Loading the method…
Start at 100. Subtract for every fact that could be used against you. A fatal fact locks the best possible score; unverified checks lock it again. The score maps to a tier. Every number on this page is read from the scoring code itself.
If it is still set, someone can print more supply into your position. A Token-2022 permanent delegate is the same power.
If it is still set, your account can be frozen so you cannot sell. Transfer hooks and frozen-by-default mints count too.
Curve, PumpSwap or a pool. If a wallet can pull the pool, it is a rug waiting to happen.
What the dev wallet still holds, and whether the launch allocation was already sold.
The ten largest wallets combined, with pools and the curve left out.
Supply bought in the creation slot and the two after it. That is where bundles live.
What this creator did on their last 10 launches. Only exits that hurt outside buyers count.
Rendered from PENALTY in src/lib/known.ts.
| Check | Condition | Penalty |
|---|---|---|
| Mint authority | still set | −40 |
| Token-2022 permanent delegate (same power as a live mint authority) | −40 | |
| Freeze authority | still set | −30 |
| Token-2022 transfer hook, non-transferable, or frozen by default | −30 | |
| Token-2022 transfer fee above 0 bps | −10 | |
| Liquidity | LP withdrawable by a wallet | −40 |
| DEX pool, LP status unknown | −10 | |
| Creator holdings | < 2% of supply | 0 |
| 2% of supply – 5% of supply | −15 | |
| 5% of supply – 10% of supply | −25 | |
| ≥ 10% of supply | −35 | |
| bought ≥ 2% at creation, keeps ≤ 20% of it, token younger than 7 days (launch allocation sold) | −25 | |
| same, token older than 7 days (historical) | −10 | |
| Top 10 wallets | < 15% of supply | 0 |
| 15% of supply – 30% of supply | −15 | |
| 30% of supply – 50% of supply | −30 | |
| ≥ 50% of supply | −40 | |
| Launch window | < 10% bought by non-creator wallets | 0 |
| 10% bought by non-creator wallets – 25% bought by non-creator wallets | −20 | |
| ≥ 25% bought by non-creator wallets | −35 | |
| Creator history | < 1 harmful exits in last 10 launches | 0 |
| 1 harmful exits in last 10 launches – 2 harmful exits in last 10 launches | −15 | |
| ≥ 2 harmful exits in last 10 launches | −40 | |
| creator flagged or banned on the wanted board | −40 |
Penalties can be offset by a clean rest of the sheet. A lock cannot. The lowest applicable lock wins, and the verdict names it. Rendered from SCORE_CAPS.
| When | Best possible | Best tier |
|---|---|---|
| Creator is banned: three or more harmful exits in their last 10 launches | ≤ 20 | RUG |
| Creator is flagged: two harmful exits in their last 10 launches | ≤ 45 | JEET |
| Mint authority is live: supply can be inflated at will | ≤ 40 | JEET |
| Freeze authority is live: your tokens can be frozen | ≤ 45 | JEET |
| LP is unlocked: a wallet can pull the pool | ≤ 40 | JEET |
| Creator already sold the launch allocation | ≤ 55 | MID |
| The creator still holds 5% or more of supply | ≤ 49 | JEET |
| The creator still holds 2% or more of supply | ≤ 69 | MID |
| Bundled launch: more than 25% of supply bought in the creation window | ≤ 50 | MID |
| Ten wallets hold more than 50% of supply | ≤ 40 | JEET |
| No market left: a bonding-curve token still under $5,000 more than an hour after launch | ≤ 49 | JEET |
| The token is under an hour old: too early for the top tier, whatever the checks say | ≤ 84 | BASED |
When the RPC cannot answer, the check reports unknown and takes no penalty. It still locks the score, because the top tier has to be earned on evidence.
Five tiers, fixed score ranges. The glyph and the word are what you see everywhere on LPX.
Authorities revoked, liquidity protocol-held, creator clean, no bundle. Still not financial advice.
Clean fundamentals with one thing to watch.
Real risk on the table: concentration, a bundle, or a creator with history.
Expect to be dumped on. Something here is built to extract.
A fatal fact is present: banned creator, live authority, pullable liquidity, or a sold launch bag.
The checks above score the token. The LPX Engine then reads the launch itself and the wallets behind it, and adjusts the same number. One read, one score, one tier, the same on the site, the bots, the badge and the share card.
The creator paid for the first buyers. A small group of linked wallets took a big share of supply at launch.
Early wallets here showed up early on other launches that rugged. Linked wallets took a smaller share at launch.
The early buyers dumped fast. Wallets keep selling into big buys. Exits cost more than they should.
Busy exchange and bot wallets, and anything the chain cannot prove.
Raw 20 after penalties, creator flagged on the wanted board and the launch allocation sold. Two locks apply; the raw score is lower than both, so the number stays 20.
Every scanned token is read again at fixed times after its first scan, and once a week the tiers are checked against what happened. The numbers below come straight from that report. A backtest in October 2026 found the tiers out of order in the middle; the weights were retuned by hand once, on under three days of data, and this is how anyone can check whether that held.
Every scan is stored as a summary: tier, score, stage, creator wallet, holdings, concentration, launch-window share, liquidity, timestamps. That is the scoreboard.
A tracker rescans tracked tokens and compares consecutive scans. Two temporal checks produce hard events: creator dumped (creator holdings fall more than 80% while price falls more than 50%) and liquidity gone (liquidity falls more than 90% after graduation).
Events attach to the creator wallet. That is the wanted board: judged on their last 10 launches only, and only exits where outside buyers were exposed count. One harmful exit puts them on watch, two flags them, three or more bans them. Launching often is never penalized on its own. A flagged creator costs −40 and locks the score at 45; banned locks it at 20. Nothing is entered by hand.
Two checks for tokens that borrow someone else's identity. Both are information on the read: neither moves the score.
A token is a copycat when an earlier token by a different creator has the same ticker and either a near-identical image or a shared identity link (X, Telegram or website). The original has to come from a different known creator: a creator relaunching their own coin is never called a copycat. Since the 21 October backtest it no longer locks the score: copycats rugged less often than average, so the lock was making the tiers less accurate.
The token's own X account, Telegram or website also appears on at least 3 other tokens from at least 3 other creators in the last 30 days. That is worth knowing, so the sheet lists them, but it costs no points and sets no lock.
Links to tweets, X communities and platforms (pump.fun, DexScreener, YouTube) are listed but never raise the flag.
Mint account (authorities, Token-2022 extensions, supply), largest token accounts, the creation transaction and the slots after it.
Bonding curve state, creator, graduation status, PumpSwap pool after migration.
Creator profile and username, where Pump.fun still answers.
Price, market cap, liquidity, 24h volume and buy/sell counts.
New-token and migration events that feed the radar and the auto-scanner.
Scan history for the scoreboard, tracker deltas, creator events and the wanted board.
Embed a live LPX badge or share card on your site, your README or your Telegram pinned post. No key, no sign-up. It shows the latest read LPX has on file for the token and updates whenever the token is read again.
Not sure where to paste it? How to use the badge has a builder for your own token and the steps for a website, a README and chat apps. The endpoints are listed below.
A small SVG badge with the tier and score from the latest read on file. A token LPX has never read shows as unknown until someone scans it.
GET https://lpx.sh/api/badge/EMJtbKt7unnxukFMZDzm9WKdMLAqbMeNtjRMSANnpumpThe 320×80 badge with the tier glyph and the verdict.
GET https://lpx.sh/api/badge/EMJtbKt7unnxukFMZDzm9WKdMLAqbMeNtjRMSANnpump?style=bigThe 1200×630 image link previews use.
GET https://lpx.sh/api/og/EMJtbKt7unnxukFMZDzm9WKdMLAqbMeNtjRMSANnpump[](https://lpx.sh/t/EMJtbKt7unnxukFMZDzm9WKdMLAqbMeNtjRMSANnpump)LPX started recording what happens to every scanned token at 1 hour, 6 hours, 24 hours, 3 days and 7 days in November 2026. The first weekly report has not run yet, so there is no accuracy number to show. When there is one it appears here as it is, good or bad.
Once a week LPX compares what it said at each token's first scan with what happened next. A rule changes only when a small, bounded adjustment did better on later tokens than the ones it was fitted on, ran unseen next to the live rules first, and held up there too. Each change is listed here with its numbers. A change that makes live results worse is undone and that is listed as well.
No change yet. LPX started measuring its own results on 7 Nov 2026 and will change a rule only when later data shows the change is better.