On Bancor: The Design, The Flaw, and The Future

Introduction

There is a huge gap of misunderstanding on how Bancor works. If you think the protocol just mints BNT to pay for IL insurance, you can learn something here in 8 minutes. In this piece, I argue that B3 (Bancor version 3) is a brilliant design, though it did not address a key tail risk. It was a critical design flaw, but I believe it is fixable. A utility token in DeFi is very rare, but the Bancor B3 design offers huge potential to have just that. I believe it could come back in a similar form and be one of the most important protocols in DeFi.

What is Bancor?

From a trader’s perspective, Bancor is a DEX (decentralized exchange). They see prices on asset pairs and can find liquidity to transact. Traders pay fees for this service.

For LPs (liquidity providers), Bancor is a pseudo-credit union. Their principal is protected (in token terms) and they receive yield for staking. Bancor imposes constraints on liquidity (7-day cooldown period and exit fee), but not like a bond. LPs also have exposure to the network (the union, in this analogy).

For LPs, calling it a DEX would be erroneous. You do not deposit into dual or multi-asset liquidity pools, just single-asset staking. You do not have impermanent loss risk to a token or token basket, instead you have tail risk of the network failing (default). LPs share trading fees, not with a protocol treasury but with a burn mechanism that secures principal protection. The only similarity is that you receive your fees from providing your liquidity to traders.

A DeFi Utility Token: BNT

Bancor’s primary offering is yield and principal protection for token depositors. It uses a utility token called BNT (Bancor Network Token) to achieve those goals. As usage of the network grows, so should yield for BNT stakers. The yield generated aligns BNT holders with network growth while the mint-burn balances the supply.

The burns are small but occur on each transaction. Mints tend to be larger but occur less frequently. The framework is similar to how an insurance company receives small premiums over time to pay out a potentially large event in the future.

BNT can also be effectively burned when a token depositor withdraws their deposit. Meaning sometimes, principal protection (also called impermanent loss insurance) leads to effectively burning BNT not minting it. Mints occur to pay out principal protection when a deposited token outperforms BNT and does not have deep liquidity on the protocol.

The protocol didn’t create an incentive or fuse to stop the recursive loop where poor relative performance increases minting and minting hurts relative performance. A potential solution would require a dynamic price for insurance or distributing tail risk back to LPs.

How Does Bancor V3 (B3) Work?

LPs can deposit as much of any single token as they like (we will refer to it as TKN) on the Bancor Network. The network provides the other side of the liquidity pair (BNT) for traders to transact. The amount of liquidity in the network therefore only depends on TKN depositors, not BNT. BNT holders can stake BNT to receive trading fees, but it does not impact the liquidity in the network.

Each non-BNT trade on B3 sends fees to three different parties.

• Fees pay TKN LPs.
• Fees pay BNT stakers.
• Fees pay for an effective burn of BNT.

The vault holds all the tokens currently on the Bancor Network. The staking ledger holds a record of all the tokens owed to LPs. Because of single-sided staking and principal protection, trading creates imbalances between the two. The actual number of TKNs in the network vault will differ from the amount of TKNs owed to LPs.

The imbalance can be a surplus or a deficit for each token deposited. If the TKN outperformed BNT in the time it was deposited, the protocol will be in deficit (protocol is TKN-poor. It has fewer TKN in the vault than owed!). If the TKN underperformed BNT in the time it was deposited, the protocol will be in surplus (protocol is TKN-rich. It has more TKN in the vault than owed!).

At each TKN withdrawal, the network uses the transaction to try and return to equilibrium. It does this by changing the liquidity available in its virtual liquidity pools. New term: VLP (Virtual Liquidity Pool).

VLPs are what traders see when they come to Bancor. Prices and tokens available. They are virtual because they don’t reflect the exact token balances on the network, but they are computed by the protocol based on those actual token balances. Note that liquidity available is never more than actual token balances in the vault. All VLPs are TKN-BNT pairs. Bancor protocol uses this tool to protect and balance the network. Here’s how.

When a TKN that is in surplus (more in the vault than owed) is withdrawn, the protocol reduces the available liquidity of BNT in that VLP by less than the withdrawal would imply. In effect lowering the TKN-BNT price. This incentivizes buying TKN and selling BNT, which is great for the protocol because it has extra TKN!

When a TKN that is in deficit is withdrawn, the protocol reduces the available liquidity of BNT in that VLP by more than the withdrawal would imply. In effect raising the TKN-BNT price. This incentivizes selling TKN to the protocol and buying BNT, which is great for the protocol because it needs more TKN!

The protocol also must uphold principal protection. When a TKN is in surplus, this is easy and harmless for the protocol. When a TKN is in deficit, the protocol first determines how big of an impact the TKN withdrawal will have on TKN liquidity in the VLP. If it is small, the whole withdrawal can be in TKN. If it is large, Bancor will mint BNT and pay part of the principal amount in BNT.

As a check on the success of this design, look at the BNT supply chart before this May. Very steady.

May and June of 2022, not so steady.

The Design Flaw

A recursive loop: when you are upset about the number on the scale. And being upset leads you to eat ice cream to feel better. Then the number on the scale goes up. So you eat more ice cream…

The BNT-TKN relative performance determines the protocol deficit. A large deficit can lead to new BNT supply from minting. New supply drives the relative price lower. A recursive loop.

In June, Celsius’s unwind was the tail event that exposed the flaw. Liquidity on the Bancor network was low as market conditions led to depositor withdrawals, especially after the UST collapse in May. This happened across DeFi and is not unique to Bancor. Celsius had a large BNT position, along with other assets. They sold hoards of BNT, driving the price down and the protocol into deficit. As Celsius entered the cooldown period to withdraw their ETH, their principal protection would cause a substantial new mint of BNT. It was clear there would be a potentially catastrophic outcome. A “run on the Bancor.”

Was this inevitable, just waiting for someone to accumulate and then sell mounds of BNT? Perhaps. There is also another factor we have not mentioned.

Bancor began liquidity mining rewards in January of 2021, over a year before B3 launched in 2022. Stakers received extra rewards in newly minted BNT. The longer you staked, the more the rewards could compound. This exacerbates the pain of deficit minting as new BNT now comes from liquidity mining rewards as well as principal protection.

Celsius had very large deposits on Bancor, earning thousands of BNT in liquidity mining rewards. In retrospect, the liquidity mining rewards sound quite foolish in a mint-burn mechanism.

The design was flawed. There is no debating it — especially while the protocol sits frozen to new deposits and principal protection. And the decision to have liquidity mining rewards was a mistake.

Theoretical Solutions

The Bancor community and DAO are working on ideas to restart the protocol and move forward. Difficulty arises from:

1) How to get the funds to maintain principal protection for existing TKN depositors, where the protocol is in deficit?

2) How to create a sustainable protocol going forward?

I contend that if you solve 2, you solve 1. Reason being if you can convince “the market” that Bancor will survive, the market will buy and stake BNT. If you get outperformance of BNT, your deficit is gone or at least greatly reduced. So I will focus on question 2.

The approaches I offer are 1) explicitly share the tail risk with LPs or 2) create dynamic insurance pricing based on network conditions.

The first. Bancor protocol could enact rules for principal protection based on BNT minting. For example, if BNT net supply change over the past 10-days is greater than 2% (the number should be based on modeling not randomly picked like I did here), principal protection on withdrawals is paused. This is like an insurance company owing more than they have in assets — they would go bankrupt and claim-holders would only get partial payment. New rules communicate to depositors their share of risks on the network more clearly, while removing the death-spiral risks.

The second. Liquidity Providers on Bancor pay a premium for principal protection in the form of fees and a liquidity premium (7-day cooldown for withdrawals). In the same way that insurance premiums rise as risks (ask Californians how much fire protection is these days), perhaps the premium the network requires should be dynamic.

For example, as the deficit on the protocol grows, the share of trading fees can shift from TKN LPs to BNT stakers. This would likely need to be paired with an increased liquidity premium to avoid LPs rapidly withdrawing when fees are reduced. It would also attract BNT stakers, which should increase the BNT price relative to the TKN pool and return the protocol to equilibrium.

Part of the fix is to create a way to stymie a death spiral. The other is to instill enough confidence in BNT holders that there won’t be a death spiral. If you accomplish the first, the second should also be solved. But, as was the case this time, it will be tested eventually. Confidence is fickle in markets.

Conclusion

The B3 design was an incredible attempt to create a new DeFi protocol with a utility token. The protocol did not solve for tail risks and the DAO increased the probability of facing such a tail with liquidity mining rewards.

Bancor faces a tough road ahead. They currently sit with an almost $40 million deficit and $113 million market cap. Perhaps even more daunting: the market has been rallying and leaving BNT behind, only making matters worse.

The core design is tremendous and can offer liquidity providers a unique product on the efficient frontier. BNT holders and stakers will have a token that truly reflects the value of the network. But they have to address the tail risks and bring confidence back to the protocol.