Guide

Game skill check and dice roll systems explained

Harbor Chronicles' Act III council vote shipped with a single persuasion check at difficulty 18. Players who invested in charisma still failed 40% of the time on a story-critical branch; failures dumped them into a generic “the council refuses” line with no alternate path. Playtesters called the moment “unfair dice,” not “hard choice.” The refactor replaced one opaque roll with a skill check system that surfaces the target number, shows modifiers from reputation and prior quests, grants advantage when the player uncovered blackmail evidence, and maps outcomes into four bands: critical success (full alliance), success (narrow approval), partial success (delayed vote with a side quest), and failure (hostile faction but not a hard block). Re-rolls cost influence points earned in Act II. Optional checks on side dialogue still use hidden rolls; only the main vote is fully transparent. Story completion rate on the council arc rose from 61% to 89%.

Skill checks are how RPG design converts stats and player choices into narrative branches. A roll is a contract: the player spends build investment or accepts risk for a chance at a better outcome. When the math is invisible or failure is a dead end, the contract breaks. This guide covers resolution mechanics (d20, dice pools, percentile), difficulty classes and modifiers, opposed and contested rolls, advantage and reroll economies, hidden vs visible feedback, critical success and failure, integration with locks and dialogue, the Harbor Chronicles council refactor, a technique decision table, pitfalls, and a production checklist.

What skill check systems are

A skill check (or ability test) compares a character's relevant attribute plus situational modifiers against a target difficulty. The engine rolls randomness, applies bonuses, and routes the result to success, failure, or graded outcomes. Core responsibilities:

Gate optional power — higher stealth opens a back route; high medicine avoids a consumable tax.
Branch narrative — persuasion, intimidation, and insight checks unlock different dialogue trees.
Resolve ambient uncertainty — search checks for hidden loot, perception for traps.
Express build identity — a thief who auto-passes tier-1 locks in lockpicking systems feels the stat investment pay off.

Skill checks are not combat damage rolls. Combat usually has its own loop (accuracy, mitigation, crit tables). Skill checks answer “can this character accomplish a non-combat task under pressure?”

Resolution mechanics: d20, pools, and percentile

d20 plus modifier (D&D-style)

Roll 1d20, add attribute modifier and proficiency, compare to a difficulty class (DC). Natural 20 and natural 1 often trigger critical success or failure regardless of modifiers. Simple to explain, wide variance on a single die — a +5 expert still fails DC 15 one time in four.

Dice pools (count successes)

Roll N dice (often d6 or d10); each die meeting a threshold (e.g., 5+) counts as a success. Compare success count to difficulty. Pools smooth variance: more dice means more predictable averages. Used in World of Darkness, Shadowrun, and many narrative RPGs.

Percentile (d100)

Roll 1–100; succeed if roll is at or under skill value (e.g., 62% persuasion). Familiar to CRPG players from classic Fallout and Warhammer-style systems. Easy to display as a percentage chance in UI.

Narrative dice (success, advantage, complication)

Genesys and FFG Star Wars use custom dice with symbols for success, failure, advantage, and threat. One roll produces mechanical and story hooks (“you convince the guard but drop your forged papers”). Higher authoring cost, richer fail-forward outcomes.

Pick one core resolver for your game and document it in a player-facing glossary. Mixing d20 dialogue with d100 lockpicking without explanation confuses players who cannot transfer intuition between systems.

Difficulty classes, modifiers, and graded outcomes

A DC (or target number) encodes task hardness. Publish internal bands so designers stay consistent:

Easy (DC 10 / 50% base) — routine tasks for a trained character.
Medium (DC 15 / 65%) — meaningful risk; failure should be interesting, not fatal.
Hard (DC 20 / 80%) — requires build focus or preparation.
Extreme (DC 25+) — reserved for capstone moments; offer bypasses via quests or items.

Modifiers stack from equipment, buffs, prior choices, environmental tags (rain penalizes acrobatics), and faction reputation. Cap total bonus swing so a +15 modifier cannot trivialize every check in the game. Prefer graded outcomes over binary pass-fail: beating DC by 5 might grant bonus loot; missing by 2 might partial success with a complication.

Opposed rolls and passive defenses

In an opposed check, two sides roll and compare totals (stealth vs perception, deception vs insight). Ties need a rule: reroll, favor defender, or favor active player. Opposed rolls feel fair in PvP contexts but hide the target number in PvE — players cannot compute odds without feedback.

Passive scores (passive perception = 10 + modifier) avoid rolling for every NPC scan. The sneaking player rolls stealth once against a fixed passive value. Reduces dice spam in stealth-heavy levels tied to stealth mechanics.

Advantage, disadvantage, and reroll economies

Advantage (roll twice, take better) and disadvantage (roll twice, take worse) express situational edge without permanent stat inflation. Grant advantage from preparation: studying a blueprint before a hacking check, flanking in dialogue with an ally, or using the right tool.

Rerolls should cost a resource — inspiration points, fate tokens, consumable luck charms — so players cannot brute-force critical branches. Show remaining rerolls in UI before the player commits to a high-stakes roll. Cooldown rerolls on the same NPC prevent save-scum loops.

Hidden vs visible rolls and player trust

Visible checks display base stat, modifiers, DC or success chance, and the die result. Players learn the system and blame build choices instead of “rigged RNG.” Use for major story gates and build- defining moments.

Hidden checks preserve surprise (insight detecting a lie, perception spotting a trap). Show outcome fiction, not math: “You sense hesitation” vs “Insight check failed.” Never hide rolls on mandatory main-quest paths without a fail-forward branch.

Optional debug overlay in accessibility settings builds trust with min-max players without cluttering default UI.

Critical success, critical failure, and fail-forward

Critical success (nat 20 or excess successes) should reward without breaking balance: bonus information, shortcut, or cosmetic flair — not a free boss kill in dialogue.

Critical failure (nat 1 or dramatic botch) works best when it advances story with a complication (alerted guards, broken lockpick jam) rather than hard-blocking progress. Pair with difficulty curves so early-game crit fails teach consequences without ending runs.

Fail-forward means failure changes the scene instead of stopping it: the council vote fails but opens a bribery side quest. Players remember outcomes; they forget dice numbers.

Harbor Chronicles council persuasion refactor (worked example)

Before: one Persuasion vs DC 18 check, binary pass-fail, no modifier breakdown, failure = generic rejection and stuck main quest.

After:

Surface DC 16 base; +2 if player presented evidence (flag); +1 per allied faction from Act II reputation (capped at +3).
Advantage if blackmail dossier completed; cannot stack with max reputation bonus (designer cap prevents +8 swings).
Outcome bands: crit success (unanimous vote + unique item), success (majority), partial (delayed vote + infiltration quest), failure (hostile council but prison break route unlocks).
Influence reroll costs 2 influence; max one reroll per playthrough on this scene.
Side dialogue uses hidden Insight checks; main vote is fully visible.

Telemetry: 89% arc completion, 12% reroll usage, 4% crit success. Players cited “I knew what I was betting on” in surveys.

Technique decision table

Approach	Best when	Skip when
d20 + DC	Tabletop-familiar CRPG, clear modifier UI	Players hate high single-die variance
Dice pool	Build scales with multiple small bonuses	Mobile UI cannot show many dice cleanly
Percentile	Show exact % chance; classic CRPG feel	Narrative dice with symbol side effects
Narrative dice	Story-first RPG, rich fail-forward text	Tight production budget for outcome variants
Stat threshold (no roll)	Deterministic mastery rewards	You want tension on pivotal moments
Player-skill QTE	Action games, reflex fantasy	Narrative RPG where build identity matters

Common pitfalls

Single-point story gates — one roll blocks hours of progress; always offer alternate routes or fail-forward.
Modifier soup — seven unexplained bonuses; show each source in a tooltip.
Inconsistent DC bands — DC 15 means easy in Act I and impossible in Act III without relabeling tiers.
Reload encouragement — hidden rolls on dialogue trees invite save-scumming; visible odds or reroll costs reduce it.
Checks where choices suffice — if the player already did the right prep, auto-success beats another die.
Combat stats on social checks — strength intimidating a child reads as nonsense; gate options by logical attributes.
No failure content — failing should spawn new scenes, not a red “failed” toast.

Production checklist

Document DC bands and when to use visible vs hidden rolls.
Cap modifier stacking; list allowed sources in a design spreadsheet.
Define tie rules for opposed checks and passive score defaults.
Map graded outcomes (success tiers) for every main-quest check.
Fail-forward every mandatory gate; verify in narrative flowcharts.
Price rerolls with a finite resource; log reroll rate in telemetry.
Playtest with min and max builds at each act boundary.
Expose success chance or DC in UI for high-stakes visible checks.
Separate combat to-hit from skill resolution code paths.
QA crit success and crit failure text for all flagged scenes.

Key takeaways

Skill checks convert build investment into narrative branches — the player must understand the bet.
Harbor Chronicles raised council arc completion from 61% to 89% by surfacing modifiers, grading outcomes, and fail-forwarding failure.
Pick one core resolver (d20, pool, percentile) and teach it consistently across dialogue, locks, and exploration.
Hidden rolls for surprise; visible rolls for trust on story-critical paths.
Fail-forward beats hard fail — complications are more memorable than game overs.