Skip to content

Awari (Oware)

Mancala · all families · data

Also known as: Oware, Awale, Awele, Wari, Ouri.

Either side can force a draw with optimal play (final score 24-24)

Result Draw
Strength Strong
Method Retrograde analysis
Year 2002
Solved by John W. Romein & Henri E. Bal (Vrije Universiteit Amsterdam)
Source Romein et al. 2002
Board 2×6 pits, 4 seeds per pit (48 seeds total)
State space 889,063,398,406
Date Size or opening Result Strength Credit Source
2002 standard game draw Strong John W. Romein & Henri E. Bal source

One row, derived from the record’s year, solver and primary source; no fuller history is recorded yet.

Write-ups

Strongly solved: Romein & Bal computed the best move and exact eventual outcome for all 889,063,398,406 reachable positions, i.e. a complete tablebase, not just the start-position value. The standard start (two rows of six pits, four seeds each, 48 seeds, capture-on-2-or-3 sowing rules) is a draw — both sides can force the 24-24 split. The computation ran ~51 hours on a 144-processor (72 dual-PIII) cluster with a Myrinet interconnect; the largest endgame database held 204 billion entries (178 GB) and the full on-disk database was ~778 GiB, the largest game database computed at the time. state_space here is the EXACT reachable-position count from the solver (≈8.89×10^11 ≈ 10^12), which is also the canonical state-space figure (log10 ≈ 12) in the Herik/Uiterwijk/Van Rijswijck 2002 “Games solved” table; that same table estimates game-tree complexity at ≈10^32 (log10 32, branching factor ≈3.5, avg game length ≈60 plies) — left null in the structured field because it is an estimate, not an exact count. Foundational retrograde work on Awari was begun by Allis, van der Meulen & van den Herik (1990) and extended by van der Goot (2000) and Lincke (2002); Romein & Bal’s 2002 run was the computation that fully solved the game. They also published an “Awari Oracle” Java applet for play against the perfect database.