How rankings and verification work

Game editorial research score

Automated game pages use mmotop-research-v1, a provisional AI-assisted research assessment. It combines documented MMO systems and Steam player reception. It is not a hands-on review, a performance benchmark or a guarantee of quality. Publisher claims may be incomplete or inaccurate. Games with different design goals may not be directly comparable.

An AI proposes feature subscores with exact official-source quotations and explanations. A separate AI audit checks the evidence, scoring anchors and whether features are available rather than planned. The application validates the quotations and computes the final score; the model cannot choose the final rating or grade.

• Persistent world — 30%.
• Progression — 25%.
• Player interaction — 25%.
• Crafting and economy — 20%.

Each criterion scores 0–4 using the anchors below. Missing evidence is unknown and excluded, never scored zero. A zero needs explicit evidence that the system is absent. At least three criteria covering 70% of the total weight are needed. Coverage is shown separately.

Feature score D = 25 × Σ(weight × subscore) / Σ(known weights). Greater documented system depth earns more points; this does not measure enjoyment, balance or execution.

Player reception P = 100 × (positive reviews + 25) / (total reviews + 50). The 50-review neutral prior prevents a few reviews from producing extreme ratings. Steam totals cover all languages and purchase types, with Steam off-topic filtering enabled. Review text and user identities are not stored. Totals are requested at most once per game per 24 hours during analysis, including failed attempts. The stored score is a dated snapshot, not a live Steam rating.

Editors rating = round((D + P) / 2, 1), on a 0–100 scale. For example, D=75 and 80 positive reviews out of 100 give P=70 and a final score of 72.5, class B. Classification uses the rounded rating: S ≥90, A ≥80, B ≥70, C ≥60, D <60. Editorial class is this grade, not a claim of AAA/AA production budget.

An unreleased game, insufficient coverage, unavailable review totals or no reviews produces Pending and no numeric rating. All automatic assessments remain provisional; a small review sample or partial feature coverage adds uncertainty. Data confidence is separate from the rating and stays medium in this source-based workflow. Human ratings are preserved when editing existing pages.

Feature scoring anchors

1. Persistent world: 0 explicit session-only play without persistence; 1 persistent character/inventory; 2 persistent shared online world; 3 persistent player changes to settlements, territory or economy; 4 lasting world changes with explicit persistence/reset rules.
2. Progression: 0 explicitly no progression; 1 one advancement track; 2 multiple tracks; 3 builds/specializations with explicit trade-offs; 4 those choices plus alternative long-term or endgame goals.
3. Player interaction: 0 explicitly none; 1 chat/parallel presence; 2 one cooperative, competitive or trading activity; 3 multiple distinct systems; 4 explicit coordinated roles or player/group dependencies.
4. Crafting and economy: 0 explicitly neither; 1 NPC purchases/loot; 2 crafting or player trade; 3 connected gathering, specialization, crafting and trade; 4 player-driven markets with explicit resource sources and currency/item sinks.

Choose the highest fully evidenced anchor. A market alone does not establish a complex economy, unlimited levels do not establish build choice, and a guild label alone does not establish coordinated roles. We retain the evidence, calculation, model, methodology version and assessment date. Steam review API documentation.