Base rates, not vibes.

We run a sample of new pools on Robinhood Chain through Collar Guardrail, then check back after 24 hours to see what happened on-chain. Public data, with the limits stated.

Read this first

This is a sample of newly created pools, not every pool.
"Failed" means a measurable failure within 24 hours of launch, such as liquidity being pulled. Other kinds of rug pulls are not measured, so the failure rate is a lower bound.
Collar's hit rate is not claimed here yet. It needs enough failed tokens to be measured.

Latest tokens

Most recent pools evaluated by Collar. Click any row for details.

Token	Collar verdict	24h result	Price change	Details
Loading…

How to read: "Collar verdict" is what the checker said at the time of the run. "24h result" is what was measured on-chain afterwards. Rows with an orange edge are ones Collar allowed that later showed a failure flag. These are individual examples, not an accuracy measure. Price change is market movement and is not counted as a failure.

Collar's verdicts

Click a segment to filter the table.

ALLOW = no hard-block rule fired (not an endorsement). WARN = one risky capability or an inconclusive check. DENY = a hard-block rule fired.

Red flags (heuristic)

Suspicious behavior detected after launch. Click a row to list the tokens.

What this is: automated heuristics over swap data (sell traps, bot-like trading, coordinated dumps, wash trading). They can be wrong in both directions.

Where the tokens come from

Loading sources…

Why it matters: if one launchpad dominates the sample, the numbers describe that launchpad more than the whole market.

Methodology and limits

Sample

Pools come from Uniswap V4 Initialize events on Robinhood Chain. A fixed-seed random sample is drawn and each token is evaluated.

What "failed" means (v0.7)

Liquidity pulled within 24 hours of launch. Only pools whose 24-hour window is complete are counted, and pools that never received liquidity are excluded.

Cumulative data

Results are cumulative. Each daily run adds new pools and preserves earlier results. Tokens older than the lookback window are re-used from prior runs instead of re-scanned, keeping the pipeline fast. The actual scan window is configurable (LOOKBACK_HOURS, default 24h after the first run).

Pipeline schedule

The pipeline runs automatically at 3:00 AM ET (08:00 UTC). New results appear in the dashboard after each run completes.

What is not measured

Other rug-pull types (for example a deployer dump or a tax raised after launch) are not part of the v0 failure definition, so the failure rate is a lower bound, not an estimate of all rugs.

Collar verdicts

Collar's v0 rules check bytecode presence, owner status, risky owner-controlled functions (mint, pause, blacklist) and a sell simulation. They do not check liquidity locks, deployer history or market risk.

Conflict of interest

Collar is our product and we run this evaluation. The bot never buys or sells tokens. The code and the raw data are public so anyone can recompute these numbers.