DeFi allocator · T1 strategy backtest · Jan 1 – Apr 30, 2026

$1M across six lending protocols

A block-by-block backtest on recorded on-chain rates: a simulated $1M, 864k Ethereum blocks, 321 switches and $67.97 of gas over the whole period. The dot on the lanes shows where our portfolio sits at every moment; the lower panel tracks its value against the alternative strategies and passive holds.

DateJan 1block 24,204,180

Portfolio is in——

Portfolio value$1,000,000+$0 vs Aave hold

Switches0of 321

day 0

Our portfolio follows T1, a gas-aware threshold rule: it moves to another venue only when the expected extra yield over the expected holding time beats the gas cost. T2 is Ornstein–Uhlenbeck optimal stopping; the greedy router always jumps to the best spot rate; a passive hold keeps $1M in one venue for the whole period.
Source: equity_t1_threshold.parquet, the simulated position state in each of 863,999 blocks, #24,204,180 to #25,068,178, replayed on recorded per-block rates. No real funds were deployed. Passive holds are recomputed block by block from the same rates, with compounding.
T3 (Cox hazard model) is identical to T1 in this replay because it fell back to T1, so it has no separate line. Its honest out-of-sample test is a separate experiment: −5.97 bp, 0 of 5 windows.