DPI liquidity on Uniswap v2 and v3
warning, long and technical (and probably littered with errors), I would put this in my blog, but it’s not fully formed…
Ok, a few members of the PWG has been looking at Uniswap v3 and how the coop can use it to improve matters for users, LPs and the coop [Forum post coming soon].
When we started, I was aware that we had ETH2-FLI liquidity on v3, but I wasn’t aware that there was a DPI:ETH v3 pool running. So I decided to do some digging:
The basic process has been to:
- Look at the pools on Etherscan (v2 and v3), between 23rd May and 05th June 2021 (14 days)
- Download all the transactions for each as a CSV
- Add hyperlinks to the actual trades and do some sorting, filtering and manual assessment of the transactions.
- Manual assessment has been focused on the larger transactions so there is a very large chance that I’ve failed to assign some correctly.
Some observations:
DPI:ETH Uniswap v2 and v3 pools 25 May to 05 June 2021.
| v2 | V3 | |
|---|---|---|
| Total transactions | 3,782 | 867 |
| No. Buys | 2,100 | 361 |
| No. Sells | 1,481 | 370 |
| No. Arbitrage (issuance or with different DEX) | 84 | 69 |
| No. DPI issuance (and swap) | 65 | 0 |
| Largest BUY (DPI) | 1459 (only v2) | 1670 (only v3) |
| Largest sale (DPI) | 1131 (only v2) | 289 (split) |
| Total $ buys (@$400) | $50.4 m | $ 5.1 m |
| Total $ sells (@$400) | $39.5 m | $ 3.9 m |
| Pool AUM (13jun21) | $41.5 m | $3.0 m |
| Average APY on volume / AUM annualised | 17% | 23% |
It appears that either there are no issuance arbitrage bots looking at the v3 pool, or there is more profit to be made going via the v2 pool (i.e. the centre of gravity for the DPI:ETH pair is still in v2)
Comparing the pool trading volume and size, the v3 pool is achieving a larger APY for the LP’s however, this will be higher for those who have used bounds closer to the market price (and lower for those with wider bounds, or zero for liquidity that does not cover the market price).
Note the excellent flipside uni v3 calculator, estimates a 16.4% APY for the last 24 hours for a market price +/- 16% range.
While v2 returned ~10% (Trade volumes have dropped significantly since the end of May.
Looking at each pool individually, and splitting the transactions into bands (lowest 10% of sale, lowest 20% of sale etc), I can get a profile for each. Assuming a $400 price for DPI, I can get a rough $ value for the trade at each position.
DPI:ETH Uniswap v2 pool, $value of each % for each type of transaction (23 May to 05 June 2021)
| Position in group | Buy | Sell | Issue* | Arb* |
|---|---|---|---|---|
| 10% | $116 | $459 | $89,588 | $12,128 |
| 20% | $400 | $1,927 | $115,958 | $43,697 |
| 30% | $830 | $5,032 | $124,450 | $98,650 |
| 40% | $2,065 | $9,562 | $142,016 | $120,870 |
| 50% | $5,221 | $13,332 | $157,413 | $136,818 |
| 60% | $11,666 | $17,227 | $183,770 | $157,413 |
| 70% | $18,043 | $22,144 | $213,400 | $183,770 |
| 80% | $30,310 | $32,000 | $240,625 | $218,440 |
| 90% | $60,005 | $56,042 | $264,199 | $251,860 |
| 95% | $131,806 | $101,483 | $328,487 | $287,732 |
| 99% | $241,342 | $255,524 | $414,363 | $414,363 |
| 100% | $583,791 | $452,691 | $542,889 | $542,889 |
*note, there is a high risk that I’ve missed some of the smaller Arbitrage / issuance transactions.
Overall, it makes sense that the arbitrage / issuance transactions are generally larger than simple buy and sells. Arbitrage / issuance (with a swap in the same transaction) are always going to be driven by profit, which is unlikely to exist on a $116 purchase.
This analysis indicates that 5% of the purchases were greater than ~$130 k, and 1% greater than ~$240 k. Likewise the sales were > $100 k and $415 k.
DPI:ETH Uniswap v3 pool, $value of each % for each type of transaction (23 May to 05 June 2021)
| Position in group | Buy | Sell | Issue* | Arb* |
|---|---|---|---|---|
| 10% | $166 | $400 | $0 | $2,295 |
| 20% | $400 | $999 | $0 | $14,250 |
| 30% | $764 | $1,733 | $0 | $18,303 |
| 40% | $1,200 | $2,663 | $0 | $27,543 |
| 50% | $2,239 | $4,402 | $0 | $34,646 |
| 60% | $4,112 | $8,865 | $0 | $36,690 |
| 70% | $8,208 | $12,762 | $0 | $41,194 |
| 80% | $14,488 | $18,814 | $0 | $46,249 |
| 90% | $30,086 | $26,618 | $0 | $68,522 |
| 95% | $46,728 | $31,039 | $0 | $73,652 |
| 99% | $63,703 | $62,400 | $0 | $171,095 |
| 100% | $667,962 | $115,548 | $0 | $250,795 |
*note, there is a high risk that I’ve missed some of the smaller Arbitrage / issuance transactions.
As noted before, I did not find any large trades that included the issuance of DPI with trade on the v3 pool. However, there were at least 62 Arbitrage trades between the v3 pool and the v2, and 6 between v3 and the Balancer pools.
This analysis indicates that 5% of the purchases were greater than ~$47 k, and 1% greater than ~$64 k. Likewise, the sales were > $31 k and $62 k.
Some other insights
I was surprised to see large value arbitrage trades that only captured a few hundred $ of profit. This means that such large pools should always be close to NAV.
Because of the concentrated liquidity on Uniswap, the overall size of the pool is not related to the trade depth at the market price. So while in theory, v3 should give better liquidity for trades, it is not symmetrical and is not easily calculated.
E.g. this is the current v3 DPI liquidity profile with lots of liquidity around 0.15 ETH to DPI (i.e. DPI that will be sold when it reaches 0.15 ETH) Note: 0.15 ETH was around the market price at the end of May 2021. So it may be concentrated liquidity placed 2 weeks ago and left passive.
While the largest purchases in each pool were single sources, there have been many purchases that used aggregators to split the trade across v2, v3, Balancer, Sushiswap, and 0x protocol. For example this trade sold 153 DPI on v2, and 246 DPI on v3, while this one purchased 491 DPI on v2, 356 DPI on v3 and 52 DPI on Sushiswap.
During this analysis, I also discovered that a single wallet was making a large number of purchases between 22nd May and 03 June. All in all, I think they spent 1,200 ETH on DPI (Worth ~$28,000,000 at the time). The majority of this is sitting cold in a second wallet, with ~$4 m on Cream as collateral against ~$900,000 in stable coins.
Please note that I’ve not examined all of the transactions, but I think that the assumption that most arbitrage and issuance will only be using larger value transactions is valid. However, I’m certain to have miscategorised some smaller buys and sells (which should be add or remove liquidity). Even so, I would expect these to be a small fraction compared to the number of swaps.
Overall, my conclusion is that while there is v3 liquidity for DPI:ETH, a significant proportion is not at the market price, and the majority of trade number and volume is still on the v2 pool.
If anyone wants to dig into this data further, it can be found here (although, you may want to download a copy and work locally).
The PWG team have been looking at uni v3 will share their findings on Uni v3, a comparison of v2 and v3 pools, automated liquidity managers and liquidity mining in the near future.