NEAR Hackathon · Investment Thesis track. Testing whether NEAR Intents
cross-chain volume is a better leading indicator of real network usage than the NEAR token price.
Data as of 2026-11-03 (rebuilt 2026-11-04). Every number below is
computed by rebuild.py from the saved CSVs in data/ — except the
clearly labelled first-party claims in Panel 5, which are attributed, not verified.
Data caution: the final fees day (2026-11-03) is flagged partial (fees below same-day revenue — impossible for a complete day) and is excluded from computed trailing-30d sums.
Intents volume has grown into a $100M+/day flow; the NEAR price rose over the same year but with sharp swings volume does not explain — daily volume vs price correlation is r = 0.50 (n = 365 days), rising to r = 0.78 for 30-day rolling volume, so the two are related in trend but volume is not a mirror of price.
Sources: https://api.llama.fi/summary/dexs/near-intents?dataType=dailyVolume and the coins.llama.fi NEAR price chart
(exact URL in data/key_stats.json). Code: rebuild.py → data/intents_daily_volume.csv,
data/near_daily_price.csv.
A single month of Intents flow (trailing-30d volume) is many times larger than the entire value locked on the NEAR chain — NEAR increasingly looks like a settlement pipe, not a vault.
Sources: https://api.llama.fi/summary/dexs/near-intents?dataType=dailyVolume (rolling 30d sum computed in rebuild.py)
and https://api.llama.fi/v2/historicalChainTvl/Near. Code: rebuild.py →
data/intents_daily_volume.csv, data/near_chain_tvl.csv.
Cross-correlation peaks at lag +2 days for 7-day rolling volume (r = +0.70) and lag +3 days for raw daily volume (r = +0.54), where a positive lag means volume was measured before price. Levels cross-correlation peaks at or near lag 0 — volume and price move together rather than volume going first; on this data volume is a coincident, not leading, indicator.
Robustness check (levels are trend-inflated, so this matters more): correlation between the 7-day % change in 7-day rolling volume and forward price returns — 1d: r = +0.11, 7d: r = -0.01, 14d: r = +0.03, 30d: r = +0.05. Weak or negative values here mean volume growth does not reliably precede price gains even if the levels curves peak at a positive lag.
Sources: https://api.llama.fi/summary/dexs/near-intents?dataType=dailyVolume and the coins.llama.fi NEAR price chart.
Code: rebuild.py → data/leadlag.csv; full method, best lags and forward-return
tests in data/key_stats.json under leadlag.
Of every $1 of gross Intents fees, about $0.34 reached NEAR as captured revenue over the last 30 complete days (capture ratio 34.2%; all-time 16.1%) — an effective revenue take of 0.054% of volume and an annualised revenue run-rate of $25.6M: the flow is large, the slice is thin.
Mechanism: Since 2026-02-23, NEAR's captured Intents revenue is used to buy back $NEAR on the open market (not burned), per the DefiLlama NEAR Intents methodology: revenue is NEAR's net captured revenue, measured as NEAR/wNEAR swept into its revenue wallets, and since 2026-02-23 that captured revenue funds open-market buybacks. Source: https://defillama.com/protocol/near-intents
Sources: https://api.llama.fi/summary/fees/near-intents?dataType=dailyFees and https://api.llama.fi/summary/fees/near-intents?dataType=dailyRevenue;
buyback mechanism per the DefiLlama NEAR Intents methodology page linked above.
Code: rebuild.py → data/intents_daily_fees.csv (partial day flagged),
data/intents_daily_revenue.csv; ratios in data/key_stats.json under value_capture.
| Chain | Intents TVL (latest) | Share |
|---|---|---|
| Near | $109.0M | 53.9% |
| Ethereum | $58.8M | 29.1% |
| Tron | $7.9M | 3.9% |
| Bitcoin | $6.8M | 3.4% |
| Binance | $5.3M | 2.6% |
| Arbitrum | $3.5M | 1.7% |
| Base | $2.9M | 1.4% |
| Ripple | $1.5M | 0.7% |
| Litecoin | $1.4M | 0.7% |
| Polygon | $1.4M | 0.7% |
Intents liquidity verifiably sits across 26 chains (largest: Near) — that footprint is reproducible from DefiLlama; the confidential-layer claims in the box are NEAR's own and are labelled, not verified.
Source: https://api.llama.fi/protocol/near-intents (chainTvls, latest point per chain).
Code: rebuild.py → data/key_stats.json under intents_tvl_by_chain.
None of these figures enter any computed statistic on this page.
| Scenario (volume vs current run-rate) | Annual volume | Gross fees | Captured revenue (= buyback $) | Captured rev. as % of NEAR market cap |
|---|---|---|---|---|
| Bear (1x) | $48.31B | $76.2M | $26.0M | 0.49% |
| Base (2x) | $96.62B | $152.5M | $52.1M | 0.99% |
| Bull (4x) | $193.24B | $304.9M | $104.1M | 1.98% |
Scenario arithmetic — NOT a price prediction. Base inputs are computed from data: annualised volume run-rate $48.31B, observed fee take rate 0.158%, observed capture ratio 34.2%. Only the volume multiplier is assumed (model_config.json). Market cap $5.26B from CoinGecko (fetched 2026-11-04).
Assumptions: model_config.json (edit the multipliers and re-run).
Outputs: data/model_outputs.json. Method and base inputs are recorded in that file.
See THESIS.md for the claim, the evidence for and against (including the lead-lag result), and what would change our mind.