Executive Summary
Uniswap v3 displays gross fee APY. It does not display impermanent loss. The gap between those two numbers is where most liquidity provider (LP) capital goes to die.
The academic record on this is not ambiguous. Loesch et al. (2021) analyzed 17 Uniswap v3 pools covering 43% of total value locked during the protocol's first five months of operation. LPs earned $199.3 million in fees. They incurred $260.1 million in impermanent loss. Net result: a $60.8 million loss compared to simply holding the underlying assets. Over 80% of pools showed LP underperformance versus holding. (Loesch et al., arXiv:2111.09192)
This paper presents a different question: not whether most LPs lose money — they do — but whether there is a structured approach to concentrated liquidity provision that generates real, continuous fee income while keeping impermanent loss in a range a deliberate investor can accept and quantify. Our answer is conditional: yes, under a specific combination of asset tier, chain, fee tier, regime gate, and range method. Outside those conditions, you are reproducing the same gross+/net-negative trap Loesch documented.
Key findings:
- B-tier DeFi coins (LINK, AAVE, UNI, ARB, AVAX) on Arbitrum at the 0.30% fee tier, with RXI regime gating, generate approximately 10–17% net fee income on LP capital in ranging market regimes in simulation. Net APY on total capital: +15.2% [95% CI: +14.3%, +16.1%] in a 5,000-iteration Monte Carlo simulation calibrated to 2025–26 live data — not a historical backtest.
- This essentially ties the 50/50 HODL benchmark (+15.7%) — within the confidence interval. The strategies are statistically indistinguishable in mean expectation. We are not claiming LP beats holding.
- The income profile is meaningfully different from HODL in character. LP returns arrive as continuous realized fee income; HODL returns arrive as unrealized price appreciation. This distinction is meaningful for income-oriented investors and advisors.
- CoinRoc's rating system inversely predicts LP suitability. A-rated coins (ETH, BTC, SOL) are the worst LP candidates. B-rated DeFi coins are the sweet spot. This finding inverts investor intuition and is a novel contribution to the published literature on concentrated LP selection.
- Ethereum mainnet is structurally unviable for LP at retail capital levels. Post-Dencun, annual gas on Ethereum mainnet runs $1,047 for a standard operational cycle. Arbitrum: $7.12. Solana: $0.04. Chain selection matters more than fee tier selection for small investors.
- All results are from a 5,000-iteration Monte Carlo simulation calibrated to 2025–26 live pool data. Tick-level historical validation is required before any investor-facing point estimates are appropriate. These are simulation results, not empirical backtests.
1. The Problem: What the UI Shows and What It Omits
When you open Uniswap v3 and look at a pool, you see a gross fee APY. For a typical WETH/USDC 0.05% pool on Ethereum mainnet, that number on the date of this analysis was 23.0%. For a 0.30% fee-tier pool on a mid-cap DeFi token, you might see 30–45%.
The formula behind that number is straightforward: Fee APY = (7d_volume / TVL) × fee_tier × 52
What the formula omits:
- Impermanent loss — the largest single cost, amplified by concentrated range width
- Gas and transaction costs — size-independent dollar costs that structurally disadvantage small positions
- JIT liquidity extraction — sophisticated actors who add concentrated liquidity in the block before a large swap and remove it after, capturing the fee (Kaiko Research: JIT attacks reached 10% of attacked trades in studied USDC pools)
- TVL dilution — new LPs entering the pool reduce your share of swap volume
- Loss-Versus-Rebalancing (LVR) — Milionis et al. (2022) formalized the structural cost every AMM LP pays when arbitrageurs trade against a stale price. LVR scales quadratically with volatility. It is not avoidable by better range selection.
2. The Impermanent Loss Mechanism in Concentrated Liquidity
In a Uniswap v2 full-range pool, if the price of a token doubles, the LP incurs approximately 5.7% IL versus holding. In Uniswap v3 concentrated liquidity, the LP's capital is deployed only within a defined price range [Pa, Pb]. The concentration factor amplifies both fee earnings and IL.
The exact Uniswap v3 concentrated LP formula:
Range: [Pa = entry × 0.70, Pb = entry × 1.40] (−30% lower band, +40% upper band)
When Pa ≤ P ≤ Pb (in range):
x = L × (1/√P − 1/√Pb) [crypto units held]
y = L × (√P − √Pa) [USDC units held]
LP value = x × P + y
When P < Pa (below range, fully in crypto):
LP value = L × (1/√Pa − 1/√Pb) × P
When P > Pb (above range, fully in USDC):
LP value = L × (√Pb − √Pa) [value fixed in USD]Our IL stress-test results using the tight range [−30%, +40%]:
| Price scenario during LP period | IL vs. HODL | In range? |
|---|---|---|
| Sideways (flat) | 0.00% | Yes — full fee accumulation |
| +10% drift | −2.3% | Yes — well covered by fee income |
| +30% sustained move | −16.6% | Yes (approaching upper bound) |
| −20% drift | −11.4% | Yes — covered by fees in ranging regime |
| −30% exit lower band | −29.4% | No — RXI gate should signal exit first |
| −50% crash | −46.7% | No — tail risk; RXI reduces frequency, does not eliminate |
| −50% crash break-even gross APY | ~234% — structural impossibility | |
3. The Simulation: What We Measured
Our analysis used a 5,000-iteration Monte Carlo simulation over 52 weeks, comparing three strategies across four asset tiers and three chains:
- Strategy A: Always-in passive LP
- Strategy B: RXI-gated LP (deploy only when RXI regime signal is ranging, exit to USDC on trending signal)
- Strategy C: 50/50 HODL (the benchmark LP investors should compare against)
Capital model: 3-3-3 allocation — 2/3 of total capital deployed in LP, 1/3 held as USDC reserve. Live calibration: GeckoTerminal API and DefiLlama data as of 2026-06-15.
Primary Results — Arbitrum (minimum $2,000 deployed LP capital)
| Tier | Representative coins | Always-In | RXI-Gated | 50/50 HODL | RXI vs. HODL | Loss Freq (RXI) |
|---|---|---|---|---|---|---|
| A-tier | ETH, BTC, SOL | +4.9% [+4.4, +5.4] | +3.0% [+2.7, +3.4] | +11.7% | −8.6pp | 41.7% |
| B-tier | LINK, AAVE, UNI, ARB, AVAX | +21.3% [+20.2, +22.4] | +15.2% [+14.3, +16.1] | +15.7% | −0.4pp (within CI) | 35.0% |
| C-tier | DOT, ADA, ATOM, NEAR, INJ | −23.3% [−24.3, −22.2] | −24.5% [−25.4, −23.6] | +21.6% | −46.1pp | 79.7% |
| D-tier | BONK, WIF, PEPE | −53.6% [−54.1, −53.0] | −60.8% [−61.3, −60.3] | +41.4% | −102.2pp | 97.7% |
95% confidence intervals on the mean. CI = 1.96 × σ / √n, n=5,000. All figures: net APY on total deployed capital, annualized.
The B-tier Arbitrum result of +15.2% RXI-gated versus +15.7% HODL is a −0.4 percentage point gap that sits inside the confidence interval. These strategies are not statistically distinguishable in mean expectation at 95% confidence.
The Solana CLMM result is marginally better due to near-zero gas. RXI-gated B-tier Solana: +15.8% [95% CI: +14.9%, +16.7%].
4. The Four CoinRoc Optimization Levers
4a. RXI Regime Gating — Measured Tail-Risk Reduction
RXI™ (Regime eXecution Intelligence) — built on a Mamdani fuzzy-inference engine — uses Hurst exponent, ADX, velocity Z-score, and drawdown-from-peak inputs to classify market regime. In LP context: deploy concentrated liquidity in ranging conditions (Hurst < 0.5, ADX < 25); exit to USDC in trending conditions (Hurst > 0.6, ADX > 25). The investor executes manually. CoinRoc is an advisory tool.
Important counterintuitive result for B-tier: RXI gating on B-tier coins reduces mean return by −6.1pp versus always-in (+21.3% always-in vs. +15.2% RXI-gated). This is expected and correct. At the 0.30% fee tier, even trending-week LP positions generate positive expected fee income in mean expectation. The RXI gate's value for B-tier is risk management, not return generation — it reduces loss frequency (42% → 35%) and tail exposure in exchange for lower mean return.
Compliance-safe framing: "CoinRoc's RXI regime engine helps investors avoid deploying concentrated liquidity during trending markets where impermanent loss tends to exceed fee income. This reduces loss frequency and improves the shape of outcomes — not the average return — compared to always-on passive LP."
4b. The Rating-Inversion Pair-Selection Rule — Publishable Novel Finding
The most counterintuitive finding of this research: CoinRoc's quality rating system inversely predicts LP suitability. The best grid-trading coins are the worst LP candidates.
The mechanism driving the inversion:
- A-tier coins (ETH, BTC, SOL) sit in 0.05% fee-tier pools — designed for deep, liquid, low-volatility pairs. The fee is 6x lower than the 0.30% tier. Deep TVL keeps Vol/TVL ratios moderate. Simultaneously, A-tier HODL benchmark benefits heavily from crypto's long-term appreciation trend. Result: A-tier LP generates +3.1% RXI-gated vs. +10.9% HODL — a −7.8pp structural deficit.
- B-tier coins (LINK, AAVE, UNI, ARB, AVAX) sit in 0.30% fee-tier pools with real organic trading volume. The 0.30% tier generates 6x the gross fee income per unit of Vol/TVL. B-tier volatility is meaningful but not extreme. This is the LP economic sweet spot.
- C and D-tier coins have volatility that overwhelms any fee tier. Loss frequencies of 79.8% and 97.7% respectively.
The practical heuristic: When selecting LP pairs: (1) Filter to B-tier candidates; (2) Prioritize coins with established DeFi protocol utility and organic trading volume; (3) Target 0.30% fee-tier pools on Arbitrum or Solana; (4) Exclude A-tier majors and C/D-tier.
4c. Fibonacci Range Selection — Targeting Price-Density Zones
Time in range is the dominant driver of LP profitability. A position spending 80% of weeks in-range at 30% gross APY earns 24% annualized gross. A position spending 40% of weeks earns 12%.
CoinRoc's fibonacciGrid.ts module performs logarithmic swing identification from daily candles, auto-detecting the historical swing low and swing high over the previous major price cycle. It places LP range boundaries at the 0.618 retracement (lower band) and the 1.272 extension (upper band) — zones where price historically spends the most time. These are zones of historical price density, which is the correct heuristic for maximizing time in range.
Directional evidence suggests the medium range [−40%, +60%] may outperform the tight range [−30%, +40%] for B-tier due to better IL absorption on price moves exceeding ±30%. Formal range-width optimization is scoped as a future research task (RES-LPIS-RANGE-WIDTH-01).
4d. Pre-Set Contingency Ranges — Conditional Income on Breakdown
The 3-3-3 capital model holds 1/3 of capital as USDC reserve outside the primary LP position. When price falls below the primary range's lower bound (Pa), the primary position converts to 100% crypto and earns no fees. The contingency range deploys this reserve capital in the band below the primary lower band, earning fees during a slow consolidation below the primary range.
What it does: Earns fees during a slow drift below the primary band — a common crypto pattern (step-down followed by sideways consolidation).
What it does not do: Hedge against a crash. A −40% to −60% price drop in days traverses the contingency range in days — fee income from a 3-day transit at 30% gross APY is approximately 0.25% of capital. Negligible against a −30% to −40% IL event on the primary position.
Critical gating requirement: The contingency range must be independently gated by the RXI. Deploy only when RXI confirms ranging regime despite price exiting the lower band. Do NOT auto-trigger on price movement alone.
5. The B-Tier Income Decomposition
The specific scenario where the LPIS thesis holds: B-tier coins, Arbitrum or Solana, 0.30% fee tier, RXI-gated.
| Component | Annual % of LP capital (deployed weeks) |
|---|---|
| Gross fee income | ~28–35% |
| Impermanent loss drag (ranging weeks) | −12 to −18% |
| Gas drag (Arbitrum post-Dencun) | −0.3% |
| Net LP return on LP capital (deployed weeks) | ~10–17% |
| Net LP return on total capital (2/3 deployed, 3-3-3 model) | ~7–11% |
Source: 5,000-iteration Monte Carlo simulation, calibrated to 2025–26 live data. Not historical backtest results.
The income-to-IL ratio in the B-tier ranging deployment: approximately 1.6–2.4x. Fee income is 1.6 to 2.4 times the IL drag during the weeks when deployed in the right regime.
6. The Income Profile Distinction — Why It Matters for Advisors
The claim that RXI-gated B-tier LP on Solana "ties HODL" requires a critical qualifier. The strategies produce similar mean returns in simulation — but through fundamentally different mechanisms.
50/50 HODL returns: Price appreciation on the 50% crypto component, recognized only when liquidated. Unrealized until sale. Subject to the full volatility path of the underlying asset.
LP fee income returns: Realized continuously as fees are collected. Accrues in the form of USDC (and crypto) deposited to the pool position, collectible at any time. No single "exit event" required to recognize the return.
For an investor who values ongoing distribution over terminal lump-sum gains, that distinction is the case for LP — not a return-superiority claim.
7. Chain Economics — The Decisive Variable
| Chain | Min capital for gas < 5% of net fees | Annual gas | RXI-Gated B-tier Net APY | Verdict |
|---|---|---|---|---|
| Arbitrum | $949 (practical minimum: $2,000) | $7.12 | +15.2% [+14.3, +16.1] | RECOMMENDED |
| Solana CLMM | $6 (practical minimum: $500) | $0.04 | +15.8% [+14.9, +16.7] | FAVORABLE — separate integration required |
| Ethereum Mainnet | $139,600 | $1,047 | +9.8% [+8.9, +10.7] | NOT VIABLE at retail scale |
Ethereum mainnet is not viable below $25,000 deployed capital. No content, advisor deck, or product description referencing CoinRoc's LP strategy should mention Ethereum mainnet without explicitly naming Arbitrum or Solana as the required execution chains.
8. What the Data Support and What They Do Not
What is supported:
- B-tier DeFi coins in 0.30% fee-tier pools on Arbitrum or Solana CLMM generate +15.2% to +15.8% RXI-gated net APY in simulation — within the confidence interval of the 50/50 HODL benchmark of +15.7%. These strategies tie in mean expectation.
- LP returns are continuous realized fee income; HODL returns are unrealized price appreciation. A genuine structural distinction.
- CoinRoc's rating system inversely predicts LP suitability: A-rated coins are the worst LP candidates; B-rated coins at the 0.30% fee tier are the LP sweet spot. Novel, publishable finding.
- RXI regime gating reduces B-tier loss-frequency weeks from ~42% to 35.0%. The RXI gate is a risk management tool, not a return generator.
- Post-Dencun, Arbitrum gas = $7.12/year. Chain selection is structurally decisive for retail-scale LP.
What is not supported and must not be claimed:
- "LP beats HODL." The data say LP ties HODL within the margin of error at best. No headline should use "beats."
- Any APY point estimate without its confidence interval. Always cite as "+15.2% net APY [95% CI: +14.3%, +16.1%] in simulation."
- "IL is manageable" or "IL is eliminated." IL is real and material.
- "This is a passive income strategy." LP at this level requires RXI monitoring, periodic fee collection, reposition decisions, and exit discipline.
- Any Ethereum mainnet recommendation at retail capital levels.
- That tick-level historical backtesting has been completed. It has not. Tick-level historical validation via Dune Analytics is required before any investor-facing net APY point estimates are appropriate.
9. Implementation Parameters
Minimum viable position (Arbitrum): Total capital $3,000 (3-3-3 model: $2,000 LP + $1,000 USDC reserve). Gas drag at $2,000 deployed: 0.47% annually — negligible.
Simulation-supported asset set: Based on simulation results, B+/B/B− rated coins with established DeFi protocol utility — including coins such as LINK, AAVE, UNI, ARB, and AVAX — exhibited the most favorable LP income-to-IL ratios in the tested period. Readers should conduct their own analysis; these findings reflect modeled conditions, not a trading recommendation.
Fee tier: 0.30% pools on Arbitrum. Not 0.05% (A-tier economics), not 1.00% (thin pools, high JIT exposure).
Range selection: Fibonacci-derived range anchored to previous cycle swing low (lower band at 0.618 retracement) and swing high (upper band at 1.272 extension). Medium range [−40%, +60%] may outperform tight range [−30%, +40%] for B-tier. Formal optimization pending (RES-LPIS-RANGE-WIDTH-01).
RXI monitoring: Check at minimum weekly. Exit to USDC when RXI transitions to trending (Hurst > 0.6, ADX > 25). Re-enter when RXI returns to ranging. Not a set-and-forget position.
Contingency range: Deploy below the primary band only when RXI confirms continuing ranging regime. Do not auto-trigger on price movement alone.
10. Relationship to Published Literature
Loesch et al. (2021) (arXiv:2111.09192): Established that LP aggregate P&L is net-negative versus holding in the studied period. Our simulation's fees/IL calibration (1.63) is more favorable than Loesch's measured 0.765, reflecting a less extreme market period. This means our results are optimistic relative to the most volatile historical period.
Milionis et al. (2022) (arXiv:2208.06046): Formalized LVR as the structural adverse-selection cost all AMM LPs pay against arbitrageurs. LVR scales quadratically with volatility. Our simulation embeds LVR implicitly in the IL model.
CrocSwap (2023): Found that large swaps (top 25% by notional) are the structural profit extractors against LP positions. Our model applies a JIT haircut per tier (6% A-tier, 4% B-tier) as a simplified representation.
Our contribution: Regime-gated LP analysis across four asset tiers with the RXI signal as the gate, combined with the novel finding that quality-rating tier inversely predicts LP suitability. No published work has connected an external quality-rating system to pool-tier selection via this mechanism.
Risk Disclosure and Simulation Notice
This document presents the results of Monte Carlo simulation analysis (5,000 iterations, calibrated to live GeckoTerminal and DefiLlama pool data as of 2026-06-15) and does not constitute investment advice, a recommendation to buy or sell any security or digital asset, or a solicitation of any investment. CoinRoc is a software tool; it is not a registered investment adviser, broker-dealer, or financial planner.
Simulation results are not backtested historical performance. All quantitative results were produced by means of a stochastic model designed with the benefit of calibration to current (not historical) market data. Simulated results do not represent actual historical LP performance and are not a guarantee of future results. Tick-level historical validation of these results has not been completed; such validation is required before any investor-facing net APY point estimates are appropriate. Actual results will vary materially based on market conditions, capital size, execution timing, pool liquidity, gas costs, and other factors not modeled in this simulation.
Past simulated performance is not indicative of future results. The value of crypto assets and LP positions can decline to zero. Fee income from LP positions is not guaranteed and depends on ongoing market activity in the relevant pool. Impermanent loss is a real and material cost of providing liquidity in automated market makers.
The RXI regime engine produces an advisory signal only. It does not predict future price movements and does not guarantee that deploying or exiting LP positions based on its signal will produce positive returns. CoinRoc does not hold user funds, does not execute transactions on behalf of users, and does not manage any pooled investment vehicle.
Regulatory status: Digital assets, including those referenced in this document, may be deemed securities under applicable law (including the Howey test as applied by the SEC). This document does not constitute legal, tax, or regulatory advice. Consult your own financial, legal, and tax advisors before making any investment or business decision.
CoinRoc compliance review completed 2026-06-15 (Matlock). External legal counsel review recommended before wide distribution, particularly regarding specific coin naming in Section 9 and Howey status of LINK/AAVE/UNI/ARB/AVAX.
LANDO-LPIS-RESEARCH-01 — Lando, Senior Content Writer & Strategist, Yodacom AI Team. Source documents: RES-LPIS-OPTIMIZE-01, RES-LPIS-EMPIRICAL-01, DARTH-AMM-LPIS-01 (Han Kessel & Darth, 2026-06-15). No new simulations run for this article. Matlock compliance edits applied 2026-06-15.