The PlayerSells X algorithm scorecard combines five signals into one 0-100 number, and it weights them bookmark power 30%, conversation spark 20%, virality 20%, reach efficiency 15% and account authority 15%. Those weights are this model's assumptions about what X rewards, taken from the ranking code X published, and not a measurement of X's live ranking system. They are printed here so a score can be reproduced instead of taken on trust.
Every engagement threshold in the model moves with audience size, because engagement rate falls as an audience grows. An account with 5,000 followers is graded excellent at a 1% bookmark-to-view rate and poor at 0.05%; an account with 1,000,000 followers is graded excellent at 0.5% and poor at 0.03%, a bar 2 times lower. Which band an account lands in matters more than it looks: across the 45,308-account sample PlayerSells measured from its 18,486,455-account X index on 24 August 2026, 10,000 followers is larger than 92.5% of them, so all but about 7.5% of indexed accounts are graded on the smallest-account band.
The median account in that same index has 550 followers, the 90th percentile has 7,085 and the 99th has 89,454, measured 24 August 2026. 1,000,000 followers is larger than 99.1% of it. The scorecard reads an account's most recent original posts, drops retweets, prefers the posts that carry a public view count, and refuses to score fewer than 3 of them rather than grading an account on one lucky post.
Follower percentiles are measured over a 45,308-row sample of the 18,486,455 X accounts PlayerSells indexes, on 24 August 2026. They describe the indexed population, not every account on X. The five weights and every benchmark rate are the model's own parameters and are not published by X.