Six dimensions.
One standard.
Every operator on Gamsites is scored against the same rubric. Below is the rubric, in public, versioned so we can be honest about how it evolves.
What we score, why we score it.
Trust
License posture, ownership transparency, time in market, public-record incidents, regulatory standing in the markets the operator targets. Operators with hidden ownership or unresolved player-fund disputes lose points fast.
Withdrawals
Reported average withdrawal time vs verified average where we can test, threshold at which the operator triggers source-of-funds review, and any documented pattern of slow-roll or staged payouts. Verified instant beats reported instant beats unverified.
Bonus value
Wagering requirements simulated against game-contribution tables, max-cashout caps, expiry windows, and the expected-value coefficient on the offer in isolation. Headline match percentage means nothing without the math under it.
Library
Game count, provider breadth, presence of provably-fair house games, live-dealer scope, and exclusive titles. Adjusted for jurisdiction, since some operators ship trimmed libraries to specific markets.
Transparency
Provably-fair implementation depth (server-seed, client-seed, nonce visible per game), published RTP per game, public terms of service that match practice, and willingness to publish change-logs when terms shift.
Support
Live-chat response time, escalation path, complaint resolution pattern (sampled from public-forum disputes), and presence of an actual human-staffed support function vs purely scripted bot responses.
How a score becomes a verdict.
Trust scores roll up to one of five tiers. The thresholds are fixed, so a score of 78 reads the same way for every operator.
Off the rubric on purpose.
A pretty UI is not a trust signal. Some of the most credible operators run plain interfaces; some of the worst look great.
A "$10,000 welcome" is a marketing number. The scored bonus is the math after wagering, contribution caps, and expiry.
Sponsored coverage is signal noise. We score the operator, not the streamer who promoted them this month.
"Provably fair", "instant withdrawals", "always paid", all need to be demonstrated. Claims without verification get treated as estimates.
Methodology v1.0
Every methodology change is published openly. When the rubric shifts, the version number bumps, the change is logged, and any affected scores are recalculated.