Wordle Letter Frequency by Position: Analysis of 2,309 Answers

Overall frequency hides useful detail. E can be common across five-letter words while another letter is more common in the opening slot. A starting guess tests both letter presence and position, so the order of the same five letters can change the feedback groups it creates.

The original analysis below counts every letter in every exact position across the site's fixed 2,309-answer model. It is generated from the same local dataset used for daily difficulty measurements, which keeps the method repeatable and prevents the article from drifting away from the tool.

How the counts were calculated

For each answer, the first letter adds one observation to position 1, the second to position 2, and so on. The percentage is the count divided by 2,309. Repeated letters count in each position where they occur; they do not receive extra weight beyond those actual positions.

This is descriptive analysis, not a claim about a private current New York Times answer schedule. The publisher can edit its active answer pool, and past lists do not reveal which word will appear next.

How to use position data

When comparing two openers with similar distinct-letter coverage, position frequency explains why their information scores can differ. The same letter may create more green outcomes in one column and more yellow outcomes in another. Use the table together with the entropy-ranked starting words, which scores complete feedback partitions instead of simply adding frequencies.

Later in a solve, position data helps order candidates but should never override hard feedback. A less common positional letter is still correct when all previous rows require it. Frequency is most valuable for deciding which uncertainty to test, not for rejecting a compatible answer.

Important limitations

The table describes one disclosed model, not the entire English language. Accepted guesses include many words that are not in the answer model, so their letter distribution is different. The analysis also treats every model answer equally and does not attempt to predict editorial selection.

For related evidence, the strategy guide includes computed one-letter trap families, while the rules guide measures repeated-letter answers and explains duplicate feedback.