Balancer: A Non-Custodial Portfolio Manager, Liquidity Provider, and Price Sensor

Paper title A non-custodial portfolio manager, liquidity provider, and price sensor.
Authors Fernando Martinelli and Nikolai Mushegian
Version and date v2019-09-19
Project Balancer
License note The paper says the Ethereum smart-contract system is GPL-3.0 licensed; it does not separately state a license for reproducing the paper text.
Official source Read the original Balancer whitepaper (PDF)
Disclaimer: This page summarizes and analyzes a technical paper for education. It is not financial advice or a verbatim copy. The original work remains with its rights holders.

A non-custodial portfolio manager, liquidity provider, and price sensor is Balancer’s 2019 design paper for an automated market maker (AMM) whose pools can hold multiple tokens at chosen weights. Its central idea is a weighted product invariant: instead of requiring every pool to hold two assets at 50/50, the pool’s rules can assign different target weights. The paper explains the original math and early contract design; it does not describe later Balancer versions or today’s governance status.

Original paper Original paper
Directories Ethereum Whitepapers & Research

What the paper is trying to do

The authors frame a pool as a portfolio whose balances change as traders exchange assets. In a traditional index fund, a manager may trade to maintain target exposures. A Balancer pool instead uses an on-chain pricing rule and traders’ arbitrage to move the reserves toward the pool’s configured relationship with outside prices. This is an AMM mechanism, not a promise that liquidity providers (LPs) will earn a profit or hold a fixed dollar value.

The weighted invariant

The paper defines the pool’s value function as V = ∏ BᵢWᵢ, where Bᵢ is the balance of token i and Wᵢ is its normalized weight; the weights add up to 1. Swaps follow the invariant, subject to the pool’s trading-fee rules. A two-token pool with equal weights resembles the familiar constant-product design. Unequal weights, such as 80/20, change the pool’s price curve and the exposures LPs choose to hold. The original Balancer paper derives the invariant and its spot-price relationship.

The pool does not autonomously buy or sell assets when market prices move. Traders compare its quote with outside markets; arbitrage can then change the pool balances until prices are closer. The direction and size of those trades depend on the pool and market. Large trades move the price, and the result can differ from simply holding the same tokens outside the pool.

What LPs take on

LPs receive pool tokens representing a share of pool liquidity. Fees may compensate them for facilitating trades, but fees are variable and do not remove market risk. A weighted mix can lose value relative to holding the assets directly when their prices move differently. LPs also face smart-contract, token, oracle or integration, liquidity, and pool-manager risks where a pool has configurable controls. The 2019 paper focuses on its invariant and contract mechanics; it is not a full risk assessment for every later pool.

What changed in V2 and V3

Balancer V2 later introduced a shared Vault that holds and accounts for assets while individual pool contracts calculate swaps, joins, and exits. The Vault can route multi-hop swaps through one transaction while keeping each pool’s balance accounting separate. This architecture is distinct from the original paper’s first-release implementation. Balancer’s V2 Vault documentation describes the split between asset accounting and pool logic.

V3 is another later system, not part of the 2019 paper. Balancer’s current V3 materials describe a Vault, routers, pools, and optional hooks that can add behavior around pool operations. Pool and hook behavior is configuration-specific; it should not be assumed from the whitepaper alone. The official Balancer V3 overview outlines this architecture. Balancer’s 2026 governance changes also make old statements about ongoing BAL emissions, veBAL fee rights, and voting weight outdated; the current governance documentation describes the revised state.

Wind-down and recent incidents: dated context

As of this page’s October 28, 2026 update, BAL holders had approved BIP-928 for an orderly wind-down. The plan schedules a withdrawals-only transition on October 30 for applicable pools; certain V3 pools with approved partner extensions may continue until November 30. The proposal says withdrawal routes remain available, but pool treatment varies by contract and deployment, so users should check the current instructions for their specific pool. Balancer’s September 29 vote update confirms the decision and dates.

A separate proposal, BIP-930, discusses recovery and reimbursement after a Balancer V1 exploit on August 31, 2026. It concerns affected V1 liquidity providers and recovered assets; it is separate from BIP-928’s plan for the DAO treasury. The forum post is a proposal and should not be read as proof that every affected user has been repaid. See BIP-930’s description of the V1 incident and proposed recovery process.

How to read the paper today

The paper remains useful for understanding weighted-pool math and how an AMM can represent a multi-asset portfolio. Keep its 2019 design separate from V2/V3 implementation choices, later token governance, and the current wind-down. Its invariant explains how a pool quotes trades; it does not establish that a pool is safe, profitable, or still operating.

Related Terms

See Also

Sources