METHODOLOGY

How ElVoid AI’s Score Works

ElVoid AI distills seven categories of public, real-time market data into two composite reads on every asset it watches: an AI Score (momentum and opportunity) and a Risk Assessment score. Every score ships with a Confidence rating, so you can see how much independent evidence backs a read — not just how strong that read is.

Signal Inputs

The exact weighting between categories is deliberately not published — publishing it would let bad actors reverse-engineer and game the score. What’s published instead is what goes in, and how confident the output is.

How Confidence Is Calculated

Confidence reflects how many of the seven input categories independently corroborate a read — not how extreme any single signal is. An asset with strong momentum but no whale or derivatives confirmation carries a lower confidence than one where several categories agree. Confidence is deliberately capped short of certainty: no combination of public signals is proof, and ElVoid AI never reports 100%.

What These Scores Are Not

• — Not a prediction. Scores describe current conditions, not future prices.
• — Not an audit. Risk Assessment reads public liquidity and trading patterns; it does not review smart contract code.
• — Not financial advice. Every score is a starting point for your own research, not a recommendation.
• — Not static. Markets move fast — scores are recalculated on every refresh and can shift within minutes.

ElVoid AI Paper Trader

ElVoid AI is ELSTAND INTELLIGENCE’s signal-generation engine for paper trading only — a simulation with virtual balance, never connected to a real exchange and never touching real funds. Every scan below runs on live candle, whale, and news data, and every output is framed as Probability, Confidence, and Risk — not a promise of where price is headed.

10 Scan Categories

From Scan to Signal

The first 9 categories each vote bullish, bearish, or neutral with a weight attached — ElVoid AI takes the side with more weighted evidence as LONG or SHORT. Entry is the live price; Stop Loss sits just beyond the nearest protective support/resistance level (plus a small ATR buffer); Take Profit 1 and Take Profit 2 target the nearest opposing liquidity levels, falling back to fixed reward:risk multiples when no clear level exists. Risk Assessment — the 10th category — doesn’t vote on direction; it evaluates volatility, R:R quality, upcoming high-impact macro events, and funding-rate crowding, then trims Confidence down when the setup is objectively riskier.

Extended AI Reasoning (TP3, RR, and 5 more checklist items)

The 2026-07 redesign added a 3rd take-profit target, a per-target Reward:Risk readout, and 5 more presentational reasoning lines — Fair Value Gap, Order Block, Funding, Open Interest, and SMT (Smart Money Divergence) — shown on every signal card’s AI Reasoning checklist alongside the original 9. These are intentionally kept out of the Confidence vote above: the original weighting was calibrated against exactly 9 categories, and changing that denominator would silently shift every historical Confidence number. SMT specifically is a simplified proxy — it compares this asset’s 24h change against BTC’s own 24h/7d trend, not a full cross-pair swing-structure comparison — labeled clearly so it’s never mistaken for more precision than it has.

Paper Trading Mechanics

• — Every trade risks a fixed % of equity (default 1%, adjustable in Settings) — a full Stop Loss is always exactly that percentage, never more.
• — When price reaches TP1, the stop moves to breakeven and the position keeps running toward TP2, instead of closing the whole position at the first target.
• — Win Rate, Profit Factor, Average RR, and Max Drawdown are recalculated from actual closed paper trades in ai_journal — nothing is simulated or estimated after the fact.
• — For strategies with at least 5 closed trades, their historical win rate nudges future Confidence for that same strategy label — capped at ±8 points, so history informs the score without ever dominating it.

Try ElVoid AI Paper Trader →

Crypto Heatmap & Token Scanner

The heatmap on the Home dashboard colors each tile by rule, not model output. Tile size is bucketed by market-cap rank. Color priority: purple if the coin is on the Top Rugpull Risk list (score ≥ 60) — this overrides everything else, since risk flags matter more than a green candle; otherwise blue if it shows meaningful net whale inflow with price that hasn’t already run (Smart Money Accumulation); otherwise green or red by 24h change, with opacity scaling by magnitude.

Token Scanner’s 7 categories are all rule-based re-reads of the same live data the rest of the app already fetches — nothing here is a separate prediction model. Top Pump Candidate and Top Rugpull Risk come from lib/scoring.ts; Top Dump Candidate, Smart Money Accumulation, High Momentum, Whale Buying, and Whale Selling come from lib/scanner-categories.ts — Top Dump Candidate is the bearish mirror of the pump-scoring rules (price decline, accelerating downside, heavy sell-side turnover, whale outflow, crowded-long funding unwinding), and Smart Money Accumulation ranks symbols by net whale inflow where price hasn’t already moved — the “quiet accumulation” read, not a chase.

Data Sources

ElVoid AI draws exclusively from public market data: exchange price and volume feeds, OHLCV candle history from Binance, on-chain liquidity and transfer data, derivatives exchange funding and open interest, aggregated financial news, DefiLlama’s public stablecoin-supply index, and FRED’s public macro series for the DXY proxy and M2 money supply. ElVoid AI does not use private, insider, or non-public information of any kind.