Research for a quick collectible card game
Research date: 3 October 2026. This document compares established card game systems and turns their lessons into choices for a new two-player game with approximately 100 unique cards. A 10–15 minute match and a five-minute first explanation are proposed targets, not measured results. The user has confirmed a tabletop game played digitally: its rules must also resolve with physical cards, ordinary tokens, and a clear rulebook. The rules system remains undecided.
The strongest starting hypothesis is that we should make resource access reliable, keep each action short, and give players competing uses for the same pieces. Strategic variety can come from timing, positioning, spending cards, and contesting the score. It does not require a large list of keywords or several unrelated victory systems. The current recommendation is to test the three-site contest first, using the 21-card pilot. The comparisons below describe a broader design space; systems.md defines the current proposals and playable rules.
Published mechanics below are cited to rules and publisher explanations. Recommendations, advantages, risks, budgets, and proposed tests are our design judgments. Historical sources establish design ideas and ordinary mechanics; they are not claims about today's competitive metagame or every exception introduced by individual cards.
What existing games contribute
| Game | Sourced mechanics | Design interpretation for our game |
|---|---|---|
| Magic the Gathering | Lands provide mana; normally one land can be played on each of your turns. Attackers attack the opponent and the defender chooses blockers. Surviving creatures lose marked damage at the end of the turn. The comprehensive rules define priority and a stack for spell and ability resolution. Basic play, rules 117 and 405 | Reusable resources and defender choices support planning and bluffing. Drawing lands adds a separate source of opening variance. A general response stack creates many timing obligations; importing it would spend a large part of our learning budget. |
| Pokémon TCG | Players normally attach one Energy card each turn. Only the Active Pokémon normally attacks; the Bench can hold five. An attack ends the turn. Taking all six Prize cards wins, with additional loss conditions for having no Pokémon or failing the opening draw of a turn. Energy is usually retained after attacks. Official rulebook, pages 7–14 | A clear focal creature makes the battle readable and creates attachment. Investing repeatedly into it raises the cost of losing it. Bench abilities and search effects can still create lengthy preparation before a single attack. A prize system combines progress with additional cards, which can compound an advantage. |
| Hearthstone | The baseline system uses increasing, refreshing mana, minions that can attack minions or the opposing hero, and usually 30-card decks. A 2020 research paper documents those ordinary rules; Blizzard's designer guide discusses the tradeoff between developing the board and drawing cards, as well as aggressive and controlling plans. Rules background in Eger and Sauma Chacón, designer guide | Automatic resources remove the need to draw dedicated resource cards. Attacker-selected trades are easy to demonstrate, but an early board lead may remove new enemy pieces before they contribute. Cheap recovery tools and delayed activation deserve testing. |
| Marvel Snap | The official quick guide describes 12-card decks, simultaneous decisions, three locations, and normally six turns. Energy grows by one each turn; controlling two locations wins. Locations have separate rules and normally reveal during the first three turns. Snapping raises the cube stakes and retreating limits losses. Official quick guide | A fixed clock and a small board are a strong route to predictable duration. Two-out-of-three scoring gives players permission to concede one area. Hidden commitments create reads with little interruption, although reveal order becomes a rules issue. Wagering adds a separate skill; it is optional for our first prototype. |
| Disney Lorcana | A player can normally put one eligible card into the inkwell each turn. A character can exert to quest for lore or challenge an opposing exerted character; the first player to 20 lore wins. Decks contain at least 60 cards, at most four copies of a card, and one or two ink types. Official quickstart, pages 12–20 | Scoring and combat compete for a character's action. Scoring exposes it to attack, creating interaction through the core rule. The ink decision creates sacrifice without dedicated resource cards, but eligibility adds another property to learn and can leave hands unable to develop. |
| Star Wars Unlimited | Players alternate single actions. Any hand card can become a facedown resource; setup resources two cards, and regroup allows another while drawing two. Units attack the enemy base or units in their own arena. Taking initiative secures the first action next round but ends your actions this round. Standard Premier uses a leader, base, and at least 50 draw cards. Gameplay explanation, official introduction | Alternating actions create frequent answers without needing to interrupt resolution. Resourcing makes each hand flexible while preserving sacrifice decisions. Initiative makes stopping meaningful. More units also mean more actions, so small turns do not automatically produce a short game. |
| KeyForge | Each turn activates one of the deck's three houses. Cards do not have ordinary play costs; the active house restricts which can be played and used. Creatures can reap instead of fight. A normal key costs six Æmber, forged at the start of a turn, and the third key wins. Publisher overview | An action restriction can replace a mana system. Delayed scoring creates an explicit response opportunity. Reaping makes noncombat progress fundamental. We can borrow these principles without adopting fixed generated decks, which would reduce the deckbuilding expression requested here. |
| Flesh and Blood | Cards have pitch values for generating resources, and combat permits defense with eligible cards from hand followed by a reaction step. Pitched cards return to the bottom of the deck during the end phase. Card properties, combat, turn structure | Reusing a card as action, payment, or defense gives a small hand many decisions. It also asks players to value several uses simultaneously. This is a promising advanced branch, but too much of that system at once would work against a gentle first game. |
The Pokémon PDF above is a Sword and Shield era rulebook consulted for its foundational rules. The Hearthstone research paper is historical and is used only for its baseline game description. Lorcana's PDF was followed from its current official resources page. The Snap source is an image guide; its core sections were inspected visually. We should recheck the current rulebooks before publishing detailed teaching material about another game.
Choose the resource system first
The resource system shapes deck construction, turn length, and how often a player can enact a plan. Changing it late can invalidate most card costs.
| Candidate | Decisions it creates | Main risk | Recommendation |
|---|---|---|---|
| Dedicated resource cards | Resource ratio, colors, mulligans, whether to keep a speculative hand | Games can be dominated by too few or too many resource draws | Keep as a comparison only; it adds friction to our short-match target |
| Automatic growth each round | How to spend a predictable budget, whether to save cards for later | Repeated on-curve play can become obvious; little resource sacrifice | Simpler comparison if banking causes confusion; represent it with a public resource track |
| Any card becomes a permanent resource | Which future option to sacrifice; when to stop growing | New players may regret a commitment before understanding its value; hands shrink quickly | Current site-game prototype choice, to test hand management |
| Discard or pitch to pay | Trade immediate options for an explosive turn | Counting and hand valuation take longer; large turns need limits | Save for a focused advanced experiment |
| One selected faction or action category per turn | Group actions efficiently or activate the most urgent piece | Hand distribution controls available actions; free chains can become lengthy | Interesting identity, but requires a separate card economy |
Automatic growth and universal resourcing are the leading alternatives. The current site-game prototype starts with universal resourcing and generic costs. If banking repeatedly confuses players, compare automatic supply within that same site game before changing the combat framework. Separate faction identity from whether a player drew the right resource color. If a resource comparison also changes the draw rate, record both changes rather than claiming we tested only one isolated rule.
Do not adopt universal resources, permanent equipment, creature evolution, and multi-use hand defense together in the first prototype. Each is individually interesting; together they require a beginner to evaluate too many future uses of every card.
Interaction without excessive interruption
There are three distinct things we can mean by interaction: choosing targets on our own action, answering an opponent between their actions, and interrupting an action before it resolves. The third is only one option.
Whole turns are easy to follow and cheap to implement. Their downside is the interval during which the opponent observes a sequence without making decisions. We can cap board size, search, and free effects to contain this.
Alternating actions offer an appealing middle ground. A summon resolves, the opponent acts, then we decide whether to attack or develop another piece. An effect that removes a unit is a response to the board rather than an interruption of a pending spell. For a first version, resolve each action completely, give each ready unit at most one activation per round, and make passing rules explicit.
Simultaneous commitments suit lane control and anticipation. They reduce back-and-forth waiting, but a digital implementation must define locks, reveals, and disconnect handling; paper needs a clear reveal procedure. Test open alternating play first if the project needs the simplest physical prototype.
A general stack permits counterspells and intricate combat tricks, but it adds repeated decisions about who can act and when. If we later need a surprise defense, a single defined response window is a smaller experiment. For example: once an attack is declared, the defender may play one reaction, it resolves, then combat resolves. Do not allow reactions to reactions until playtesting demonstrates a need.
Combat is a choice about agency
| Structure | Player experience we expect | Design pressure |
|---|---|---|
| Defender assigns blockers | The defender has agency even with no hand cards | Evasion and removal must break board stalls; group blocking adds arithmetic |
| Attacker chooses targets | Each action has an immediate, legible outcome | A leading board can repeatedly remove new units; guard and recovery tools matter |
| Attack only used units | Scoring or attacking exposes a piece | A player may refuse to activate; the scoring clock must reward taking risks |
| One active fighter plus a reserve | Strong attachment and a clear focal battle | Replacement fighters and upgrade access must keep a knockout from ending participation |
| Compare power at objectives | Positioning and timing matter without routine damage counters | Movement, sacrifice, and limited removal must prevent every deck from simply playing the most power |
These are design predictions, not measured comparisons. For our short game, a board cap of roughly four to six units per player is a reasonable test parameter. Persistent damage makes several small attacks meaningful and preserves injury stories; clearing damage reduces tracking. We should select one convention and show it directly on the prototype.
Win conditions and genuinely different plans
A game can support many strategies while sharing a single victory condition. Aggression, attrition, tempo, engines, and combination decks can all pursue damage or points through different decisions. Separate victory rules create variety only when the opponent has meaningful ways to contest them.
For a first set, choose one main progress system and allow at most one visibly telegraphed alternative. A progress race can permit attacks to slow a scorer; a damage duel can include a public project that takes several rounds to complete. Avoid an alternative that asks the opponent to run one obscure answer or lose.
Suggested strategic families to support during testing:
| Plan | What the pilot does differently | What the opponent can exploit |
|---|---|---|
| Aggression | Converts early cards into immediate progress | Runs out of cards or loses efficiently to a defense |
| Tempo | Moves, delays, or removes a blocker while advancing | Delay loses value if it fails to produce points or damage |
| Midrange | Makes sturdy threats and pivots between racing and trading | Can be outpaced or outscaled |
| Control | Spends cards slowing the opponent, then closes with a threat | Must surrender some early progress; cannot answer everything |
| Engine | Invests in pieces that improve later actions | Setup pieces are vulnerable before the payoff |
| Sacrifice and recursion | Converts used pieces into a second benefit | Limited recovery targets and explicit limits prevent repeated loops |
| Mobility or evasion | Changes where pressure appears | Low raw strength and position-dependent rewards |
| Combination | Assembles two or three pieces for a larger effect | Visible setup and several broadly useful disruption tools |
Not all eight need independent prebuilt decks on day one. The current pilot begins with two decks; after its rules and pacing work, expand toward four contrasting plans and see whether mixed decks produce new decisions. Count a strategy as distinct when its sequencing and priorities change, not merely when its card names or color change.
Blizzard's class-identity discussion explicitly treats strengths and weaknesses as useful design constraints. Our application is to give every faction tools it excels at and a problem it solves poorly, while ensuring neutral cards do not erase those weaknesses. This is a design principle drawn from the publisher's explanation, not evidence that any particular faction allocation will be balanced. Blizzard on class identity
Keep uncertainty interesting
Separate opening uncertainty, decision uncertainty, and outcome randomness. A shuffled deck provides the first. Hidden hands or simultaneous choices provide the second. Coin flips and random targets add the third. We do not need much outcome randomness if the first two already make matches unpredictable.
Our proposal is a generous, simple opening redraw and a high density of useful early cards. A card with two useful modes can soften poor draws, but every mode adds reading. Put early selection on a small number of straightforward cards. Avoid repeated whole-deck searching, both because it slows physical play and because it can make combinations too reliable.
As an illustrative calculation, in a shuffled 30-card deck with no redraws, searching, or extra draws, the probability of seeing at least one particular card after seeing 12 cards is:
| Copies in the deck | Probability of seeing at least one |
|---|---|
| 1 | 40.0% |
| 2 | 64.8% |
| 3 | 79.9% |
| 4 | 88.8% |
These are exact hypergeometric calculations rounded to one decimal place: 1 − C(30 − copies, 12) / C(30, 12). They do not predict win rates or the chance of completing a multi-card combination. A two-copy limit leaves room for discovery; it also means a deck cannot depend on seeing a single named enabler every game. Support important plans through several interchangeable cards.
Spend the complexity budget deliberately
Mark Rosewater distinguishes comprehension complexity, board complexity, and strategic complexity. A card can be short yet make the whole battlefield much harder to reason about. His discussion of New World Order describes limiting the first two at common while preserving strategic choices. New World Order
The current pilot uses only Guard and Arrival as reminder terms. The set plan reserves at most two additional named mechanics for the first 100 cards, one primary effect on most cards, and a small number of visible ongoing effects. These are starting budgets to test. A short triggered sentence is preferable to a new keyword unless the keyword meaningfully saves repetition across the set.
Across the systems surveyed, examples of inexpensive strategic depth include placement, choosing between a scoring action and combat, committing a card as a resource, and taking initiative. These are alternatives, not a combined rules list: the current site prototype scores exhausted units and alternates initiative automatically. Examples that consume more tracking include several different counters on the same unit, global modifiers affecting many subsets, effects remembered for multiple turns, and optional triggers nested inside other triggers.
For a digital tabletop prototype, surface ready status, valid targets, remaining resources, score timing, and the consequence of passing. Keep full card text readable without depending on art, tiny symbols, or color alone. An attractive frame that obscures the next legal decision works against the learning target.
Every rule must also resolve on a physical table. Track resources with a dial or tokens; use a small, consistent set of counters; define the order of simultaneous effects. Avoid secret random generation from an external collection, permanent modifications that disappear into a shuffled deck without a physical marker, and decisions that require a software-only history log. Browser automation may help players follow rules, but it must not be required to understand what happened. Validate sample cards at actual print size as well as on screen.
What approximately 100 cards can support
Treat 100 as unique functional designs. Alternative artwork, foil treatments, tokens, and reference cards should be counted separately. A compact first set benefits from broad reusable effects that serve several decks, because isolated named combinations consume scarce slots.
For example, four factions of 20 cards plus 20 shared cards gives 100. Five factions of 16 plus 20 shared also gives 100. With a two-faction deck rule, those models offer 60 or 52 eligible designs before other restrictions. Four factions offer six possible pairs; five offer ten. These are combinatorial choices, not promises of six or ten viable archetypes. More factions mean more identities to teach and fewer designs per identity.
Before filling the set, make a card-role spreadsheet. Record cost, faction, type, primary role, strategies served, intended answer, rules text, and complexity. Each intended strategy needs early plays, payoffs, and recovery; each strong permanent needs a broadly usable answer. A rarity label should express presentation and acquisition, not excuse an unanswerable power level.
Do not assume the game needs randomized packs to be collectible. The first prototype should provide every test card. Later distribution can be discussed separately: complete sets, faction packs, or randomized boosters make different demands on card access and duplicate limits.
Three current structures to compare
These summarize the proposals in Mechanic structures, which is the source of truth for rules and test order. A and C are sketches; B has the first playable rules. Theme can be tested independently of them.
| Structure | Starting test parameters | Why test it | Main question |
|---|---|---|---|
| A — Focus duel | 30 cards, two-copy limit, automatic energy capped at six, full own turns, five units and one relic per player, 20 stronghold health | Familiar creatures and tactical answers; fallback if site tracking proves too demanding | Can matches finish in 10–15 minutes while slower strategies still have room? |
| B — Three-site contest | 30 cards, two-copy limit, bank any card once per round, three sites with two friendly slots each, six rounds and three alternating actions per player per round | Placement, removal, movement, and card sacrifice create competing priorities within a fixed clock | Does scoring reward early involvement without making a lost round unrecoverable? |
| C — Creature relay | 24 support cards plus three openly chosen creatures, fixed three focus per turn, one Active and two reserves, training and switching, first to three opposing knockouts | Strong character attachment; an alternative if individual creature development becomes the main product goal | Does training feel exciting without requiring the right named card in hand? |
The current site game scores every round, with site values of 1, 1, 2, 2, 3, and 3. Exhausted units contribute their full current power: scoring requires no separate activation. Normal movement costs one supply and exhausts a ready unit. Tied sites award nothing, and a final score tie is a draw. End-only scoring or two or three checkpoints remain possible later experiments, not rules for the first test.
For the creature relay, a universal training rule can test the feeling of growth before investing in many named evolution chains. A roster that is always available reduces draw dependence but concentrates repetition and balance pressure into those starting creatures. The proposed fixed focus budget should be tested separately from the banking model in B.
Practical prototype and validation sequence
- Use the 21-card pilot, which supplies two complete 30-card decks for B. Begin on paper or a plain browser tabletop.
- Run two open-hand learning games and resolve genuine contradictions before tuning numbers.
- Record six games as three paired matches, swapping decks and starting initiative. Log duration, unused actions, resource friction, scoring progression, and when the outcome felt inevitable.
- Ask each player to explain resources, placement, and an alternative action they considered. A rules author testing alone cannot establish learnability.
- Make one focused revision and play another six games. Compare automatic supply within B if banking causes repeated confusion; compare the simpler A structure if site play remains too demanding or passive.
- Grow the chosen core's card-role matrix only after its basic rules and pacing work. Expand where a deck lacks a necessary role or a new card opens a distinct decision.
Use small batches pragmatically. Ten exploratory matches can expose slow turns or a glaring imbalance; they cannot establish a balanced metagame. Match time, turn count, first-player outcomes, resource failures, repeated mandatory lines, unused cards, and archetype matchups all deserve a simple log. Small samples should retain raw counts rather than confident balance percentages.
The Star Wars Unlimited team describes tuning base health, unit stats, and removal thresholds together to reach its own game-length goals. That is directly useful methodology: our health total is not an independent speed knob. A threat's cost, the turn it can act, the price of an answer, and the number of opportunities to use it jointly determine pacing. Their published numbers are specific to their system and should not become ours by default. Behind Unlimited on development standards
The next step is hands-on testing of B with its small pilot. Art direction and sample frames can advance in parallel through the theme comparison, provided sample rules remain marked as provisional. No timing, learning, strategic-diversity, or balance target in this document has yet been validated by playtests.