How we decide which tokens to cover (and why most never make it)

Posted by the operator of RebateMax

Every day, an automated pipeline pulls the 100 trending Solana tokens from Jupiter's organic-score feed and measures each one against the same frozen checklist. On a typical day, zero to two of them clear it.

That failure rate is the point. This post is the walkthrough of the filter — the exact thresholds, where the numbers come from, and what happens to a token after we write about it. Nothing here is secret; the full checklist lives at /research-criteria and every covered token lands on the public ledger, where the call is re-measured daily whether it aged well or not.

The checklist, in plain terms

A token has to clear every one of these. There is no scoring, no vibes, and no making up for a failed check with a strong one elsewhere:

• At least 45 days old. Most rugs happen in week one. A token that has traded for six weeks has survived the window where most of them die.
• At least $150,000 of liquidity, with $75,000 in one pool. Thin liquidity means you can buy in but not out.
• 24h volume no more than 4× liquidity. Higher than that reads as wash trading — volume manufactured to look popular.
• Top 10 wallets hold no more than 25% of supply. Concentration is exit risk: ten wallets that can end the chart on their own.
• Mint authority revoked and freeze authority revoked. If either is live, someone can print more supply or freeze yours. We never waive these two.
• A deployer wallet that is not a token factory. No more than 25 tokens minted and 5 migrations from the same wallet. The single most rejected-on check that scanners skip: we once measured a "clean" token whose deployer had minted 140,483 tokens and pushed 1,078 of them to market. The contract was fine. The wallet was a production line.
• At least 90% of material liquidity locked or burned, and the deepest pool at least 95% locked. We weigh money, not pool counts — a token with three pools where the deep one is locked and two dust pools are not is safer than the reverse.
• A discoverable X profile. No verifiable public presence, no coverage.

Where the numbers come from — and the rule that keeps us honest

Three public sources: Jupiter (holders, concentration, authorities, deployer history), DexScreener (per-pool liquidity, volume, pool ages), RugCheck (liquidity locks). Anyone can check them; our posts link each figure to its source.

Sources disagree constantly. When they do, we record the lower number. When we covered TROLL, Jupiter said $1.32M of liquidity and DexScreener said $2.35M — the ledger says $1.32M. A filter that quietly takes the flattering source is not a filter.

The same logic runs the other way: "unknown counts as failed." If a source cannot tell us the lock status or the holder concentration, the token fails that check. It does not get the benefit of the doubt.

Clearing the checklist is not a recommendation

This is the part most token content gets wrong, so we put it in every post: clearing the checklist means measured and not disqualified. Nothing more.

The two tokens that have cleared it so far — TROLL and aura, both memecoins — were covered with that exact sentence attached. A memecoin that passes every safety check is still a memecoin. What the filter tells you is that the specific, checkable failure modes — printable supply, freezable balances, withdrawable liquidity, factory deployers, exit-sized concentration — were measured and not found. It says nothing about whether the price goes up.

And a human makes the final call on every token. The pipeline finds and measures; it never publishes on its own.

The part that keeps us accountable

Anyone can publish a checklist. The harder commitment is the ledger.

Every covered token goes on /research-ledger with its numbers from coverage day — market cap, liquidity, holders — and gets re-measured automatically every day after. Both columns are public. If a call ages badly, the row stays, with the bad numbers next to the date we made it. Nothing is deleted, a covered token cannot be re-covered inside 90 days (so the ledger cannot be padded by re-covering winners), and corrections are published dated, with the original and corrected figure side by side: /research-corrections.

That is the whole system. A frozen checklist, public sources, the lower number, a human gate, and a ledger that remembers. If you take one thing for your own research, take the deployer check — it costs nothing to look up and almost nobody does it.

---

RebateMax research carries no referral links and is not financial advice. The criteria are versioned at /research-criteria; changing them is a new version with a date, never a silent edit.