Has IL issuance and sales stopped?

I would love to say yes, however, from what I have seen on chain, I don’t think so (and I would love to be proven wrong)…

So, what does it mean, what can we do?

What is happening?

Looking at the (excellent) IL protection Dune we see IL issuance of BNT has spiked in the last week:

I think ETH is our main issue (and I’m going to focus on ETH), however, I suspect all the other TKN:BNT ratios are showing similar trends.

Coingecko shows BNT : ETH fairly steady over the last 90 days, then dropping in the last week.

Note, this weeks volume is higher than recently, but not greater than other periods in the last 90 days. However, this time, we see more price impact. What’s happening is rather than buys and sells being about equal, there are more sells, and so sells are going into the protocol and changing the # of BNT in each pool and pushing the BNT price down.

So, a key problem is simply more sellers than buyers.

Are Celsius selling?

I believe so, and I think that (one of) the Celsius wallets selling is this one 0x3B65.

If we look at the last 3 days we see the following pattern each day: Exit B3 withdraw queue receive ETH and BNT, Sell BNT for ETH, transfer ETH.

Looking at the transactions we can calculate the ETH to bnETH ratios:

Date transaction bnETH redeemed ETH received % ETH BNT received BNT $
14 June 11:20 link 1999 1418 71% 835,000 $680,000
15 June 12:03 link 1999 1362 68% 1,000,000 $740,000
16 June 12:45 link 1999 1331 67% 1,097,000 $648,000

Note, these values are significantly smaller than the global BNT volume (~$35 M, which may miss B3 volumes) so there is lots of other things happening (Bancor 2.1 IL issuance, other Bancor B3 IL issuance).

As the BNT : ETH ratio drops (i.e. ~ BNT price falls in $USD), then ETH IL in that pool increases, so Bancor issues more BNT.

This is the key to our short term problems:

When there are large withdraws at a time when BNT price is low, we issue more IL BNT, if it’s large enough, then resulting sale pushes the BNT price down further.

Remember, IL is vs BNT for all other tokens, BNT dropping away from the market, is bad.

Celsius are not the only people causing large IL protection BNT issuance

Note, these values are significantly smaller than the global BNT volume (~$35 M, which may miss B3 volumes) so there is lots of other things happening (Bancor 2.1 IL issuance, other Bancor B3 IL issuance).

In fact, looking at this bundle of (not verified! ) Celsius wallets, they have only sold 5.5 M BNT (for 3,724 ETH ~ $4 M) in June.

According to this query, there has been ~36 M BNT issued for IL in June. So, the identified Celsius wallets are 15% of the IL issuance in June (or the wallet bundle has missed some Celsius wallets… :roll_eyes:)

Have the sales stopped?

I’m assuming that most of the large v2.1 withdraws have been done as they only include a 24 hour lock up…

Looking at this chart:

Protected balances in v2.1 dropped for ~140 M on 30 May to ~$95 M on the 13th June.

(Note, I don’t know what happens when tokens migrate to B3, does the BNT IL debt get transferred, or is it paid in some way and B3 has a fresh start :thinking: )

Looking at selected tokens in the B3 withdraw queue:

Bn number Price (1:1) Value in queue
ETH 30748 $1,095 $33,669,060
Link 2653358 $6.46 $17,140,693
wBTC 41 $20,979 $860,139
DAI 7905495 $1.00 $7,905,495
USDC 2000014 $1.00 $2,000,014
Total $61,575,401

i.e. There is another $61 M to exit with the majority being ETH…

However, this is much less than in the queue at the start of the month (but we could get more IL issuance as BNT price has moved…).

How much is Celsius?
In some respects, this is irrelevant, an LP is withdrawing, and is expected to get BNT which we expect they will sell.

However, we can see from here that the suspected Celsius wallet has ~14,877 ETH in the withdraw queue. So about

  • ~$15 M of ETH to be removed,
  • IP protection paid in BNT,
  • and presumably sold by Celsius. (50% of total ETH exiting B3)

So we are expecting IL issuance and sales over the next week (some Celsius, but most others).

What can we do?

Firstly, BNT is one of the most liquid tokens. The vast majority of BNT in existence is sitting in constant product AMM pools.

This means that a $1 M sale (or purchase) of BNT has a much smaller effect than a $1 M sale of anything else.

So, we are in a better place than others to weather this storm.

I see a few options for what we could do in the immediate future. None of them are great, and they will all have effects.

1. Do nothing

Bancor is a robust protocol, this scenario has been modelled. BNT will get issued to cover IL, but nothing breaks (However, BNT holders get diluted)

2. Build more liquidity on Bancor.

More liquidity means two things:

  • In pools below the liquidity limit (or cap on v21.) sales have a smaller impact on BNT price, and so reduce the IL paid to those leaving any pool.
  • In B3 pools above the liquidity depth, we have more surplus TKN which dilutes IL for the withdraws from that pool.

In both cases, it effectively pushes back the IL to a future date / spreads it around.

3. Increase trading depth on B3 pools with a surplus.

Increasing depth, reduces price impact. So BNT $ value drops less for a given trade. Opening up more ETH / USDC / DAI to trade, we reduce BNT issued for IL protection. So less dilution.

However, once again, while reducing IL overall for Bancor (for other pools), it increases it in those pools adjusted (which will need to be paid in some way in the future).

  1. Increase pools fees when we expect BNT sales.
    If we increase v2.1 and B3 pool fees (e.g. ETH from 0.1% to 1%) for the next 2 weeks it will have a few effects (and probably some I miss):
  • We capture more fees to the protocol for IL protection and vBNT burns.
  • We penalise those selling (and those buying) BNT by taking a larger slice of the trade.
  • APY’s on pools increase, so we may gain LP’s and cause some to stay in the pools.
  • We will loose some volume to other DEX’s / CEX (mainly on TKN to TKN trades).

Please note, I’m not an expert on Bancor economics, so there may be critical things I’ve missed. I’m happy to be corrected.

Any comments?


Just brainstorming here:

  1. What about turning up the vortex on V2.1 to 50%? Put some buy pressure on vBNT.
  2. Set the withdrawal fee on V3 to 0.5% - (Still a tiny price to pay for IL protection)
  3. Do nothing, my APYs will increase if people panic run away.

Thank you for posting this @OverAnalyser - very interested to see the response from the community on all of these options.


Great review, first and foremost. Exciting to see such effort taken to discuss such topics.

The other issue we are facing is not just IL, but large withdrawals that had a great deal of LM rewards from the 2.1 era. Reporting wise, these can be indistinguishable.

These being turned off now are a great thing for Bancor moving forward and will soften the landing of any future forced withdrawals that we might incur.

I believe most of our major exposure by large players running into collateral issues is over. But no way to know for sure since depositors don’t come with labels. Either way, Bancor will continue to be battle tested and rise from the market ashes stronger than ever.


I’ve been working with MikeBurgersburg on Twitter to track Celsius DeFI wallets, and I’m fairly certain Celsius only had three Bancor wallets. They are:

0xf73C37dF30571b9b58dA19C7473964369011DBeC stayed on Bancor v2.1, and has minted a lot of BNT there because of IL.

0x3B6543ebe26824Bd8156A103063F56cE50F88080 migrated to v3 before withdrawal, and so we were a little more protected from IL minting by v3’s design

If you have doubt about a wallet being connected to Celsius, it’s pretty easy to verify by looking at the flows on bitquery and cranking up the depth and detail levels of the graph. The Celsius backend has a very obvious fingerprint, a bunch of small wallets feeding into/out of a hub that connects to other hubs with a bunch of small wallets.


Good questions:

Increasing the vortex should help long term, but doesn’t stop the BNT sliding further (which increases the number of BNT issued to cover LP IL).

Also, isn’t the vortex burn denominated in BNT?

APY’s increasing will attract more AUM (also, 7 day APYs probably look at current pool AUM, and 7 days income, so a 50% drop in pool AUM displays a doubling of APY).

1 Like

Thank you @Linksemper

I admit that I don’t have a good understanding of the v2.1 rewards that are being claimed. However, the v3 exit queue shows exit volume for the next week.

Whether the market is buying at the time we will see. Otherwise, it will be BNT liquidity absorbing it.

I’m sure the protocol will survive, and I’m fully in support of not issuing more BNT as farming rewards.

1 Like

I can appreciate that, I’m just a guy with a spread sheet. SO thank you for the Dune dashboards. They help shine a light into what’s happening under the hood.

I would actually argue the opposite. BNT should have a longer queue. Then if there is a mass redemption and IL issuance, BNT LP’s can NOT front run the IL issuance (and consequent BNT price dip).

IMHO, BNT liquidity is a key part of our safety backstop. Keeping it in the protocol (and as liquidity) reduces selling pressure at times like this. AVVE locks staked $AAVE for a week so they can be slashed within a week of any even.

If you were a v2.1 BNT staker at this time, you could withdraw within 24 hours, sell BNT, wait for the mass redemptions in the B3 queue to happen then rebuy more BNT at a lower price.


Thanks, Thats very interesting.

At first glance, it looks like Celsius only have the ETH wallet with any exposure to Bancor at the moment. So it’s just the ~14,000 ETH to be withdrawn with IL protection…

The end is in sight.

1 Like

thats comforting - i also saw a proposal to disable deposits on CEL


(Note, I don’t know what happens when tokens migrate to B3, does the BNT IL debt get transferred, or is it paid in some way and B3 has a fresh start :thinking: )

FYI, the migrations from v2.1 to B3 uses the “fully protected” value which is the initial deposit + fees + full coverage from IL (including any loss due to rebalancing).

1 Like

I assumed as much.

What i was unsure was if a LP position was in deficit.

E.g. if a v2.1 position was 100% IL protected but there was insufficient TKN available (i.e if the position was withdraw the LP would get TKM and BNT) does:
a) 100% TKN migrate (and the v2.1 pool uses BNt to buy THN???)
b) the TKN and BNT migrate

I suppose the question is “does B3 inherit the IL of a migrated position or is it created with zero IL”?

AFAIK, 100% of the tokens for the position migrate.

If the LP position is such that his tokens are in deficit:

e.g. he deposited 50 TKN but only 25 TKN are available. If he were able to withdraw from v2.1, then he would have received the 25 in TKN and the remaining in $BNT.

When this position migrates to B3, 25 tokens will get migrated but his balance of 50 will be recorded (plus any fees). This is essentially his staked balance. Bancor 3 has a “stakedbalanced” for every TKN pool which is essentially the cumulative deposits for all LPs of that TKN plus any fees that they have earned. This can be compared to the balance of that token that exist in the Bancor 3 master vault. The difference between these two numbers will tell you if there is either a deficit or a surplus for a pool.

Can confirm that my position that was in deficit that I migrated showed exactly this in the Etherscan tx. 40% of the TKN was migrated but the UI is tracking the 100%+Fees amount of TKN

1 Like