DEV Community: khalid ali

Baloot rules for developers: contracts, projects, doubling and qaid scoring

khalid ali — Fri, 11 Sep 2026 07:45:34 +0000

Baloot is Saudi Arabia's most popular card game, and most English-language explanations get the scoring wrong. This is a compact, implementation-oriented reference we used while building Arb3a Baloot (Unity client, Node.js server, RL-trained opponent).

Deck and seating

32 cards (7 through Ace in four suits), four players in two fixed partnerships, eight cards each. The dealer rotates counter-clockwise. Bidding starts with the player to the dealer's right.

Contracts

Sun: no trump. Card order per suit: A, 10, K, Q, J, 9, 8, 7. Point values: A 11, 10 10, K 4, Q 3, J 2, others 0. Round total is 130 plus 10 for the last trick.
Hokm: one trump suit. In the trump suit the order becomes J, 9, A, 10, K, Q, 8, 7 with J worth 20 and 9 worth 14. Other suits keep the Sun order. Round total is 162 including the last-trick bonus.

Bidding

Two rounds. In the first round, players may take the face-up card as trump (Hokm), call Sun, or pass. In the second round a player may name a different trump suit or Sun. If everyone passes twice, the hand is redealt. The side that buys must score more than half the points or it loses everything to the opponents.

Projects (melds)

Declared during the first trick, shown after the second. Sira (3 consecutive cards, one suit), Fifty (4), Hundred (5, or four of a kind), Four Hundred (four Aces, Sun only), Baloot (K and Q of trumps). Only the side with the strongest project scores its projects.

Doubling

Double, Triple, Quadruple and Qahwa multiply the round result. Qahwa ends the match on that hand.

Scoring (qaid)

Points are converted with rounding: in Hokm divide by 10 and round (with special rules at .5), in Sun divide by 5 then round. A team wins the match at 152 or more, provided it is ahead. Kaboot (winning all eight tricks) is worth 44 in Sun and 25 in Hokm.

Try it

Android: https://play.google.com/store/apps/details?id=com.e_moh.balootgame
iOS: https://apps.apple.com/app/id6759184291
Full rules and glossary (Arabic): https://baloot.kriyadh.com/rules

Training a reinforcement-learning agent to play Baloot (and shipping it to Android)

khalid ali — Fri, 11 Sep 2026 02:04:49 +0000

Baloot is a 32-card trick-taking game played in Saudi Arabia by four players in two teams: two contracts (Sun and Hokm), two bidding rounds, declarable melds called projects, doubling up to four times, and a scoring system (qaid) with rounding rules that catch even experienced players out. Building a bot that plays it well turned out to be a much bigger project than the card game itself. This is a write-up of the approach used in Arb3a Baloot (Unity client, Node.js server), which shipped on Android this week.

Why the rule-based bot failed

The first opponent was a hand-written heuristic: count trump strength, bid Hokm with jack and nine, lead high in Sun. It played legal Baloot but lost to humans, mostly because bidding depends on your partner's likely hand and the dealer position, not only on your own cards. Every hand-coded fix created a new exploitable habit.

State encoding

The encoder produces a 571-dimensional vector per decision:

the 32-card hand as a bitmap
trick history as ordered plays with seat indices
suits each opponent is known to be void in (inferred from failed follows)
confirmed projects
bidding context: who bid, what, in which round, and the doubling level
the running match score, because the right play at 140-60 is not the right play at 60-140

Training loop

PPO with self-play on a simulator that reuses the server's engine code, so the bot trains on exactly the rules it faces in production. Bidding and card play are two policy heads on a shared trunk. Rewards are shaped per round from the qaid delta rather than only at match end, which roughly halved wall-clock time to a competitive policy. A distillation stage compresses the best checkpoint into a small network that runs on-device for offline play.

Evaluation

A candidate is promoted only when it beats the current champion over a fixed set of 960 dealt games with seats rotated, so both sides see the same cards. A search-based player (lookahead over sampled hidden hands) beats the champion network by a few points but is too slow to run for every seat on the server, so it stays a benchmark.

Things that bit us

The model loader logged "model loaded" before verifying the file, so one deploy ran a random policy in production for days. It now prints the checksum and the evaluation win-rate at startup.
Turn timers lived only in memory, so a server restart during a doubled hand stranded the table. Timer flags are now persisted with the game state.

Shipping

The Android build is a Unity IL2CPP app bundle (arm64 + armeabi-v7a). The game is free with no ads.

Google Play: https://play.google.com/store/apps/details?id=com.e_moh.balootgame
App Store: https://apps.apple.com/app/id6759184291
Rules and terminology (Arabic): https://baloot.kriyadh.com/rules

Happy to answer questions about the encoder or the training setup.

Baloot Goes Digital: Building Saudi Arabia's Card Game with a Real RL-Trained AI

khalid ali — Tue, 18 Aug 2026 12:33:07 +0000

Baloot is Saudi Arabia's national card game — 32 cards, 4 players, 2 teams, and a bidding system (Sun / Hokm) that rewards reading your opponents.

Most Baloot apps are either stuffed with ads or built around buying coins. Arb'a Baloot takes the opposite approach: completely free, no ads, no in-app purchases.

The interesting part: a real AI

Instead of hard-coded rules, the AI opponent was trained with reinforcement learning over millions of self-play rounds — the same family of techniques behind AlphaGo. It bids Sun or Hokm like a strong human, counts cards, punishes mistakes, and works fully offline.

Features

Online multiplayer with private invite-code rooms
Full rule set: projects (Sira, Fifty, Hundred, Four Hundred, Baloot) and doubling up to Qahwa
Automatic score counting with correct rounding
Weekly tournaments with a public leaderboard
Rated 5.0 on the App Store · iPhone, iPad, Mac & Vision Pro