This tool does not tell you when a generic audience is online. It buckets one account's own recent posts into a 168-slot grid - 7 days by 24 hours - and reads the average engagement each slot earned, where engagement is likes plus reposts plus replies on that post. It reads up to three pages of recent posts, drops reposts, and refuses to run on fewer than 5 posts with a usable timestamp, because a heatmap drawn from four posts is a picture of noise.
Two numbers come out of that grid. The timing score starts at a baseline of 50 and moves on four things the account controls: posting between 1 and 5 times a day, how steady the gaps between posts are, how many distinct hours and how many distinct days the account is active in, and whether there were at least 40 posts to read. 85 and above is graded "İdeal"; below 25 it is graded "Zayıf". Consistency is the coefficient of variation of the gaps between posts: under 0.3 the gaps barely vary and the account is called çok düzenli. Those bands are this model's parameters, not X's.
Scale is why the answer has to be personal. Across the 45,308-account sample PlayerSells measured from its 18,486,455-account X index on 24 August 2026, the median account has 550 followers, the 75th percentile has 2,046, the 90th has 7,085 and the 99th has 89,454. Posting advice written for the top of that distribution does not transfer down it: an account at the 99th percentile has over 162 times the audience of the median one, and its best hour is a fact about its own readers rather than about X. That is why this tool reads one account's own posts instead of publishing a single best hour to tweet.
Follower percentiles are measured over a 45,308-row sample of the 18,486,455 X accounts PlayerSells indexes, on 24 August 2026, and describe that index rather than every account on X. The timing bands, the consistency thresholds and the 168-slot grid are this tool's own parameters. X publishes none of them.