128ordle Opener Analysis
Six opening words, compared using the 2,377-word standard answer pool. Figures below are calculated from the word list shipped with the game.
| Opener | Feedback patterns | Entropy (bits) | Expected answers left | Largest group |
|---|---|---|---|---|
| RAISE | 132 | 5.882 | 62.5 | 173 |
| AROSE | 122 | 5.770 | 67.8 | 188 |
| STARE | 137 | 5.812 | 73.5 | 234 |
| SLATE | 147 | 5.852 | 74.0 | 229 |
| TRACE | 150 | 5.817 | 77.0 | 255 |
| CRANE | 142 | 5.726 | 82.0 | 273 |
What we measured
We compared six common opening guesses against all 2,377 words in the standard answer pool shipped with this version of 128ordle. Each answer is weighted equally. The results are calculated from that list, not from today’s puzzle or player submissions.
How to read the table
A feedback pattern is a five-tile combination of green, yellow, and grey. More patterns is not necessarily better: the sizes of the groups matter too. Higher entropy means more information on average. Expected answers left is the average number of candidates still consistent with your feedback, including the actual answer. Largest group is the worst-case remainder. Lower is better for the last two measures.
What the comparison shows
RAISE has the highest entropy (5.882 bits). RAISE has the smallest expected remainder (62.5 answers), and RAISE has the smallest worst-case group (173 answers). These results compare only the six listed guesses; they do not identify the best word among all accepted guesses.
Why this does not solve 128 boards for you
This is a single-board, one-guess calculation. It does not simulate a full run, account for a second opener, or measure time spent scanning. On multiple boards a confirmed answer can also gather information elsewhere. Use the table to choose an opening candidate, then use the feedback and your remaining guess budget to decide what comes next.
Calculation method
The analysis is calculated from the game’s standard answer pool during the site build. For every opener and answer pair, it marks exact matches first, then allocates yellow matches only from unused answer letters. It groups identical patterns, computes entropy as −sum(p × log2(p)), expected remainder as sum(group size²) / answer count, and the largest group.
Limits of the sample
The model treats every standard answer as equally likely. It excludes the separate Hard and external answer pools and does not claim to measure real-player win rates. A change to the standard answer list changes the results. A smaller average group can still leave a difficult pattern on one of your boards.