The Bull Case for Bancor

Published in

Smart Money — DeFi Studio

19 min readApr 8, 2021

Disclaimer: Both Smart Money Ventures’ Yield Farming Fund and I hold positions in Bancor (BNT). None of what is discussed here should be taken as investment advice.

Despite the sharp increase in valuation in recent months following their V2.1 release, Bancor remains one of the most misunderstood projects in the decentralized finance space. Bancor is arguably quite differentiated from other AMMs, but the system is shrouded in a layer of complexity which makes it difficult even for many fairly technical DeFi natives to understand. In this analysis, I will attempt to simplify the core pieces of the Bancor system and provide some insight into their value proposition to users, their tokenomics and eventually the value accrual mechanisms that affect the price of $BNT, their native token.

The History of Bancor

To briefly touch on a bit of history, the original Bancor founders started with a mission to create local currencies, but found that in order to enable a less homogenous and locally empowered monetary system, there would need to be a market to swap between them for purposes of trade. This led them to the idea of an automated market maker or AMM. To solve this issue, Bancor was founded in 2016 with the goal of creating an automated market maker (AMM) to replace the traditional order book model found in traditional equity markets. The name Bancor actually refers to John Maynard Keyes’ idea for an International Clearing House (ICU), also called Bancor, which was proposed as a supranational currency to stimulate international trade at the Bretton Woods conference in 1944. Unlike a traditional market maker model where a centralized entity seeds liquidity for a given market at a predetermined range of prices and matches bids and asks in exchange for a spread, an AMM prices according to a bonding curve determined by the equation x*y=k and anyone can provide liquidity to any token pair for a share of trading fees.

However there’s an issue — how does an AMM bootstrap their network with ample liquidity before there are sufficient traders for liquidity provision to be attractive? There’s a clear analogy here to many Web2 companies who had to overcome the infamous chicken and egg problem; Driving on Uber is not valuable without a critical mass of riders, nor is riding without a sufficient supply of drivers. A social network is useless without other users to socialize with. It’s generally acknowledged that the supply side of the marketplace / network is often much harder to bootstrap and therefore many companies will strategically focus their efforts there — bribing with signup bonuses, high wages for early users etc. Within the world of Web3 and specifically De-Fi applications, there is often a critical mass of liquidity that is required for a well-functioning application, so many applications will choose to incentivize the provisioning of liquidity through a process of yield farming or liquidity mining. Through this process, users are rewarded with the protocol’s native governance tokens which have the characteristics of (pseudo) early-stage equity. Bancor is no different. To specifically incentivize the provision of liquidity, Bancor has strategically decided to focus their recruiting efforts on the supply side, offering rewards of $BNT and a high fee take rate to the key laborers of their network, the LPs, who are required to supply either $BNT (common denominator between markets) or a risk asset (basically any ERC-20 that has been whitelisted), not both as is typically the case with liquidity provisioning.

A Smarter Incentive Model For AMMs

Many liquidity mining programs are designed entirely around the bootstrapping phase, but often little thought is given to incentivizing the right kinds of actions that provide long term value to the network and retaining liquidity providers after rewards start to collapse. Mercenary capital comes in to “farm and dump”, but the best protocols view their liquidity mining program as just the first step in building a sustainable protocol. By creating a model where the token itself is essential to the functioning of the protocol, farm-to-dump becomes farm-to-stake and the rewards incentivize specific actions that increase the utility of the network and the native token itself. Anyone who has worked in recruiting knows that it’s not only necessary to just attract great talent, but you must also develop and retain that talent to build a sustainable company. Bancor rightly recognizes their dependence on their liquidity providers as their key talent and provides the right incentives in the form of rewards in their native governance token and high swap fees.

Bancor’s Token Model

With the release of V2.1, Bancor has added additional incentives in the form of impermanent loss protection (asset insurance) and single-sided liquidity provisioning (LPs seed one side of the pool and not both). Essentially, liquidity providers are rewarded with ownership in the network (which they can re-stake for compounded ownership), 50% of swap fees and are further incentivized with asset insurance (LPs don’t have to worry if their TKN or BNT changes in value as the difference between holding and LPing is refunded) and single-sided liquidity provisioning (you no longer have to sell off your treasured token to seed a market with 50% TKN and 50% BNT).

Single-Sided Liquidity provisioning and Impermanent Loss Insurance are the very core of the Bancor model and are enabled under the hood by something called elastic supply.

Elastic supply refers to the protocol’s ability to mint and burn BNT as needed. In the case of SS liquidity, the protocol mints BNT alongside other BNT LPs in order to match incoming TKN (risk asset) liquidity to increase the amount of SS liquidity space and to cover any kind of impermanent loss. Each individual pool has a co-investment limit that limits the proportion of the BNT LP side that can be protocol minted.

Initially, you should rightfully be concerned. What do you mean they’re just minting $BNT out of thin air? Won’t I just get diluted?

First, let’s start with single-sided liquidity. Just as there is a bootstrapping issue of liquidity for the overall network that must be solved with liquidity mining rewards, liquidity pools often initially face a microcosm of the same issue as they initially offer unattractive yields to LPs given that there isn’t enough transaction volume to generate sufficient fees. That’s where the protocol minted BNT comes in.

Whenever a new TKN LP comes in and wants to supply to the pool, there is (almost) always a matching $BNT there to meet them (up to the co-investment limit of the pool) which allows the pools to scale liquidity linearly as TKN LPs enter. The minted $BNT acts as sort of “welcome squad” for this new TKN liquidity and works to deepen the liquidity pool to a point where it would now generate enough organic rewards that $BNT holders want to stake the other side. As soon as LP $BNT comes in or a TKN LP leaves the pool, the protocol minted $BNT is burned along with its share of fees that it accrued while in existence.

Now that’s very interesting. This $BNT came from nowhere and went to nowhere. If any trades happened in between minting and burning and fees accrued to the protocol minted BNT, the burning has a net deflationary effect on the system.

The interpretation of this new supply in the market is perhaps the largest source of confusion in the market currently as Bancor is being valued according to its total supply on analytics sites like CoinGecko when in excess of 80m of that supply has been minted with the sole job of increasing single sided liquidity space. Furthermore, the process happens entirely within the Bancor system and the minted BNT is insulated from price discovery on the open market where an influx of supply could cause the price to drop, as is the case with most Farm and Dump schemes when rewards dry up.

Sure, but that doesn’t account for the impermanent loss insurance that must be paid out by minting $BNT right? Surely the inflationary issuance feedback loop must dominate the deflationary burning mechanism. The short answer is yes it could in any given pool, but averaged out across ALL the pools, the law of large numbers will lead to the burning feedback loop dominating the issuance feedback loop. The best way to look at this is like an insurance mutual — where risk for the mutual is diversified across all pools. Some pools will experience large amounts of impermanent loss while others will experience almost none. The impermanent loss protection itself is structured in a way that gives it the opportunity to benefit from the law of large numbers. This asset protection is granted in a progressive manner with a 30 day cliff where no coverage is granted, after which 30% coverage is granted on the 31st day and then 1% coverage is given every day until full protection is achieved at day 100. The rewards to BNT and TKN LPs (both denominated in BNT) are also granted in a progressive manner with a rewards multiplier that starts at 1x and increases by .25x every week, capped at 2x rewards after 4 weeks. If an LP pulls their stake out at ANY time, ALL rewards across ALL pools reset to 1x. Once liquidity pools are of a large enough size, a larger percentage of the IL payout will come from fees, which will have a longer time to accrue given that LPs are incentivized with progressive IL insurance and rewards multipliers to not pull out.

While individual pools may experience net inflation of $BNT, the pools together are likely to experience a net deflationary effect. Up to this point, only $.07 of every $1 of IL insurance has had to be printed by the protocol, so this has yet to be a problem so far. Even if some BNT has to be minted, the team sees this as the cost of doing business to serve their LPs in the right way and can even be a trojan horse for creating new BNT holders if they chose not to swap for their original TKN and stake their compensatory BNT for more rewards.

It’s a dirty secret in DeFi that most of the volume on Uniswap comes at the expense of arbitraged LPs. As a Bancor LP, once impermanent loss is covered, fees and rewards are pure profit — guaranteed passive income. So in summary, the Elastic supply creates the possibility of IL insurance and additional SS liquidity space while allowing the protocol to invest in itself through the burning of fees that accrue value back to the token holders. But there’s more. We’re assuming currently that the only non-inflationary mechanism for paying IL are the fees that stick to the protocol minted $BNT in each pool, but there’s actually several other sources that could either cover IL or unlock more SS liquidity space.

The first way that the protocol may come into possession of BNT or any other ERC-20 (TKN) is as a result of arbitrageurs rebalancing a pool, for example ETH / BNT. Since this is a decentralized exchange, ecosystem participants are necessary for performing this action in a decentralized fashion, in this case it’s an arbitrageur looking to make a quick buck. Remembering that x*y=k is the equation that governs the bonding curve, we can see that equal amounts of token A and token B in each pool are required to meet that constant. In other words, the amount of token A x amount of token B = some constant. Typically traders make money buying low and selling higher with time in between, but in this case the profit comes not from the timing of the sale but the location of the sale. Let’s call token A here ETH and token B will be BNT. Many trades that happen within this pool won’t be large enough to move the market prices, but some certainly will. In this scenario, let’s say that a trader swaps BNT for ETH and now there is more BNT in the pool and less ETH. ETH increases in price in the BNT pool because there is less of it and BNT decreases in price because there is more of it. To bring the pool into balance since the quantity of token A and B must stay constant, an arbitrageur would come in seeing this and decide to buy the cheaper BNT off the market and sell it on another exchange and pocket the difference. However, both sides of the pool are equal partners in losses and gains, so the ETH LPs will effectively come into possession of some BNT since one side can’t shoulder the losses fully. Likewise if BNT were to increase in price, BNT holders would come into possession of ETH. As some of the BNT side is protocol minted BNT, the protocol will come into possession of some amounts of other ERC-20 tokens, a reserve which amounts in the hundreds of millions currently and which can be used to pay out for impermanent loss without minting more BNT.

The other way by which the protocol may effectively come into possession of $BNT is as a result of the Bancor Vortex. As a result of the V2.1 release, $BNT stakers will now be issued vBNT tokens which can be used for governance purposes. However, the team recognized that the Bancor system is quite complex and not everyone would be sophisticated enough or even interested in participating in governance. As a result, they gave the vBNT token additional utility.

vBNT effectively turns BNT into collateral — Vote, Stake, Sell (Leverage). When a BNT holder stakes BNT in any liquidity pool, say BNT/ETH, they will receive 1 vBNT. Instead of using it for governance, they may swap it in the vBNT / BNT pool for more BNT to gain 1x leverage with no liquidation risk. This has the effect of increasing the space for single-sided exposure space and compounding the rewards of the $BNT holder further. The holder might also choose to use it as a leverage / flexible credit token. Selling vBNT is like borrowing against staked BNT. Since you have to buy back all vBNT you have sold to unlock the staked BNT eventually, the vBNT acts as a sort of key. In this case, the holder may sell the vBNT for any other ERC-20 token, perhaps ETH or Dai to participate in a yield farming opportunity.

This becomes very interesting as it’s now a way to gain leverage with no liquidation risk. However, there is some risk to the endeavor. The risk is that the cost of the round trip will be higher than never gaining leverage in the first place. If the price of BNT has increased (increases the price of vBNT as it is denominated in relation to BNT) and/or the ratio of vBNT:BNT has tightened or both, the cost of leverage will increase, requiring the user to market buy more vBNT to unlock their original BNT position. The amount of vBNT required to unlock the position never changes, which prevents any crazy short squeeze situations, but the cost of said vBNT can increase between its sale and buyback in $ value.

Currently, the price of vBNT sits at about .50 vBNT per BNT, but a more natural peg will be found once Full Vortex is initiated (just happened), effectively placing a slight upward pressure on the price of vBNT. Starting with a burn of 5% of swap fees taken from the TKN side of liquidity pools and increasing to 15% over time, vBNT will be purchased from the vBNT:BNT pool and burned. This has the effect of locking BNT in the protocol forever. The reason that is the case is because there is now more BNT than vBNT in circulation required to unlock it. It’s as if we all had our cars parked by a valet, but a certain percentage of the keys were incinerated, locking the car in the lot forever. This has the effect of not only decreasing vBNT in the pool, but also increasing the amount of protocol owned BNT which can now be used to increase the space for single sided exposure in the system and pay for impermanent loss protection.

Another way by which the protocol might effectively come into possession of BNT is if this upward pressure on the bonding curve is not counteracted with enough traders taking leverage, presenting a scenario where vBNT > BNT. This creates an arbitrage scenario for restoring the peg. Let’s assume that vBNT is now 10% higher than the price of BNT. An arbitrageur could purchase $100 of BNT, stake it for $110 of vBNT, swap the vBNT for BNT in the vBNT / BNT pool and swap that BNT for $110 of ETH. The interesting scenario is that it is unclear if the arbitrageur would ever come back for their staked BNT. They may be happy to take their $10 profit and leave and never participate in the Bancor ecosystem. So this BNT is now at the disposal of the protocol to increase SS liquidity space and pay out IL insurance.

With a small upward pressure on the price of vBNT as a result of the Full Vortex vBNT burn and a natural mechanism to return the peg if vBNT overshoots BNT, vBNT will eventually settle into a less volatile peg where the cost of leverage will be more predictable.

In conclusion, pool rebalancing by arbitrageurs, Full Vortex fee burns, leverage taken while vBNT price is increasing, and situations where vBNT > BNT could all lead to situations where the protocol comes into possession of BNT or any other ERC-20 token. This protocol controlled value or PCV can be used to increase the space for SS liquidity or as a backstop for paying out IL if protocol owned BNT swap fees aren’t sufficient before having to print new compensatory BNT.

How Bancor’s Token Model Aligns Incentives For $BNT Holders

What struck me about the Bancor model was not only the elegance of their tokenomics, but how there seemed to be some very clear feedback loops that could be modeled out in a flow chart as well as in a discounted cash flow model. The idea for modeling out these feedback loops were sparked by a prescient twitter thread from last summer by Andrew Kang of Mechanism Capital (@Rewkang). Recognizing the chicken and the egg problem of bootstrapping a market maker, Andrew correctly identifies how providing the right early incentives for liquidity providers is necessary to getting a liquidity feedback loop churning as higher liquidity reduces slippage, making prices better, attracting traders, increasing transaction volume and profits for market makers and so forth. However, this thread was produced well in advance of the release of V2.1, so I hoped to update the model with the new parts of the system and show how they directly lead to increased utility to the native token BNT and lead to increased value capture over time.

To frame this discussion on well aligned tokenomic incentives, Wangerian (@Wangarian1) made an interesting point in his thread “Can An Old Dex Learn New Tricks” (https://twitter.com/Wangarian1/status/1367087459255738370?s=20) that LPs and tokenholders in the Bancor system are one and the same. The BNT holders are owners of a market-making cooperative. What incentivizes BNT holders / LPs and attracts TKN LPs makes the platform more useful and valuable and by extension, the LP’s BNT more valuable. Bancor needs to incentivize those LPs correctly and the rest will follow.

Protection against impermanent loss and single sided liquidity provision are the core of the system and essential for attracting LPs. Not only could you lose money as an LP if your impermanent loss is greater than the pool yield, but it actually completely destroys the portfolio construction strategies of most DeFi investors. Similar to early stage venture capital, investors make a series of bets that have a limited 1x downside, but also have the potential for unknown asymmetric upside, so the expected return across all bets is hopefully positive. This is why we HODL, or hold on for dear life. We expect these tokens to “moon” in value. With impermanent loss, your tokens can only increase in value by the square root of x. A 1000x becomes a 100x, a 100x is limited to a 10x and so forth. This breaks the math of portfolio construction and is an invisible risk that costs LPs more than they are currently aware of. Not only might Bancor benefit from an influx of LPs as they wake up to the true costs of providing liquidity provision, but Uniswap V3 has now made this a professional’s game with concentrated liquidity. The products have completely diverged.

It’s possible that Uniswap will not only expect their LPs to shoulder this cost, but also share fees in the future with passive token holders, so it will be interesting to see what effect that will have on their system. The lack of single sided liquidity protection on Bancor is initially what has given Uniswap an early lead over Bancor and others. LPs were hesitant to sell off any of their holdings for BNT and Uniswap was able to solve that by making ETH the common asset on all token pairs as many DeFi users already owned ETH. However, requiring equal parts ETH and risk asset may require LPs to sell off some of their tokens and regardless, both sides are exposed to impermanent loss. This leaves LPs effectively rooting against their tokens increasing in value while they’re providing liquidity. With zero utility currently for the Uniswap token in the functioning of the Uniswap exchange, the most capital efficient use of Uniswap tokens is actually to provide liquidity to the UNI / BNT pool on Bancor.

Why It Works

So let’s jump into the model. Providing impermanent loss protection (asset insurance) and requiring LPs to seed only one side of a pair (single sided liquidity provision) leads to more LPs moving their liquidity to Bancor, which conversely increases SS liquidity space as some of those will be BNT LPs and increases the possibility of impermanent loss protection as the pools will be larger and accruing more fees. More LPs leads to more liquidity and greater pool depth will mean there is less slippage on trades = better prices, which should attract more traders, create more transaction volume and lead to more fees.

Briefly detouring to the inner circle, more fees means more burned vBNT as Full Vortex will burn 5–15% of swap fees taken from the TKN side and burn the equivalent amount of vBNT, locking BNT into the protocol to increase single sided liquidity space or to be used to pay for impermanent loss and additionally increasing the price of BNT. Continuing on the outside of the loop, more fees will lead to higher market maker / LP profits which will lead to more LPs, but it also means more BNT rewards are being paid out over that time. For each BNT that is rewarded, vBNT is also rewarded and a certain percentage of both (currently about 80%) will be re-staked for compounding rewards which again increases space for new TKN single sided liquidity to come in and increased impermanent loss protection as pool fees will increase.

Some of the created vBNT will lead to locked BNT — either from the Full Vortex fee burn, vBNT holders taking leverage as vBNT is increasing in price or arbitrageurs leaving their money staked after vBNT exceeds the peg. This BNT locked into the vortex forever can be used to increase SS liquidity space for TKN LPs to come in or to pay out impermanent loss as a backstop if protocol accrued fees are not sufficient before having to mint any BNT. At the core of the loop, a higher price of BNT itself increases the $ value of the reserves which can pay out IL and increases the $ value of the BNT side LPs which has the effect of increasing space for TKN LPs to come in. Finally, the existence of IL protection itself and the ability to only provide liquidity to one side of a pair increases the utility and value of the protocol itself which is represented in the value of the BNT token.

Conclusion

Automated market makers and lending are arguably the only products that have achieved real product market fit in Defi. Centralized exchanges like Coinbase will be the portal into Defi, but they will go the way of Aol, the easiest first interaction with this new technology platform, but not as expressive and useful compared to the cambrian explosion of Defi primitives that allow you to borrow, lend, stake, yield farm, provide liquidity etc. You can sit on some coins in a wallet, but that’s not too fun is it? That’s why so much Ethereum has been exiting centralized exchanges in recent months.

As Coinbase goes public at a valuation of $100b and exposes millions of new investors to crypto, I anticipate this will benefit decentralized exchanges across the board as centrally bought coins exit and look for more productive use cases. It will take some time for the competitive landscape of composable DeFi middleware protocols to settle, but in that time, we know one thing for sure, coins will be bought and sold on decentralized exchanges. To benefit from randomness, both the Yield Farming Fund and I have chosen to invest in one of the arms dealers of the industry, the exchanges — who will generate real fees regardless of what tokens win which in turn helps to set a valuation floor. Despite being in the top 10 for total value locked (TVL) and top 5 of Defi protocols for fees, Bancor is valued at just 1.3B fully diluted, a mere fraction of Coinbase’s $100b valuation, Uniswap’s 30b FDV, Sushi’s 3.5B FDV and even PancakeSwap’s 2.5b FDV. Additionally, most other market makers are being valued according to revenue that includes both supply side LPs and holders of the protocol token. This is akin to valuing Uber or Doordash according to GMV and not subtracting out driver revenue to get the app’s take rate. After accounting for impermanent loss, Bancor pays LPs 50% of swap fees, compared to 16% for SushiSwap and 0% for Uniswap. As I look for investments that are mispriced by the market, have asymmetric upside, are positioned to benefit from randomness, have a great product and a dedicated team, Bancor checks all those boxes.

Thank you the Bancor team including Mark Richardson, Nate Hindman and Michael Herzyk. Mark in particular spent several hours additional hours on Zoom with me helping me understand their system and they all answered countless follow up questions via telegram. I have attached a few of the best resources I have found below and would invite you to follow me on twitter @patfitzgerald1 for more research notes in the future. You can follow the rest of the Smart Money Team @Smart_Money__ and @_SmartFunds_ .

Resources

Reading

Twitter Threads