How the grade works

This is the rule everything else follows from. Most facts about a Solana token minutes old are simply unknown: the holder table has not been read, the launch analysis has not run, no third party has published a report. A screener that substitutes a neutral value for each of those produces a confident-looking number for a token nobody has checked.

So every pillar is tri-state at the source. A pillar with no inputs returns null, not 50. The weighted sum is taken over the pillars that were measured and renormalised to their own weights, and coverage reports how much of the token the number actually describes. Below 60% coverage the score is shown as provisional and auto-trade refuses to act on it.

On screen this reaches you as a stone with a facet missing: an unmeasured pillar is a chip cut out of the shape, never a short vertex. A dashed chip means a check that has not run — not a check that came back clean.

Safety and launch dominate because they are the two that take the whole position rather than part of it. Momentum is weighted lowest on purpose: it is the one a manipulator can manufacture cheaply, and the one already visible on the chart you are looking at anyway.

A critical finding clamps the total rather than being averaged. A token with locked liquidity, a fair distribution and good momentum whose freeze authority is still live would otherwise score in the sixties, because four fifths of it is genuinely fine. It can still freeze your account.

The reading this product exists for. Splitting a bag across twelve wallets defeats top-10 concentration, creator percentage, holder count and every other distribution metric there is, and on Solana it costs a fraction of a cent per wallet. What it cannot hide is that all twelve were funded from the same place.

• Bundle — supply taken in the token’s first traded slot.
• Snipers — supply taken within fifteen seconds. Deliberately weighted far below the bundle: a fast buyer is not an insider, and a signal that fires on a third of all buyers is not a signal.
• Clusters — wallets sharing a funding source. The strongest of the three, because it is the one that survives the wallets being split up.

Two sources, in order. What already happened: three or more unrelated wallets having sold in the last six hours is a record, not a simulation, and nothing can fake it. What would happen: for a token too new to have any, a real-size exit is quoted against the real router, and judged on whether it routes and on what it costs — a route that exists but returns four percent of the money is a honeypot with extra steps.

Neither answering leaves the check null, which is not a pass.

A score is a continuous opinion that changes every minute, so it is never wrong. A call is a moment: at 14:02, at a $38K market cap, this was worth buying. It can be measured afterwards and it can be wrong.

A call is rare by construction. It has to clear the grade, a coverage floor, a closed launch window with the bundle actually analysed, a confirmed sell, real liquidity, a minimum age, and no call at the same tier in the last twelve hours. Every call is then tracked forever — where it was called, what it peaked at, where it is now, how far it fell on the way, and how it ended.

The record counts every call in the window the same way, the ones that ran and the ones that went to zero, and it sits above the feed rather than behind a tab. It reports the median peak and the median current multiple: the gap between those two is the honest measure of a feed like this one, because a product that publishes only its peaks is showing the best moment of every call it ever made. Medians rather than means throughout — one 400x carries a mean and says nothing about the typical call.

The engine buys what the product called; your entry rules are a filter on top of that, never a competing opinion. Every refusal names its rule and both numbers, and is recorded whether or not anybody is watching — “why didn’t it buy that one” is the first question anyone asks, and an engine that cannot answer it is one people turn off after a day.

Exits are evaluated worst-news-first: a rug signal beats a stop, a stop beats a take-profit, and the time limit is last. Position size, concurrent positions, hourly spend and the daily loss cap are all required and all bounded — there is no value of any of them meaning “unlimited”, because the failure mode of automated buying is not one bad trade, it is forty bad trades in ninety seconds. The daily cap disarms the engine rather than pausing it.

Arming requires $500 of ROCK held — a position, not a subscription. A subscription is a payment that leaves; a holding is something you keep, and it aligns the one feature that spends money by itself with the token whose price it moves. ROCK in your engine wallet and in any wallet you have linked both count.

It is not advice, and nothing here predicts a price. It is a set of measurements about what a contract permits, who holds the supply, how it was distributed at launch and whether an exit exists — each of them stated with its own uncertainty attached. Most tokens on this chain go to zero. A good grade means the specific ways this one could already be rigged were checked, not that it will go up.