working definition of situationally optimal. Option A produces the lottery problem (no strategic content). Option C is the design-by-stat-budget mistake. Option D is irrelevant — costs are part of the equation, not the goal.

A3. DPS = 80 / 1.6 = 50 DPS.

The mechanical step you should be able to do in your head. Damage per hit divided by interval between hits.

A4. EHP = HP / (1 − damage reduction) = 200 / (1 − 0.25) = 200 / 0.75 = 266.67 (effectively 267) EHP.

The percentage-reduction formula. Note that 25% reduction does not give 25% more EHP; it gives 33% more, because the defender takes 75% of incoming damage. This nonlinearity is why armor scaling is delicate.

A5. TTK = EHP / DPS = 267 / 50 = 5.34 seconds.

Combine the two derived metrics for the player-facing number. If your fights last around 5 seconds, this matchup is in the right range; if your design target is 2-second fights, this is a slugfest.

A6. The human brain cannot hold and compare twenty-plus variables across forty units. The spreadsheet stores the full data state (current values, derived metrics, comparison columns) so the designer can query and reason about it without having to remember it. When one cell changes, formulas cascade and the designer can immediately see the downstream effect of a change before committing to it. Without the spreadsheet, the designer relies on memory and heuristics, which produces unbalanced games.

A7. Per-elo balance refers to the observation that a champion's effective power varies with player skill. Champions with high skill ceilings (Riven, Azir) are stronger in expert hands because they reward execution that lower-skill players cannot perform. Champions with high skill floors (or with kits that are easy at low skill but punishable at high skill) are stronger at lower elos because the punish never comes. The same numerical kit therefore produces different win rates at different ranks. Riot solves this by collecting data per rank and targeting changes at the aspect of a kit that scales with skill — nerfing animation cancels hurts pro-level Riven much more than bronze-level Riven, while nerfing base damage hurts both equally.

A8. D. Who complained about it. The four parts are: what changed, why it changed, what you expect, and what you will watch. Naming the complainant is community-management theater, not balance documentation, and often counterproductive (it singles out one voice and makes the change feel reactive rather than data-driven).

A9. The over-nerf cycle is a pattern where a dominant unit is nerfed too hard, becomes weak, gets buffed, becomes dominant again, gets nerfed again — oscillating around the balance line because each correction overshoots. To avoid it, make smaller changes. A 3% nerf is almost always preferable to a 10% nerf, because it is easier to extend a small nerf next patch than to walk back a large one. Patience and small step sizes prevent oscillation.

A10. In symmetric games, every matchup is mirrored — if X beats Y, the win-rate data for X-vs-Y also gives you Y-vs-X (just inverted). In asymmetric games, the two sides have completely different toolkits, so X-as-killer-vs-Y-as-survivor is a fundamentally different scenario from Y-as-killer-vs-X-as-survivor. Each pairing must be evaluated independently in both directions, and the data matrix grows multiplicatively with each added option, multiplying the number of playtest hours and bot matches required to characterize the balance state.

A11. Effective multiplier = (probability of crit × crit multiplier) + (probability of no-crit × normal multiplier) = (0.3 × 2.5) + (0.7 × 1.0) = 0.75 + 0.7 = 1.45×. The unit deals 45% more damage on average than a no-crit equivalent.

This is the calculation that lets you compare crit-builds to flat-damage-builds at the design level. A 30% crit chance at 2.5× sounds bigger than a flat +45% damage but they are mathematically equivalent on average.

A12. Two of any of the following are acceptable: - Designers (and non-coders) can tune the game without touching code. - Hot-reload becomes possible — change the CSV, see the results without recompiling or restarting the engine. - Source-controlled diffs of balance changes are clearer (you see "hp: 100 → 150" rather than "an integer changed somewhere in a script"). - All balance-relevant numbers live in one place, so audits, comparisons, and pattern-spotting are easier. - The CSV can be edited by tools the team already uses (Excel, Google Sheets), enabling parallel work by multiple designers. - The same CSV can drive simulations and external tools (e.g., a Python balance simulator) without re-implementing logic.

A13. Balancing for skill expression aims to make every option viable so the player's choices and execution determine the outcome — the goal is meaningful play. Balancing for monetization extraction aims to make weak free options and strong paid options, so spending money is the path to power — the goal is conversion. Same techniques (spreadsheets, derived metrics, per-segment analysis), opposite ethics. Chapter 33 explores this in depth.

A14. False. The chapter explicitly warns against this — "spreadsheet idolatry." If players universally report a unit as feeling terrible, the players are right and the model is missing something the model does not capture (psychology, learning curve, expected behavior, frustration). Update the model to include the missing variable, and rebalance from the new model. Player feeling is data, not a verdict, but it is real data.

A15. Sample size of one. A single complaint is an observation, not a pattern. Halving the boss's HP based on it would be an over-correction motivated by anecdote. The proper response is to aggregate — wait until you have data from five or more playtesters before drawing a conclusion, and let the pattern (or its absence) tell you whether the boss really is too hard, or whether one tester had a bad run. Acting on n=1 produces the over-nerf cycle.

Scoring (for instructor use)

• 13 questions × 7-8 points each = ~100 points.
• Multiple-choice questions (Q2, Q8, Q14): 5 points each.
• Computational questions (Q3, Q4, Q5, Q11): 7 points each, partial credit for correct method.
• Short-answer questions (Q1, Q6, Q9, Q10, Q12): 8 points each.
• Short-essay questions (Q7, Q13, Q15): 10 points each, graded on insight and accuracy.

A passing score reflects that the student understands what balance is, can do the math, can recognize common failure modes, and can connect balance to the broader ethical and production context.