Skip to main content

Scoring engine

The scoring engine is what makes the MBER feed different from scrolling X directly. It ranks every post by genuine signal, so the strongest content rises to the top no matter how big or small the account is. It is the difference between a feed sorted by who is loudest and one sorted by what is actually worth your attention.

Why raw numbers mislead​

If you sort by likes, big accounts always win. A post with five thousand likes from a huge account will outrank a post with eighty likes from a small one, even when the smaller post earned far more attention relative to its reach. Raw engagement measures audience size as much as it measures quality, so a feed sorted that way is really a follower-count leaderboard in disguise.

MBER scores for quality instead. A sharp post from a smaller account can, and often does, outrank a viral post from a megaphone.

What MBER rewards​

Without exposing the recipe, the scoring engine weighs a few ideas against each other:

  • Substance. The real engagement a post earned, and not as a flat like count. A repost carries a post further than a like, a reply or a quote means someone stopped to respond, and a bookmark means someone saved it for later. MBER reads those as different strengths of signal, so a post that sparks conversation or gets saved counts for more than one that collects easy likes.
  • Efficiency. How much engagement a post earned relative to how many people saw it. This is the great equalizer: a small account with sharp engagement can outscore a large one coasting on reach. MBER guards it so a fluke does not pass for quality. A couple of replies on a handful of views is not trustworthy yet, so the signal eases in as a post gathers enough views to actually mean something.
  • Reach. How many people a post got in front of. Reach is rewarded, but with diminishing returns, so a huge view count on its own cannot bully its way to the top.

One idea runs through all three: MBER deliberately compresses big numbers. A post with a hundred times the raw engagement of another does not earn a hundred times the score. That is what keeps a sharp small post in real contention with a viral one, instead of letting sheer volume decide everything.

The result is a single ranking that surfaces quality regardless of follower count.

Score pills: ENG, VRL, and QUAL​

Each post also carries three pills, scored from 0 to 10:

  • ENG, overall engagement,
  • VRL, virality or shareability,
  • QUAL, depth of interest.

The important detail: these pills are relative to the posts you are currently looking at, not to the entire database. A 10 means "near the top of this view." When you change filters, the meaning shifts with them, by design. The pill answers "how does this compare to what is in front of me right now."

Because the pill is a rank within that view and not a raw rate, one freak post cannot peg the scale and flatten everyone else into the low end. The same grow-with-confidence guard applies, so a big number on a tiny sample does not leap to the top.

When your current view is too small to compare fairly, MBER shows a neutral mark instead of a misleading number.

info

This page explains the philosophy behind ranking. The exact weights, thresholds, and math are part of what makes MBER work, and stay under the hood.