Word Cube 3: How This Guide Was Derived

Why this page exists

This page documents where the guide’s rules came from, how they were tested, and which details remain uncertain. Word Cube 3 is a Scrabble-style word game on Cash Out Club, a different platform from the Skillz games covered elsewhere on this site. Cash Out Club is operated by Tether Studios, the studio behind Solitaire Cube and other Skillz titles.

You can use the results below to check the research before or after buying the guide.

---

The scoring model

I built a scoring model for Word Cube 3 and fit it against real recorded games, reading each game’s on-screen score and checking that the model reproduced the score change for the exact tiles placed. The model is standard Scrabble scoring plus a per-turn length bonus:

play score =
  for the main word and each cross-word formed this turn:
    ( sum of tile values, with letter multipliers on newly placed tiles )
    × any word multiplier on newly placed tiles
  summed across all words
+ length bonus by tiles placed

game objective (points per second):
  total banked = sum of play scores
              + (seconds remaining when you finish) × 1

What the conformance test confirmed. Run over 169 clean single-turn plays across recorded games, the model reproduced the on-screen score change for the tiles placed. That is what pinned:

• The length-bonus staircase: 5 tiles placed = +10, 6 = +25, 7 = +50, and 1–4 tiles = no bonus. The seven-tile rung is the all-rack “bingo.”
• The digram tiles: two letters in one square, worth a flat 5, except QU, which is 7.
• The special tile: a marked single letter worth 7.
• The high-value letters: the single Q and Z at 10, X and J at 8, confirmed against the printed tile values.

Other confirmed mechanics:

• The five-minute clock starts on the first move, not at load, and there is no turn limit.
• Unspent time banks at one point per second, and the game can be ended voluntarily at any time to claim it. Both were confirmed in live play: ending with time remaining paid out at exactly one point per second. This is what makes the real objective points per second, and it yields a clean endgame rule the guide teaches: play another word only while it out-scores the seconds it takes.
• Premium squares (triple/double word and letter) count only under tiles placed that turn, exactly as in Scrabble. The premium layout itself is fixed from game to game; the guide prints the full map.

---

The move-finder

On top of the scoring model I built an exhaustive legal-move generator: for a given board and rack, it produces every legal play, including cross-words, digram tiles, and blank/wildcard placements, and scores each one with the model above. That let me do two things a person cannot do live:

1. Find the single strongest available play in a real position, and
2. Compare it against what was actually played.

The computer’s advantage is completeness: it never misses a legal word or mis-scores a placement. The guide translates the useful parts of that search into human habits: scan premium squares first, pursue the length bonus, place expensive tiles on multipliers, and keep tempo.

---

Key findings

• The length bonus has a steep staircase. Reaching five, six, and seven tiles can outweigh a few points of face value on a shorter word. An all-seven bingo adds 50 points.
• Premium squares often separate good plays from the best play. In the reviewed positions, double-digit scoring gaps repeatedly came from a missed premium square or longer word.
• Tempo affects both turn count and banked time. With no turn limit and a point for every unused second, finding good words quickly can beat finding slightly better words slowly. The guide therefore includes off-board speed drills.
• The finish is a decision. Remaining seconds pay a point each and the game can be ended at any time, so the endgame has a precise stopping rule rather than a vibe. The guide teaches it.
• The rack is renewable and swaps are cheap (three per game, any number of tiles, no penalty), so dead racks should be swapped rather than ground out.

---

Limitations

Word Cube 3 has several limitations that do not apply to the Solitaire Cube and 21 Blitz research:

• There is no full-game win-rate benchmark. Word Cube 3 does not expose its tile distribution, so I could not run a faithful full-game self-play sweep the way a same-deal card game allows. The strategy is derived from a measured scoring model plus optimal-play analysis on real positions, not from thousands of simulated games. The scoring and mechanics are measured with high confidence; the strategy follows from the observed sources of points rather than a statistical head-to-head.
• A few individual tile values are less certain than the headline ones. The high-value tiles (Q, Z at 10; X, J at 8), the digrams, the special tile, and the blank are confirmed; a couple of mid-value letters are inferred and could differ by app version.
• Whether a blank counts toward the length-bonus tile count is unconfirmed. The staircase itself is solid; the blank interaction had no clean test case in the data.
• The one-point-per-second bank is verified on real finishes near the end of a game. Ending with several minutes still remaining is untested, which is why the guide frames the finish rule as an endgame decision.
• The accepted-word list is the app’s own. It is not identical to any printed Scrabble dictionary; the game accepts some words a Scrabble dictionary omits, and may reject a few it includes. The guide teaches word-finding skill, not a promise that a specific word will be accepted.

---

The paid guide contains the tile-value table, full premium-square map, Score-Density Scan routine, endgame finish rule, worked examples, and vocabulary drills. Its printable word-list appendix uses the same research lexicon: all 101 two-letter words, a parallel-play card, hook lists, and rescue lists for expensive tiles, each priced in Word Cube 3 values rather than Scrabble values.