Vowl · Strategy

Every opener, scored

Candidate first guesses for all three word lengths, ranked by how much of the dictionary they actually eliminate. Computed against the full answer set, not a sample.

How the scoring works

For a given opener, play it against every possible answer and record the colour pattern it produces. Answers that produce the same pattern are answers you cannot tell apart yet, so they form a group. If the opener produces one group of 400 and forty groups of 5, then most of the time you end up in a small group and occasionally you end up in a big one.

The words left column is the size of the group you land in, averaged over every answer. It is the number you should care about, because it is literally how many candidates you are still holding after guess one. The patterns column is how many distinct colour patterns the word can produce at all, which is a rough measure of how finely it can cut.

Those two mostly agree but not always, and where they disagree the words left column wins. A word can produce a lot of rare patterns and still dump you in a giant bucket most of the time.

Five letters

8,636 possible answers. TARES leaves you with 196 of them, which is a little over 2 percent of the dictionary from one guess.

Four letters

3,903 possible answers, and the spread between openers is much tighter. Short words carry less information per letter and there are fewer of them to separate, so your opener matters less here than anywhere else. SANE edges it at 286.

Note that the S effect is weaker at four letters: only 814 words end in S, 20.9 percent rather than 31.2. Plurals of three letter words are a smaller share of the list.

Six letters

15,232 possible answers, and paradoxically the easiest opening. More letters means more constraint per guess, so ORATES gets you down to 153 candidates out of fifteen thousand. The good six letter openers all produce over 400 distinct patterns, roughly double what the best five letter words manage.

The fifth slot at six letters is E 30.8 percent of the time, driven by the -ERS and -EST endings, and the sixth is S 28.1 percent. Openers that cover both do well.

Caveats worth knowing

• This assumes every answer is equally likely, which in Vowl it is. The server picks uniformly from the list, so there is no hidden weighting toward common words to exploit.
• A lower score is not automatically a better game. These numbers measure guess one in isolation, and a word that leaves 196 candidates in one big cluster is harder to follow up than one that leaves 205 spread out. In practice the top few are close enough that it does not matter.
• The candidate list is hand picked, not exhaustive. Scoring all 8,636 words against all 8,636 answers is 74 million comparisons, which is fine to run offline but not what this page is for. If you have a favourite opener that is not here, the method is written above and it is about fifteen lines of code.

← Play Vowl