built on nautilus trader·open strategy research·19.6K ★ foundation·crowdsourced intelligence·multi-source token intelligence·don't trust, verify·contribute strategies·backtest with real data·collective intelligence·7 data sources·built on nautilus trader·open strategy research·19.6K ★ foundation·crowdsourced intelligence·multi-source token intelligence·don't trust, verify·contribute strategies·backtest with real data·collective intelligence·7 data sources·
// fathom · open research · collective intelligence

Crowdsourced Strategy Research

The infrastructure exists. The data is public. But finding profitable strategies for Solana memecoins requires more minds, more experiments, and more compute than any single trader can run alone.

Fathom is an open strategy research platform — built on nautilus_trader's architecture, adapted for the trenches. Contribute strategies. Backtest on real data. Get paid for what works.

0%
win rate (so far)
0+
mints scanned
0
data sources
0K
nautilus ★
// origins

Built on Nautilus Trader

We didn't build a trading engine from scratch. We adapted one of the most battle-tested open-source platforms in existence.

From Nautilus

Event-driven architecture (EventBus, typed events)
Strategy base class with backtest/live parity
Adapter pattern for exchange connectivity
Paper trading with simulated fills

Built for the trenches

+ 7-source token discovery + real-time monitoring
+ Multi-factor scoring model (5 signal categories)
+ On-chain holder analysis via Solana RPC
+ 7-source discovery pipeline (DexScreener, pump.fun, GeckoTerminal)
// the bottleneck

Why Open Research

Data is expensive

Helius RPC, DexScreener, pump.fun — rate limits everywhere. Meaningful backtests need weeks of continuous collection across thousands of tokens. One person's free tier doesn't cut it.

Strategy space is massive

Holder analysis, momentum, mean reversion, social signals, deployer forensics — the combinations are infinite. No single trader can explore them all.

The edge compounds

A strategy that works for one person works better when refined by many. Shared backtests, shared data, shared learnings. Collective intelligence beats solo grinding.

“If it really works, why share it?”

Because finding what works requires more experiments than any single person can run. The infrastructure is here. The strategies are the hard part. That's what we're crowdsourcing.

// live data

Token Feed

Live Solana tokens scored by fathom's multi-factor model. Every token links to its source. Updates every 30s.

LIVE FEED
live tokens · 0s ago
0 trade0 skip
scanning tokens...
scoring: momentum · holders · liquidity · dev · snipersrefresh 30s
// thesis

Why On-Chain Filtering Works

Thousands of tokens launch on Solana every day. Most are noise. The alpha isn't in finding tokens — it's in knowing which ones to ignore.

01

Multi-source discovery

7 data sources feeding one pipeline — pump.fun, DexScreener, GeckoTerminal, Helius, Jupiter. Tokens surface before CT catches them.

02

Holder forensics

On-chain holder distribution, dev wallet activity, sniper detection, top-10 concentration. The data that separates real projects from exit scams.

03

Liquidity validation

Fake billion-dollar mcaps with $10K liquidity? Ratio analysis catches them instantly. Real depth vs paper valuations.

04

Conviction scoring

5 signal categories weighted into a single score. Not binary trade/skip — graduated conviction with dynamic position sizing.

Speed bots will always be faster. MEV will always front-run. But being first doesn't matter if you're buying a token that dumps to zero.

The edge is intelligence — knowing what to skip before you ever place a trade.

What gets filtered

Fake mcaps

Golden: $1.07B mcap, $12K liquidity = 89,166:1 ratio

Insider concentration

If 5 wallets own 90% of supply, they dump on retail

Serial deployers

HgaD5ZSN launched 7 fake tokens in 24 hours

// backtest

Real Results

347+ mints through the default scoring model across multiple collection runs. Early. Promising. Needs more minds on it.

Win Rate
58.3%
Return
+56.8%
Trades
38 / 67
taken / scanned
Max Drawdown
-30%
fathom backtest
$ python -m fathom backtest --data data/collect-20260216.json
Top performers:
OLO +50.0% TP hit score=78 $88K mcap
FIRST +50.0% TP hit score=74 $34K mcap
TEDDY +50.0% TP hit score=71 $18K mcap
Duck +50.0% TP hit score=68 $13K mcap
Filtered out (scam detection):
Golden $1.11B mcap, $0 liq ratio ∞ SKIP
Vooerk $1.05B mcap, $0 liq ratio ∞ SKIP
C0IN $740M mcap, $0 liq deployer HgaD5ZSN
Strategic $556M mcap, $0 liq ratio ∞ SKIP
Strategy: take_profit 50% | stop_loss 20% | trailing 15% | max_hold 10min
347 mints scanned | 67 graduations | Completed in 1.847s

// NOTABLE PICKS

High-Conviction Captures

Tokens the scoring model flagged with high conviction at graduation — tracked from bond to peak.

MOMOMOMO$MOMOSCORE 82TRADE

High buy ratio, strong holder distribution, massive volume surge post-graduation

+808%
bond → peak
MCAP $322KVOL $4.3MTXNS 57K
PreguntalePreguntale$PREGSCORE 78TRADE

Clean on-chain metrics, low top-10 concentration, sustained momentum

+712%
bond → peak
MCAP $401KVOL $2.5MTXNS 20K
SAVESAVE$SAVESCORE 74TRADE

Highest transaction count in dataset — massive organic activity signal

+173%
bond → peak
MCAP $143KVOL $269KTXNS 97K
TRENCHTRENCH$TRENCHSCORE 71TRADE

Strong liquidity health, clean holder profile, steady accumulation pattern

+105%
bond → peak
MCAP $296KVOL $201KTXNS 8.3K

Performance measured from graduation to 24h peak. Past performance does not guarantee future results.

// trust layer

Scam detection

Wallet HgaD5ZSN... deployed 7 fake tokens with billion-dollar mcaps and ~$10K liquidity. All caught.

C0IN$399M$11Kmcap/liq ratio 34,694:1BLOCKED
CION$2.6B$20Kmcap/liq ratio 130,000:1BLOCKED
Golden$1.07B$12Ktop 10 hold 100% of supplyBLOCKED
MrBeast$235M$9.2Ksame deployer: HgaD5ZSN...BLOCKED
XMoney$719M$10Ksame deployer: HgaD5ZSN...BLOCKED
Strategic$556M$10Ksame deployer: HgaD5ZSN...BLOCKED
Vooerk$975M$10Ksame deployer: HgaD5ZSN...BLOCKED
// quickstart

Start exploring

Clone. Collect data. Test a hypothesis. Share what you find.

$git clone https://github.com/late-build/fathom.git && cd fathom
$cp fathom.toml.example fathom.toml
# collect real token data
$python -m fathom collect --hours 24
# backtest — reproduce our numbers
$python -m fathom backtest --data data/collect-latest.json
# try your own strategy params
$python -m fathom backtest --tp 0.4 --sl 0.15 --data data/collect-latest.json
# paper trade live
$python -m fathom run --mode paper