How a call gets made

Every part of a call, explained on a real one from Oct 12, 2026.

01
Market data
Every pool it trades in
02
Safety scan
Traps and who holds it
03
Launch scan
Young coins: the first buyers
04
AI read
One fixed scorecard
05
The call
Odds against the costs

1
The coin
Name, ticker, contract address (click to copy) and chain. The logos open it on DexScreener, GMGN, Axiom and Fomo. Under it: age and launchpad.
2
Market data
Price, market cap, liquidity and 24h volume, added up across every pool it trades in. One mispriced pool can't set the price. Hover the chart for any point in time.
3
Safety scan
Can you sell, can the owner mint or freeze, taxes, locks and who holds the supply. Exchange and burn wallets are counted apart from the top 10. Coins under a month old also get a launch scan: snipers, bundles, the dev's buy.
4
Social
What X and Telegram say about the coin. Coming soon.
5
Cost to sell now
What selling $100, $1K or $10K right now would cost you: price impact plus sell tax.
6
Analysis
An AI read on a fixed scorecard: momentum (-100% to +100%), liquidity risk and manipulation risk (0% to 100%), a short summary and red flags. It feeds the odds. It doesn't make the call.
7
The call
Buy, Don't buy or Wait, with three odds over the stage's window: up X% first, down Y% first, or neither. Then the measured start for coins like it, the break-even after costs, what moved the odds and the plan.

Token Battles$FIGHT

on Solana

Graduated from pump.fun, 22 hours old

Market data

Price: $0.0002695
Market cap: $235.41K
Liquidity: $46.01K
Volume 24h: $1.22M

1h -23.52%6h +6.07%24h +489.00%

Safety scan

6,219 holdersBubble map

Pass: 13% of the supply burned
Pass: Supply is fixed
Pass: No freeze switch
Pass: Top 10 wallets hold 12%
Pass: 100% of the liquidity is locked

Launch

Risk: The dev bought 14% at launch
Watch: 1 wallets bought 3% in the launch block
Watch: 2 snipers took 5% in the first seconds
Watch: Those early wallets still hold 0.0%

Social

What X and Telegram say about it: coming soon.

Cost to sell now

Sell $100: 0.43%
Sell $1K: 4.2%
Sell $10K: 30.3%

Price impact and sell tax, at best.

Analysis

Momentum: +20%
Liquidity risk: 65%
Manipulation risk: 60%

This token graduated from pump.fun and sits at a $235k market cap with $46k in liquidity. Momentum is mixed as an earlier surge cooled with a sharp drop over the past hour despite strong overall volume. High volume relative to liquidity and heavy same-block buying at launch introduce notable execution and manipulation risks.

Volume is $1.2M against $46k liquidity, a high volume-to-liquidity ratio
Selling $10,000 incurs a 30% price impact loss
Up to 22 buyers entered in the same block during launch

Call

Don't buy

Lottery: only money you can lose

It's more likely to drop 50% than to rise 100% first.

Odds over the next 3 days

Measured on actively traded coins under a day old: up first 12%, down first 24%. To break even after costs, up first needs 24% (a $100 trade costs about 1.8% in and out).

Against: The creator already sold most of their launch buy
Against: 22 wallets bought in one block early on: coordinated buying
For: Its own socials are linked (X, a website)

A real call, run Oct 12, 2026 at 11:47 PM UTC. Not a recommendation. Run it yourself

Read for its age

Each stage has its own targets and window: up this much first, or down this much first, within this time.

Stage	Up / down first	Within	Risk
Still on its launch curve	+100% / -50%	1 day	Lottery
Under a day oldThe example	+100% / -50%	3 days	Lottery
1 to 7 days old	+60% / -35%	3 days	High risk
A week old or more	+40% / -25%	a week	Risky
$100M+ and deep liquidity	+20% / -15%	a week	Volatile

Odds from measured data

We measure each coin's potential against base rates from thousands of real coins at the same stage, then move the odds with the evidence. The weights come from our own research and from published studies, peer-reviewed papers among them.

The start: actively traded coins under a day old, over 3 days

Up first 12%Down first 24%Neither 65%

What moved it

The creator already sold most of their launch buy
22 wallets bought in one block early on: coordinated buying
Its own socials are linked (X, a website)

Where it landed

Up 100% first 1%Down 50% first 61%Neither 37%

The three always add up to 100%. Signs that can be faked, like dust holders or wash volume, count less, and anything we can't read pulls the odds back toward the start.

Buy, Don't buy or Wait

A Buy doesn't need better than 50%. It needs up first to beat the break-even: the point where the trade pays for itself after the cost of getting in and out.

Buy

Up first beats the break-even by a clear margin. Every Buy comes with a plan: the market cap to aim for, where the setup breaks, and when to take your initial out.

Don't buy

A trap from the safety scan, or the odds against you: down first is much more likely, or the trade loses money on average after costs.

Wait

Not enough edge yet, or too much we couldn't read. Wait always says what would change it.

The example, worked out

A $100 trade costs: 1.8%
Break-even: 24%
Up 100% first: 1%
The call: Don't buy

Any one of these is a Don't buy

You can't sell it, or the sell tax is over 10%
It's been flagged as rugged
The creator can still mint more or freeze wallets
A big share of the liquidity unlocks within a week
The creator or insiders are dumping right now
Under $5k of liquidity

Already holding it?

Add your amount and average entry market cap. You get your profit or loss, the cost to sell it all and a call of your own, based on what comes next, not on what you paid.

Buy more

Only on a clear new edge, never just because it went up.

Hold

You already paid to get in, so it takes a worse outlook than a buyer's to sell.

Sell part

Take your initial out at about 2x after costs, or trim a bag that's too big for the pool.

Sell all

When staying is likely to lose more than selling costs, or a trap shows up.

The math behind Neurolith

What every call computes. The inputs and their weights come from our measurements and the research below.

01Three outcomes

o=\begin{cases}U & \text{if } p_t/p_0\ge 1+X \text{ first}\\ D & \text{if } p_t/p_0\le 1-Y \text{ first}\\ N & \text{if neither by } t=T\end{cases}

Every call asks one question: from today's price, does the coin go up $X$ first, down $Y$ first, or neither within the window $T$ ? All three depend on the stage.

02The odds

\begin{aligned}P(k)&=\frac{e^{z_k}}{e^{z_U}+e^{z_D}+1}\\ z_k&=\ln\frac{b_k}{b_N}+c_k+\sum_i w_{k,i}\,x_i\end{aligned}

The odds start at the base rates $b$ , measured on real coins at the same stage, corrected by $c$ , fitted on how our own past calls turned out. Each piece of evidence $x$ moves them by its weight $w$ . The three always add up to 100%.

03Expected return and break-even

\begin{aligned}\mathbb{E}[R]&=P_U X-P_D Y+P_N r_N-c\\ P_U^{*}&=\frac{(1-P_N)\,Y+c-P_N\,r_N}{X+Y}\end{aligned}

$c$ is the round trip's cost at your size, $r_N$ the usual return when neither line is hit. A Buy needs $P_U$ above the break-even $P_U^{*}$ by a margin, not above 50%.

04Cost to sell

\begin{aligned}\text{impact}(s)&=\frac{s}{L/2+s}\\ \text{loss}(s)&=1-\bigl(1-\text{impact}(s)\bigr)(1-\tau)\end{aligned}

Selling $s$ dollars into a constant-product pool holding $L$ dollars of liquidity, with a sell tax $\tau$ . A best case: one thin pool loses more.

05Your position

\begin{aligned}m&=\frac{\mathrm{MC}_{\text{now}}}{\mathrm{MC}_{\text{entry}}}\\ \text{PnL}&=C\,(m-1)\\ \frac{1}{m}&=f\,\bigl(1-\text{loss}(f\,C\,m)\bigr)\end{aligned}

$C$ is what you put in, $m$ how many times it has grown. $f$ is the share to sell to take your initial out after costs, solved numerically.

06Checking the odds

\mathrm{BS}=\frac{1}{N}\sum_{t=1}^{N}\ \sum_{k\in\{U,D,N\}}\bigl(p_{t,k}-o_{t,k}\bigr)^2

The Brier score: each call is scored against what happened ( $o$ is 1 for the outcome that came first). Lower is better, and 40% should come true about 40% of the time.

Read the paper

Meme Coin Factories: Uncovering Large-Scale Manipulations on pump.fun

Szwajcok, Tsuchiya, Liu, Soska, Payer, Christin

A large study of pump.fun: wash trading in at least 17% of all trades, copycat coins, and a small set of creator networks behind most launches. Signs a call has to discount.

ACM CCS 2026, peer-reviewed

Sources

1.Liu, Tsyvinski, Wu. Common Risk Factors in Cryptocurrency. The Journal of Finance 77(2), 2022Peer-reviewed
2.Zaremba, Bilgin, Long, Mercik, Szczygielski. Up or down? Short-term reversal, momentum, and liquidity effects in cryptocurrency markets. International Review of Financial Analysis 78, 2021Peer-reviewed
3.Szwajcok, Tsuchiya, Liu, Soska, Payer, Christin. Meme Coin Factories: Uncovering Large-Scale Manipulations on pump.fun. ACM CCS 2026Peer-reviewed
4.Gneiting, Raftery. Strictly Proper Scoring Rules, Prediction, and Estimation. Journal of the American Statistical Association 102(477), 2007Peer-reviewed
5.Brier. Verification of Forecasts Expressed in Terms of Probability. Monthly Weather Review 78(1), 1950Peer-reviewed
6.López de Prado. Advances in Financial Machine Learning (the triple-barrier method). Wiley, 2018Book
7.Marino, Naviglio, Tarantelli, Lillo. Predicting the success of new crypto-tokens: the Pump.fun case. 2026Preprint
8.Mongardini, Mei. A Midsummer Meme's Dream: Investigating Market Manipulations in the Meme Coin Ecosystem. 2025Preprint
9.Angeris, Kao, Chiang, Noyes, Chitra. An analysis of Uniswap markets. 2019Preprint
10.Neurolith's own measurements: 3,500 pump.fun graduates followed through each stage's window (October 2026), re-measured as new coins come in.

Checked against what happened

Every call is saved with the inputs it saw and scored on what the price did next.

Call
Saved with everything it saw.
+1, 6, 24 hours
The real price is recorded.
End of the window
Which came first: up, down or neither.
Checked
Did 40% odds come true about 40% of the time?
Tuned
Starting odds re-measured, weights adjusted.

See it on a real coin

Paste any token's contract address and get an analysis and a conclusion with the odds, in seconds. Free, no wallet needed.

Try It Now

Reading a call

The coin