Methodology
How QuantScan actually works
Full transparency on the data, the algorithms, the grading formula, and the limitations. No black boxes. No “proprietary AI.” Just rule-based scanners you can verify.
The pipeline
Data → signal → grade
Universe scan
Every refresh cycle (~60s) we pull live OHLCV + funding + OI for 500+ coins from Hyperliquid and Binance. One unified candle store powers every scanner.
Per-scanner detection
Each of 51 scanners is a deterministic Python-style function. Inputs: candles, indicators, funding, OI. Output: zero or more signals with direction (long/short), entry, timeframe, and a raw scanner score.
Confluence scoring
Coins are bucketed by symbol. We count how many scanners fired on the same coin in the same direction across distinct categories. More categories agreeing = higher confluence.
Grade assignment
A composite score combines scanner strength, regime alignment, multi-timeframe agreement, volume confirmation, bias alignment, and confluence rarity. Score is mapped to S/A/B/C grades.
Surface + cache
Top setups by grade are published to /today and /dashboard. Results cached in Redis (60s TTL) so every Vercel instance shows the same numbers — no instance-bounce.
Data sources
Where every number comes from
All data is from public APIs. We don't buy proprietary feeds, we don't scrape dark pools, and we don't make up numbers. Everything below is verifiable.
Hyperliquid
Primary perp dataLive mark price, prev-day mark, funding rates, open interest, and order book depth across the full HL universe. Polled directly from api.hyperliquid.xyz/info.
Binance Futures
Backup + breadthUSDT-margined perp tickers used to extend coverage when HL doesn't list a coin. 24h klines feed our breakout / momentum scanners on the long tail.
CoinGecko
Macro contextTotal market cap, BTC dominance, 24h volume. Used only for the market-pulse strip and regime sanity checks — never for individual signal decisions.
Alternative.me
SentimentCrypto Fear & Greed index (0-100). Displayed as context — not used as a signal input.
CryptoPanic
News pulseAggregated headlines tagged by sentiment + impact. Surfaced in the /today brief, never auto-traded on.
Scanner categories
51 scanners across 6 strategy types
We organize by strategy type, not timeframe. A bull flag is a chart pattern whether you trade it on the 15m or the 4h. This taxonomy makes confluence meaningful: when a Trend scanner and a Smart Money scanner both fire on the same coin, that's different evidence than two trend scanners agreeing.
Grading
S / A / B / C
Every signal gets a composite score (0-100) combining six factors: scanner strength, regime alignment, multi-timeframe agreement, volume confirmation, directional bias alignment, and cross-scanner category diversity. The score maps to a letter grade.
Elite. 4+ category agreement, high rarity, regime-aligned.
Strong. 3 category agreement with volume + regime support.
Solid. 2-3 scanner agreement, decent context.
Watch. Single scanner or weak confluence; only show if clean entry.
Regime detection
Know the market state before you trade
Every scanner output is tagged with the current regime. Some scanners only run in certain regimes (e.g. mean-reversion is suppressed in CHAOTIC). Regime is derived from BTC 24h change on the perp markets — the same number we trade.
BULLISH
BTC 24h > +2%Trend strength is firm with broad participation. Momentum + breakout favored over mean-reversion.
BEARISH
BTC 24h < -2%Downside pressure dominates. Defensive scanners + short-bias setups favored.
CHAOTIC
abs(BTC 24h) > 5%Volatility elevated, direction unstable. Reduce size, prefer breakout-confirmation.
NEUTRAL
everything elseNo clear directional edge. Range-bound conditions favor mean-reversion + tight risk.
Refresh cadence
How fresh are the numbers?
Caches are deliberate. Hitting Hyperliquid's API every page load would burn rate limits and cost us money for no user benefit — the underlying data doesn't change meaningfully on a sub-minute scale for setup detection.
Limitations
What QuantScan is not
Not financial advice. We surface setups. You decide what to trade and how much to risk. Past scanner performance does not guarantee future results.
Not a signal service.We don't tell you when to buy or sell. We tell you what scanners fired and how strong the confluence is — the rest is your call.
Not predictive. Every scanner is a detector of a present condition (compression, volume spike, regime shift). None of them predict the future. They identify situations where historically certain outcomes have been more probable.
Not exhaustive. We cover the 500+ coins on Hyperliquid + Binance Futures. Smaller alts, spot-only tokens, and pre-listing assets are out of scope.
Not infallible.Upstream APIs go down. Funding rates spike on illiquid pairs. We log issues and surface a “demo mode” banner when live data is unavailable. Always sanity-check before sizing into a position.
Questions about the methodology? Email support@quantscanai.com.
See today's scan →