Every DeFi transaction generates data — churn patterns, purchase intent, engagement decay. It's public, permissionless, and almost entirely untapped by systematic strategies. BlockSight extracts predictive intelligence from on-chain activity and turns it into tradeable alpha.
Four prediction types. Zero correlation with existing strategies.
Derived from on-chain engagement patterns — not price, volume, or sentiment data.
Churn Probability
Will this wallet stop engaging? Configurable horizons (7/14/30/90d). Aggregate churn risk across a protocol is a leading indicator of TVL decline — the signal arrives before price moves.
Purchase Affinity
Predicted next-purchase category and timing per wallet. Aggregate purchase intent correlates with future demand for a protocol’s token or ecosystem assets.
Engagement Decay
The velocity at which wallet interaction frequency declines. Systematic decay across a protocol’s user cohort predicts TVL outflow weeks before it materializes.
Behavioral Credit
Creditworthiness derived from engagement history, repayment patterns, and cross-protocol behavior. Goes beyond collateral ratios for DeFi lending counterparty risk.
Why quant firms aren't trading on-chain predictions yet.
The alpha is sitting in plain sight. On-chain engagement patterns are structurally analogous to the factors that drive traditional quant — order flow, microstructure, sentiment — but generated by pseudonymous wallets across decentralized protocols.
The problem isn't data quality. It's infrastructure. Getting clean, structured, cross-chain data from blockchains is an engineering nightmare that most firms won't touch.
Archive nodes per chain — 2TB+ storage, constant syncing, DevOps overhead. Multiply by every chain you need.
Raw transaction parsing — logs, events, internal calls in different formats. Months of engineering before one data point.
Cross-chain normalization — a swap on Uniswap looks nothing like PancakeSwap. Every protocol, a different schema.
Continuous maintenance — new protocols, contract upgrades, chain forks. Your pipeline breaks weekly.
One API call. BlockSight delivers structured, prediction-ready data across every major EVM chain — without running a single node or maintaining a single pipeline.
From raw chain data to tradeable insight in under three seconds.
Multi-Chain Data Ingestion
Real-time ingestion across 6+ EVM chains. Raw transactions, event emissions, LP positions, governance votes — decoded, normalized, and unified into a single schema.
BlockSight Encoder
Transformer for temporal sequences + Graph Neural Network for cross-protocol relationships. Produces a dense embedding per wallet that captures engagement patterns across the entire DeFi ecosystem.
Specialized Prediction Heads
Each prediction type has its own head optimized for that task. LSTM + XGBoost for churn. Collaborative filtering for purchase affinity. Ensemble methods for credit scoring. Not an LLM wrapper.
REST API + WebSocket. Query any wallet for churn risk, purchase intent, engagement quality, credit tier.
Where BlockSight creates edge.
Each use case maps a BlockSight prediction to a concrete trading or risk strategy.
DEX / LP Strategies
Churn + DecayPredict which LPs will withdraw liquidity and when. Time entries and exits around pool depth changes before they hit order flow.
Lending Protocols
Behavioral CreditAssess borrower reliability beyond collateral ratios. Enable undercollateralized lending with engagement-based risk tiers.
Governance / DAOs
Engagement DecayPredict vote outcomes before voting closes. Identify proposal momentum from engagement patterns of participating wallets.
MEV / Prop Desks
All signalsIdentify whale accumulation phases, protocol migration flows, and systematic trading patterns before they manifest in mempool activity.
Institutional Allocators
Churn + CreditPortfolio-level analytics for DeFi positions. Counterparty assessment, engagement-based due diligence, compliance-ready reporting.
Risk Management
All signalsReal-time risk monitoring across lending, LP, and governance positions. Early warning system for protocol-level deterioration.
Others describe.
BlockSight predicts.
Existing on-chain analytics describe what happened. Social sentiment tools measure what people say. Neither predicts what users will actually do.
BlockSight's predictions are derived from what wallets do — longitudinal engagement patterns that are structurally uncorrelated with price, volume, and sentiment-based strategies.
Purpose-built multi-task learning system. Transformer + GNN encoder branching into specialized prediction heads. The methodology originates from 30+ years of NASA-funded research on forecasting rare, high-consequence events from longitudinal time-series data.
“Attention correlates with short-term price. Behavior predicts long-term economic participation.”
Built, validated, shipping.
Quant science meets chain infrastructure.
Multiple products shipped to production. Smart contract architecture, payment systems, product strategy.
Architected AI-driven trading engines across decentralized markets. Led Studio Chain L2. Deep ML infrastructure + on-chain systems.
GSU Professor. Artemis program AI systems. The prediction methodology behind the platform originates from his work on forecasting rare events from longitudinal time-series.
Predictive intelligence for DeFi, delivered via API.
BlockSight API access is available to qualified quantitative firms, institutional allocators, and DeFi protocols. We're onboarding partners who want early access to an entirely new factor class.
devon@blocksight.nl