24H
TRANSPARENT BY DESIGN

What goes into a Scout score?

The score has two parts. Measured data (75%) combines engagement (50%), earning sustainability (30%) and momentum (20%). An AI assessment (25%) rates each game on a published rubric. The weights change only after owner review.

One formula, out of 100

Overall = 0.75 × data score + 0.25 × AI score.

Ads never affect scores or rankings. Advertising buys a labelled placement and nothing else.

The AI assessment

Our team writes each assessment with an AI model when we list or re-research a game. It works from our measurements, independent player verdicts, the game's declarations and its own checks of the chain and official sources. It scores six categories from 0 to 100: economic sustainability, gameplay, real users, token and holder structure, transparency, and risk (higher is safer). The AI score is their average. Each assessment shows its rationale, evidence, who wrote it, the model, date and confidence, and is published with the listing. Game text and web pages are treated as untrusted data, so a project cannot instruct the model. A game without an assessment scores zero for this part, like any missing input. Submissions are researched by our team, and the owner decides what is listed.

Activity that means something

Engagement considers meaningful active wallets, returning wallets, meaningful actions and approved player reviews. Players give a thumbs up or down with a written experience; the game page shows the share of thumbs up as a rating out of 5. Reviews are a small part of the score: 15% of engagement, about 5% of the overall score, damped until a game has 10 reviews; a game with no approved review yet has this input left out, not scored as zero. Wallets are not verified human players. Each game’s reviewed integration defines the measured action. Custody participation proxies are labeled separately and receive the existing 25% estimate credit; they do not establish movement sessions or human players.

Numbers a game reports about itself (players online, total players, players in the last 24 hours, a reward pool) are shown on its page as “reported by the game”, beside our on-chain numbers and never merged with them. They are display-only, with one exception: when no on-chain collector measures a game’s active players at all, the game’s reported players in the last 24 hours stand in for active wallets at 25% credit. Where an on-chain collector exists, its value is always the one scored, even while it is stale or under review. Counts from public room rosters are labelled partial coverage: visible online sessions only, never total players or verified unique people. Online counts also appear in the directory. Where wallet activity is not measured, the player tile and chart show reported online players; each chart day is the latest accepted sample for that UTC day, not a daily total or unique-player count. Missing days stay blank.

Holder base (2.5% of engagement, for games with a Solana token): GMGN’s holder count on a log scale, from 0 at 100 holders or fewer to 100 at 50,000 or more.

An economy we can inspect

Sustainability considers operating revenue relative to paid rewards (rewards other players pay directly, such as marketplace sales, count as funded), funded reward runway, payout continuity, treasury strength (the declared treasury's SOL, stablecoins and own tokens, with its own token at half value) and modest credit for buybacks traced to operating revenue. Token trading volume is not game revenue; internal treasury transfers are not new revenue. A documented fee share that stays inside a game's own wallet, with no separate transfer to trace, counts as revenue at 25% credit. Burns and emissions are disclosed where supported.

Capped until measured. Reward funding and funded runway are scored only when every earning route paid from a reserve, new tokens or an unknown source is completely measured by a reviewed collector. While one is not (often because the game keeps that route off-chain), both inputs are left out and the other sustainability inputs carry the weight, rather than reading as 0 or a full 100. Keeping reward funding internal is rated once, in the AI assessment’s transparency and risk categories. A reward reserve held in the game's own token counts at half value, as in treasury strength. Burns that are a share of a player's own payment (a shop item or a marketplace sale) are spending, not buybacks, and earn no buyback credit. A treasury whose own-token holding has no price is recorded as incomplete, never as complete without it.

Holder spread (2.5% of sustainability, for games with a Solana token): the share of supply held by top-10 holders outside the listing’s game wallets and GMGN-labelled exchanges, from GMGN’s sampled holder list. 20% or less scores 100 and 80% or more scores 0.

Momentum without tiny-sample hype

Six-hour and daily gameplay growth, reward-token trading growth, and qualified organic site interest are damped when samples are small. Inputs no collector measures yet, such as organic interest or trading growth for a token without a tracked pool, are shown as not collected and left out of the weights. Tokenless games have an alternate allocation. Social does not affect the launch score.

Organic trading (3.75% of momentum, for games with a Solana token): 100 × (1 − the shares of top-trader volume that GMGN tags as bundler, insider or entrapment trading).

GMGN estimates: a small fixed weight

Holder base, Holder spread and Organic trading are supplementary third-party estimates from GMGN. Each carries a small fixed weight and counts at its actual value, and the other inputs in its category share the rest of the weight in their usual proportions. Good GMGN data raises a score and poor GMGN data lowers it. A GMGN input that is missing, failed or more than 48 hours old changes nothing: it is left out of the weights, so the score is the same as without it. GMGN never affects coverage or whether a score is provisional.

Where does the data come from?

Several collectors for one game: when a game’s economy needs two reviewed shapes (for example token withdrawals and SOL fees into one treasury), each collector owns a declared set of measurements. Each measurement comes only from its owner, so one collector never overwrites another, and each has its own freshness and review state on the game page.

Totals from separate routes: a few dollar totals (revenue, rewards paid, player-funded rewards, burns) can arrive through two separate routes, such as marketplace fees paid in USDC to one account and in the game’s token to another. Each route’s collector measures its own accounts, never the same one twice, and the game page shows the sum with each part in the score breakdown. The total is as fresh as its oldest part, and its evidence credit is weighted by value (a verified $90 plus a reported $10 earns 92.5%). When one part is missing or its collector is held for review, the total is marked partial and left out of the score rather than counted as the whole.

What does each data status mean?

Awaiting research
Listed, but our team has not verified the facts yet.
Facts verified
Our research verified the listing; on-chain collection has not started yet.
Partly measured
Some measurements are verified on-chain, but less than 60% of the data weight is measurable or part of it is not yet verified.
Verified on-chain
Everything we can measure for this game is measured: at least 60% of the data weight is measurable, and at least 80% of that is complete, independently verified chain data.
Reported / estimated
Most available numbers are project reports or indirect participation estimates and count at 25%.
Data stale
Collection missed its schedule; the last complete values are shown and ranking credit fades.

Our team researches every listing, whether a player submitted it or we found it ourselves, and moves it up this ladder as measurements come online.

Unknown stays unknown, and is not a penalty

Some data can never be measured for a game: a route it keeps off-chain, a metric nobody collects yet, or reviews it has not received. A game is scored on everything we can measure for it, to the fullest, so these inputs are left out and the measured inputs in the same category share their weight. They are listed as missing in the score breakdown, never guessed. Reproducible, independently verified measurements count fully; developer reports and indirect estimates (such as custody participation) receive 25% credit. Thin evidence guard: when less than 60% of the data weight can be measured for a game (not counting inputs no game can have yet: organic interest, which nothing collects, and reviews before the first one is approved), the reallocation applies only in proportion (at 30% measurable, half of it), so a game cannot rank high on one or two strong numbers. A score is provisional when less than 60% of its data weight is measurable or less than 80% of what is measured is complete, independently verified chain data.

How fresh is the data?

Core measurements refresh every 2 hours and are marked stale after 12 hours without a complete update. Trading-volume candles update daily and are marked stale after 36 hours without a new candle; since a token with no trades has no new candle, that marks only the volume collector as overdue, not the whole game. Ranking credit stays intact for 48 hours after the last complete observation, then fades linearly to zero at 120 hours. Historical gaps are visible; they are never filled with invented activity.

Common questions

What is a play-to-earn game?

A play-to-earn (P2E) game lets players earn something of real value that can leave the game, such as tokens, stablecoins or NFTs, by playing it. P2E Scout lists only games where that is possible. Betting and casino apps, and games whose currency can never be cashed out, are not listed.

How is the Scout score calculated?

Overall = 0.75 × data score + 0.25 × AI assessment score, out of 100. The data score combines engagement (50%), earning sustainability (30%) and momentum (20%), measured on chain wherever possible. The AI assessment rates six categories on a published rubric. Ads never affect scores or rankings.

What does "active wallets" mean?

Active wallets are distinct wallets that made a meaningful in-game action on chain in the last 24 hours, as defined by each game’s reviewed collector. One person can use several wallets, so active wallets are not a count of human players.

Are play-to-earn games free to play?

Some are. A game page says Free only when nothing has to be bought to play or to earn. If earning needs an NFT, a token holding or another purchase, it says Paid and names the assets needed, without dollar amounts.

Why is some data marked unknown?

When a game keeps an earning route off-chain, or nothing collects a metric yet, P2E Scout shows it as unknown instead of guessing. Unknown inputs are left out of the score weights rather than counted as zero, and a score with less than 60% of its data weight measurable is marked provisional.

How often is the data updated?

Core on-chain measurements refresh every 2 hours and are marked stale after 12 hours without a complete update. Trading-volume candles update daily. Ranking credit stays intact for 48 hours after the last complete observation, then fades to zero at 120 hours.

How do I get a game listed?

Submit its website on the Submit page. Our team researches the gameplay, earning routes and on-chain evidence, and the owner decides what is listed. Accepted discoveries earn Scout points, with a bonus for finding a game within 30 days of its launch.

This method and every game’s latest measurements are also published as plain text for AI assistants: /llms.txt (summary) and /llms-full.txt (full reference).

Formula version: scout-v3.6. Updated . Scores describe evidence and activity, and cannot promise player profit.