Figure 5 — Yodacom Research · LPIS Series · 2026-06-15
RXI Regime Gating: Loss Frequency Reduction by Tier
% of weeks with a net loss (fee income minus IL minus gas < 0) · Always-in passive LP vs. RXI-gated LP · Arbitrum · 5,000-iteration simulation · RXI parameters: 40% ranging frequency, 15% Type I error
0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% % of simulated weeks with net loss (fee income − IL − gas < 0) Always-In Passive LP RXI-Gated LP ~44% 41.7% A-tier ETH, BTC, SOL 0.05% fee tier −2.3pp lift RXI gate mainly improves tail shape for A-tier, not mean loss frequency. Mean lift: +1.6–3.7pp. ~42% 35.0% B-tier LINK, AAVE, UNI, ARB, AVAX 0.30% fee tier B-tier: Best LP outcome in simulation −7.0pp lift Largest measured reduction. However, RXI-gated mean APY is −6.1pp vs. always-in (RXI trades mean return for tail-risk reduction) ~82% 79.7% C-tier DOT, ADA, ATOM, NEAR high volatility −2.3pp lift RXI gate cannot fix structural IL problem ~98.5% D-tier loss frequency is structurally ~98% regardless of RXI gating — no viable LP scenario D-tier BONK, WIF, PEPE — NO-GO
RXI gate lift sensitivity (A-tier Arbitrum) — measured range from simulation
Ranging Frequency Type I Error Rate Always-In APY RXI-Gated APY Lift
30% ranging 10% false+ −0.5% +3.2% +3.7pp
30% ranging 20% false+ −0.5% +2.4% +2.9pp
40% ranging (baseline) 10% false+ −0.5% +2.8% +3.3pp
40% ranging (baseline) 20% false+ +0.4% +2.5% +2.1pp
50% ranging 10% false+ +1.0% +2.8% +1.8pp
50% ranging 20% false+ +1.0% +2.6% +1.6pp
Measured lift range: +1.6pp to +3.7pp on total portfolio return (A-tier, Arbitrum). This is a risk-reduction benefit, not a return-generation benefit. The RXI gate reduces the probability of catastrophic IL weeks — it does not convert LP into an alpha-generating strategy. Replaces the prior qualitative estimate of +15–35pp, which overstated the effect.
Simulation disclosure: Loss frequency = % of simulated weeks where (fee income − IL − gas) < 0. RXI gate parameters: 40% ranging frequency (40% of weeks classified as ranging/mean-reverting), 15% Type I error rate (15% of trending weeks misclassified as ranging), 12% Type II error rate (12% of ranging weeks missed). Always-in A-tier and C/D-tier loss frequencies are implied from the simulation distribution; B-tier always-in loss frequency (~42%) is stated explicitly in RES-LPIS-OPTIMIZE-01, Section 2a. RXI gate lift of −6.1pp mean return for B-tier (RXI-gated +15.2% vs. always-in +21.3%) reflects the opportunity cost of sitting out fee-positive trending weeks at the 0.30% tier — this is a correct and expected finding, not an anomaly. Sensitivity table from RES-LPIS-EMPIRICAL-01, Section 4, Table 4. All Arbitrum chain. All figures represent 5,000-iteration Monte Carlo simulation outputs, not historical backtest results. Tick-level historical data validation pending (Priority 1, RES-LPIS-EMPIRICAL-01). Not a projection of future performance.