What the "Run the research" button does
It restarts the whole chain from scratch: queries the sources again, recomputes the
scores at this second's price and rewrites the list. Trending lists and safety reports
are cached briefly (from 20 seconds to 5 minutes) so fast refreshing doesn't burn
through the sources' limits; prices and candles are always fresh. The first run takes
10–25 seconds, later ones much less.
Checking a coin you found yourself
Paste a token's contract address (CA) into "Check a coin" and the radar runs that one
token through the same chain: the hard filters, the RugCheck safety report, Jupiter's
holder and trader data, the price candles and the press. You get the same card, verdict
and ticket as a token from the list. The only difference: a coin that fails the hard
filters isn't hidden, the card says exactly which filter it fails. The verdict uses the
market reading from the radar's last run, so the check waits for the first run to
finish. Checked coins aren't added to the signal log.
Auto-refresh and notifications
With "Auto-refresh (10s)" ticked, the page presses the button by itself every 10
seconds while you stay on the radar; if a run hasn't finished yet, it waits for it.
Press "Turn on notifications" once and you get a test notification right away. From then
on, every time a new token gets the "enter now" verdict you get a system notification
with its symbol, entry, stop and first target. Keep the radar open in a tab: with
notifications or sound on, it keeps scanning while you're in another tab or in Axiom (the browser
slows hidden tabs down to about one scan a minute). Clicking the notification brings you
back to the token's card. With "Sound alert" ticked, the page also plays a short chime
when a new token gets "enter now", even from a background tab; the browser only allows
sound after you've clicked somewhere on the page once. A token that has alerted you isn't
announced again until you reload the page. Phones can't show these notifications: on Android the button says so,
and on iPhone the browser doesn't offer them.
The verdict: "enter now", "almost" or "no"
Every token gets one of three verdicts, and every condition stays visible under
"Verdict conditions", with its threshold. A verdict you can't recompute by hand isn't a
verdict, it's an opinion.
- Enter now: every condition is met. The card gives you the full
ticket: entry price, stop, position size, the two targets, the trailing stop and how
long you stay by the chart.
- Almost: every safety condition passes, but exactly one of the
others is missing. The card says which one, and the plan tells you what to do on the
next run.
- Not now: at least one safety condition fails, or two of the
others do.
The non-negotiable conditions: safety above 0.85, authenticity above
0.85, at least three hours since launch and since migration, no vertical move (under +30%
in an hour and under +80% in six), one hour's volume below liquidity, holders
growing between +0.3% and +3% per hour, risk to the stop below 30%, and a
memecoin market that isn't being sold across the board. The others: at least 50 unique
traders per hour, more buys than sells, liquidity above $30,000 and a score of at least
55.
Adoption is a hard condition because it's the only one that clearly separated results
in the forward test: tokens whose holders weren't growing ended in profit 11–12% of the
time, those with steadily growing holders two to three times as often.
"None right now" will often be the answer, and that's correct. If the
page produced a recommendation on every run, it would be making up signals to look
useful.
The trade ticket and the exit ladder
The stop sits below the last hour's support, at least 8% below the price, so a
memecoin's normal noise doesn't hit it. Position size is calculated from it: if the stop
is hit you lose about 1% of your total budget, never more than 5% of the budget on a
single token.
- At +30% sell half and move the stop to your entry price. From that
moment the position can't lose, and the first target is close enough to be hit
often.
- At +80% sell another quarter.
- Leave the rest with a trailing stop 30% below the highest price
reached. The rare big moves come from this last quarter.
The time stop: if it hasn't reached +30% within 4 hours, sell
everything at market, even if the price stop hasn't been hit. No movement is an answer
too, and staying by the chart with no deadline isn't patience, it's hope.
What "Watch" means on each card
It means that during that window you actually keep your eyes on the
token and check the points in the list. When the window ends you aren't left
hanging: every card has three outcomes, and one of them always applies.
- Enter: on the next run it gets "enter now"; you follow the
ticket.
- Drop: it broke the stop, liquidity fell 30%, or a vertical move
appeared. You take it off your list.
- Extend: nothing broke, but the verdict didn't change either. You
give it one more hour; if it still gets no verdict, you drop it.
Having waited is not a reason to enter.
How the score is calculated
score = quality (0-100) × safety (0-1) × authenticity (0-1) × timing (0-1)
Quality adds up eight components: adoption (how many new holders
appear per hour), participation (how many unique traders), maturity, liquidity, a move
with no vertical, buying pressure, volume that's lively but not frantic, and declared
presence. All of them are visible, with their points, under "Why this score". Where the
tests showed that "more" turns bad (holders growing by over 3% an hour, over 400 traders
an hour, volume above liquidity), the points drop again after a peak.
The other three are multipliers, not penalties. A token with unsecured
liquidity isn't "a bit weaker", it's untradeable, and a ten-point deduction would leave it
at the top of the list.
- Safety: from the RugCheck report (locked liquidity, revoked
authorities, transfer fee, concentration in the top ten wallets, how much the creator
holds, linked wallets), completed with Jupiter's audit and the creator's history:
someone who has launched thousands of tokens is a factory, not a project.
- Authenticity: two independent measures: how much price a dollar
of volume moves (money passed around in circles moves nothing), and Jupiter's organic
score, which tracks wallets one by one and separates bots from people.
- Timing: hype signals lower the multiplier: a vertical move in the
last hour, a wave of new holders, volume above liquidity, a launch a few minutes
old.
What changed from the old version, and why
The old version rewarded acceleration: exploding volume, a curve climbing faster and
faster, buying in the bottom third of the hourly range. We tested it in two ways:
- On history: 235 memecoins × 17 days of 5-minute candles (2.7
million candles), entering at the open of the candle after the signal, with 3% costs.
The old rule won 25% of the time, at −3.5% per trade: worse than entering at random on
the same tokens (−2.2%). Acceleration, curve shape and position in the range predicted
nothing.
- Forward, in real time: every candidate recorded at the moment of
the scan, with the result measured hours later, so without knowing in advance which
ones survive. The hype signals were exactly what came before the biggest drops: +30% in
the last hour (−28% over the next three hours), volume above liquidity (−27%), tokens
under three hours old (almost −50%). The best were tokens 2–7 days old with steadily
growing holders.
Hence the current grid: it looks for sustained, organic adoption, not hype. And the old
"enter now" signals in the log, replayed on candles, ended in profit in 1 case out of 6,
which is why the history above marks them separately.
What the platform checks, and what's left for you
The RugCheck report and the Jupiter audit are requested automatically for every
finalist, so the basic safety checks are already done. What's left to do
in Axiom before any entry:
- Confirm the verdict is current. If you're looking at a list from
twenty minutes ago, run it again.
- Look at the real chart. If the entry zone looks like a falling
knife, don't enter.
- Check the slippage for your size. Total liquidity doesn't tell you
what your order costs.
- Read the community. A dead channel or an X account with three posts
doesn't show up in any number here.
What the platform can't see
There's no live Twitter/X feed, because there's no free API for it, so
"declared presence" means the links the project put up, not how much it posts. The press
is read from seven sources and matched strictly by symbol and by the distinctive words in
the name, but a three-hour-old token can't appear in the press: no news isn't a negative
signal.
Candles come from Jupiter's charts, and where Jupiter doesn't respond, from
GeckoTerminal. The Jupiter chart endpoint is the one jup.ag uses, not a publicly
documented one: if it changes, the radar automatically falls back to GeckoTerminal, which
is slower.
And most important: the score describes the present, not the future.