How SPA Works
The mechanics behind systematic, risk-gated DeFi yield management.
What SPA Does
SPA is an autonomous yield optimizer for onchain stablecoins. It runs daily scans of whitelisted DeFi lending protocols, applies a deterministic risk gate to every proposed allocation, and rebalances between protocols when drift exceeds a threshold. There is no discretionary human override in the execution path. There is no LLM in the execution path.
What SPA Does Not Do
Daily APY Scan
Every day, SPA executes a full scan of supported protocols to collect current yield data.
Data Sources
Primary
DeFiLlama Yields API
Aggregated APY/TVL across all supported pools
On-Chain
Protocol adapters (read-only)
Direct contract reads for position verification
Oracles
Chainlink / Pyth price feeds
Collateral checks for Aggressive strategy
Logging
All data sources logged per cycle
Full audit trail in timestamped JSON
Protocol Adapters
Each supported protocol has a dedicated read-only adapter. Adapters normalize APY, TVL, and pool metadata into a standard format for the allocation engine. No write access at scan stage.
CURRENT ADAPTER REGISTRY — 35 adapters (sample shown)
Evidence Levels (L0–L6)
Every yield number on this site carries an evidence level — how much real confirmation stands behind it. Below L2, no APY reaches any public surface. Paper is L3 and is always labelled "paper", never "live". This is the canonical scale (ADR-YL-009); other pages only use the tags.
Where SPA is now: our evidenced paper track is honest L3 ("paper-tracked over N evidenced days"), not L4+. The conservative "~3.3% real" baseline and the Paper APY on /track-record are L3 numbers, labelled "variable · not guaranteed". We never call paper "validated" or "live-tested"; promotion is one level at a time.
RiskPolicy v1.0 Parameters
A deterministic, hard-coded gate. No human, no AI, no strategy can override these parameters.
| Parameter | Value |
|---|---|
| TVL floor | $5M minimum per pool |
| Per-protocol cap (Tier 1) | 40% of portfolio |
| Per-protocol cap (Tier 2) | 20% of portfolio |
| Tier 2 total cap | ≤50% of portfolio |
| Cash buffer | ≥5% |
| APY lower bound | 1% (avoid dust yields) |
| APY upper bound | 30% (sanity cap) |
| Kill switch — SOFT (peak drawdown) | ≥5% → de-risk (halt new, no liquidation) |
| Kill switch — HARD (peak drawdown) | ≥10% → all to cash |
Risk-Gated Rebalancing
Strategy Tournament
Multiple strategies run in parallel and are evaluated on risk-adjusted performance. Tournament output is advisory-only — all allocation decisions still go through RiskPolicy gate. No strategy can override the gate.
Participants
Multiple strategies run in parallel on paper
Scoring
Risk-adjusted APY (Sharpe-weighted)
Promotion
Top performers promoted to paper track record
Current Leader
Balanced strategy — current tournament winner
Strategy Profiles
SPA offers three strategy profiles: Conservative, Balanced, and Aggressive. Each has a distinct risk/return target and validation status.
Paper tracked since June 22, 2026 (evidenced anchor). Verified performance data available on dashboard.
Target profile defined. Paper tracking not yet started. No historical performance data.
Supported Assets
SPA operates exclusively with stablecoins. Minimum asset set for private MVP:
USDT
- TRC-20 (Tron)
- ERC-20 (Ethereum)
USDC
- ERC-20 (Ethereum)
Compounding & Logging
GoLiveChecker
29 criteria must pass before live deployment.
Failing criteria visible in Dashboard → GoLiveChecker tab.
Go-live target: July 2026
Versioning & Changelog
Capital Preservation Algorithm (CPA)
CPA is a layered validation framework that governs how strategies move from research to live capital. Every step requires verifiable point-in-time evidence — no retrofitted data, no hindsight.
What is Point-in-Time Evidence?
SPA uses only data that would have been available on the simulation date. A protocol launched in 2023 is invisible to a 2022 backtest — no exceptions. This eliminates look-ahead bias, the most common cause of inflated backtest returns in DeFi.
Example: GMX launched Q3 2021. Any simulation dated before that launch treats GMX as non-existent, even if historical APY data was later published. The whitelist is date-gated per protocol. Each entry in the point-in-time whitelist carries a launch_date field — the backtest engine rejects any protocol whose launch date is after the simulation date.
4-State Gate System
Backtest Gate
Point-in-time simulation over historical on-chain data. CLEAN_INCLUDED sources only. No forward-looking inputs.
Pre-Paper Gate
Strategy tested on live market data with zero capital at risk. Formal approval before advancing.
Paper Trading
Full strategy run on virtual capital. Requires ≥ 30 clean evidence points via PaperEvidenceTracker.
Live Capital
Own capital only. Requires GoLiveChecker all 29 criteria pass (currently 28/29) + 7+ consecutive days READY + Owner manual review (ADR-002).
Source Quality Framework
| State | Meaning | Examples |
|---|---|---|
| CLEAN_INCLUDED | Full historical data, on-chain verifiable | Aave V3, Compound, Morpho |
| PENDING | Data found, under audit | Morpho Steakhouse |
| RESEARCH_ONLY | No historical data, model-based projections | GMX exposure, Concentrated LP |
| SOURCE_NEEDED | No data source identified yet | Gold proxy, BTC pools |
Why 86.97% Cash Drag?
In strict point-in-time mode across 2022–2026, most DeFi protocols had not yet launched or lacked verifiable on-chain data at the simulation date. CPA does not extrapolate or assume retroactive availability.
The result is a high cash drag — capital that cannot be deployed to a CLEAN_INCLUDED protocol is held as USDC. This is intentional: it is better to acknowledge limited deployment opportunity than to overstate historical returns using protocols that were not yet verifiable at the simulation date.
A 13% average deployed allocation (86.97% cash) is the honest answer for the 2022–2026 period. As CPA sources are audited and promoted from PENDING to CLEAN_INCLUDED, the deployable universe grows and cash drag falls.
Research vs Production Strategies
RS-001 and RS-002 are tagged RESEARCH_ONLY. Their APY figures are forward projections based on model assumptions — not verified historical data. They cannot be promoted to paper trading until a CLEAN_INCLUDED data source is identified and audited. Evidence points from research days score 0.3 (vs 1.0 for clean days).
Production strategies use only CLEAN_INCLUDED data sources. Each allocation target has on-chain verifiable APY history and a confirmed launch date. These are the strategies eligible for paper trading evidence accumulation (1.0 point per clean day, 1.5 per extreme-market day) and eventual live capital deployment.
Continue your due diligence
Personal research project in paper validation — not investment advice, not a regulated service, not raising capital; results are simulated. Full disclaimer & risk disclosure →