Spendii

Room setup

The host selects a target score from 6 to 21, a game clock, a deck, position assignment, anonymity, the total number of players (default 4), and whether the final position is NO PLAY. Choose 2-5 participants normally, or 3-6 with NO PLAY, with at most five humans and four bots. There must be 2-5 active playing positions.

The Deck dropdown defaults to Standard. In the testing environment, it also offers Short test deck for shorter test games.

• Random positions: positions are shuffled and all active players begin at 0 points.
• Betting positions: each participant bids points for every position before choosing a starting position.
• NO PLAY: the participant assigned this position watches the game without taking turns.
• Move advice: when turned on in Advanced settings, active human players can privately ask what the bot strategy would do. During a player's live turn, Move advice shows only the current stage and recalculates after each submitted stage because a main action can reveal a held or replacement market card that may affect later excess-coin returns or bonus-tile choices. Predicting with face-down server information would leak it, while ignoring that information could make the advice inaccurate. During another player's turn or gameplay playback, Move advice can instead show the requesting player's latest relevant completed turn as an ordered list reconstructed from the authoritative stage snapshots. The same private history remains available when viewing a completed game. Viewers and the participant assigned NO PLAY receive no preview.

Waiting room

Create Room opens with only its creator, ready by default. The waiting host can select Add bot in the host-only actions row below the player list, up to four bots within room capacity. Bots show Automatic bot; Remove reopens their slots. The host selects Start Game once every slot is filled and all humans are ready. Adding a bot never starts the game. Spendii uses online play and has no Play mode setting or same-device players. Recreating a completed game follows the same waiting-room flow.

Viewing and joining

Every room is public to view. For a registered online room that can accept another human, the host chooses Joining: Open joins eligible players immediately, Host approval lets the host Accept or Decline each request, and 4-digit passcode requires the exact code chosen at creation. A player may cancel a pending request, and stays in any existing waiting room unless the request is accepted. Only the host can see the passcode. Bot additions and removals preserve the chosen joining policy. Guest rooms are public to view but never accept remote players, including while their slots are empty; a guest adds 1-4 bots before starting.

Betting for positions

Each bot submits its position bids after a random 450–550 ms delay from Start Game. Bots receive distinct random submission times so tied bot bids do not favor room entry order. The server retains those scheduled times even if processing is delayed.

With Position assignment set to Bidding, every bid must be an integer between the negative and positive target score, and all of a participant's bids must total 0. Lower bids receive better rank. Tied human and bot bids favor the human. Other ties favor the earlier server-recorded submission, then the lower-ELO participant when submission times are identical, then the participant who entered the room earlier when ELO is also tied. One submission time applies to all of a participant's position bids. The server measures it from when the host clicks Start game. During bidding, each visible participant who has submitted shows Submitted in Ns using whole elapsed seconds, and that relative time remains in the participant's player box during starting-position selection rather than appearing in the ranking-result cells. Hidden participant cards keep the neutral betting status, and Hide identities set to Players and bots continues to hide submission times during starting-position selection. Beginning with first rank, participants privately select an available position from one of their highlighted ranked cells. Selections are revealed together after everyone required for that rank has chosen.

Board setup

• Active players act in position order: A, B, C, D, then E.
• Four development cards from each of the three tiers are face up.
• The board has one more bonus tile than active players.
• The bank begins with 2-9 regular chips of each color and 0-6 gold wild chips, as selected in the room settings. Regular chips default to 7 and use player-count presets of 4, 5, 7, and 8 for 2, 3, 4, and 5 active players; gold defaults to 5.
• A betting-mode player's starting points equal the bid for their selected position.

Your turn

Choose exactly one main action:

1. Take chips: take 1 chip; 2 or 3 different available colors; or 2 matching chips when at least 4 of that color are in the bank.
2. Hold a card: hold a face-up card or the top hidden card from a tier. You may hold at most 3 cards and receive a gold chip when one is available.
3. Buy a card: pay for a face-up or personally held card. Permanent discounts reduce colored costs and gold covers remaining colored costs. The room's Card payment method defaults to Use least gold by spending matching colored coins first. If the room uses Choose manually and more than one payment is possible, choose the number of each held coin to spend before confirming the purchase.
4. Pass: take no main action. Any required follow-up still has to be completed.

Turn limits and bonus tiles

• You may hold at most 10 chips after your turn. If you exceed the limit, return exactly the excess.
• After the main action and any chip return, the game checks bonus-tile eligibility using permanent card discounts.
• If one or more tiles are available, choose exactly one before the turn ends.
• The game clock continues during required chip returns and bonus-tile choices. The increment is added only when the entire turn finishes.

Resigning and leaving

During betting, position selection, and gameplay, a non-forfeited participant sees one Resign departure control. Its confirmation offers Resign to forfeit and remain in the room or Resign & Exit to apply that same forfeit and return immediately to the lobby. Both keep the participant in the roster and score-based ELO result and make remaining decisions automatic; Exit occurs only after the forfeit succeeds. When the player remains, the same navbar control changes to navigation-only Exit. A direct active Exit request has the same forfeit result. Top-left navigation, refresh, and closing the page change presence only and preserve the reconnect grace period.

Game end and ELO calculation

• Reaching the target score starts the final round. Play continues until every active player has received the same number of turns.
• The game ends immediately if every active player cleanly passes in the same round. A pass is clean only when it needs no chip return or bonus-tile choice.
• Otherwise, the game ends when the same coin distribution occurs three times at round boundaries. Compare each active player's regular and gold coins and the bank's coins. Count the initial position and completed rounds; occurrences need not be consecutive. Finish all mandatory returns and bonus-tile choices before comparing. Clocks and round numbers do not affect the comparison.
• Buying or holding a card, claiming a bonus tile, a timeout, or a forfeiture resets the repetition count. Counting resumes at the next completed round. There is no round limit. Repetition uses the existing scores and normal ranking and ELO calculations.
• Spendii ELO uses point totals and participant ratings; purchased development cards are not included. Let pMax be the highest point total among all joined humans and bots. For each entity i, diff_i = pMax - points_i and playerScore_i = diff_i / sum(diff_j), where the sum includes every human and bot. If all entity point totals are equal, every playerScore is 1 / entity count. The expected loss-like score is expectedScore_i = 1 / sum(10 ^ ((effectiveELO_i - effectiveELO_j) / 400)), also summed over every entity. Humans use their current rating and every bot uses the fixed default ELO of 1500.
• Every entity's hypothetical raw change is -40 * (playerScore - expectedScore). The combined raw change of all bots is divided equally among the humans and added to their target ratings. Floor-aware reconciliation keeps every human at or above ELO 1 while preserving the total human ELO. Bots remain at 1500. The same calculation applies with only one human: scores are still calculated, and redistribution naturally leaves that human's ELO unchanged.

Guests, bots, and ELO

Guests can create bot rooms and spectate public rooms, but their rooms cannot be joined remotely and their games are unrated. Bots play automatically, participate in rated calculations at the fixed default ELO, and do not receive a persisted ELO change. Rated games update human participants, including a human assigned NO PLAY, when the game finishes. In the final table, rated registered-human ELO changes and new ratings use exactly one decimal place. Increases are green, decreases are red, and unchanged changes are white. Guest, bot, and other unrated rows show dashes in both ELO columns.