How the rating works
A program score is not a plain average. One review should not decide the fate of a program, recent experience matters more than three-year-old experience, and the voice of someone the site knows counts for more than a day-old account. Here is exactly how it works.
How the score is built
The score of a single review
The author rates five criteria. They are combined with different weights: payouts and reliability 25% each, profitability 20%, support and promo materials 15% each. The result is one number from 1 to 5 — the score of that review.
The weight of a review
Reviews carry different weight. A refchat member and the SDT team count three times as much as a confirmed member — there is verified work with the site behind them. The role weight is fixed the moment the review is published and never rewritten afterwards: leaving the refchat later does not rewrite the past.
Decay over time
Programs change their terms, so an old review weighs less than a fresh one. After eighteen months a review is worth half of what it was, and it keeps fading — but never to zero: the floor is 30% of the original weight. A three-year-old review is still heard, only quieter.
Smoothing
Five "virtual" votes with the site-wide average are added to every program. That is why a program with a single enthusiastic review does not outrank one with fifty good ones: the more real reviews there are, the weaker the virtual ones become. This is Bayesian smoothing — the same technique large catalogues use.
What does not count
- Reviews awaiting moderation, rejected and hidden ones.
- Reviews their authors deleted.
- Reviews by program representatives — they have an interest in the outcome and no vote.
- Reviews by banned members, including the ones written earlier.
- Reviews from deleted accounts, on the contrary, stay: otherwise a rating could be brought down by deleting accounts in bulk.
Why a program may have no score
The number appears once three reviews count towards it. Until then the card says "Not enough reviews": one or two scores say nothing about a program, but would read as a verdict or as an advert. Such programs still take part in sorting — they simply have no public number yet.
Breakdown by group
Below the overall score the card shows three separate averages: from the SDT team, from refchat members and from everyone else. These are reference numbers — plain averages with the same decay over time, but without role weights and without smoothing. A group with fewer than three reviews is shown as a dash. The breakdown takes no part in the overall rating or in sorting.
Scale and recalculation
Scores are given in stars from one to five and shown on a 0 to 100 scale — easier to compare. A rating is recalculated as soon as something changes: a review appears, a moderator makes a decision, an author is banned. Once a day the whole site is recalculated — decay requires it: even with no new reviews, the weight of the old ones changes every day.
The SDT score is separate
Next to the community score the card shows the editorial one. It is set by hand and takes no part in the rating at all: these are two different opinions, and merging them into one number would be unfair to both.