Proposal : Include ICHI Pool within Almanak-managed pools

:sparkles:Hi Bancor community! :sparkles:
The ICHI team recently approached Almanak team to assess the opportunity of including the ICHI pool within the current scope of Fee/TL recommendations to boost minimization of the ICHI pool deficit on Bancor v3.

Summary

Almanak would like to propose a partnership with the ICHI DAO, where Almanak would deliver risk parameters updates for the swapping fee and the trading liquidity of the ICHI pool on Bancor v3. Approval of this proposal would result in the weekly update of these parameters with the other 4 pools currently serviced by Almanak (ETH, LINK, DAI, WBTC). The cost of that service would be supported entirely by ICHI DAO, no additional cost for Bancor except the gas fees of the weekly update deployment.

EXPECTED VOTE DATE: before Jan 8th, 2023.

FOR:
Implement Almanak’s weekly trading fee and trading liquidity recommendations for the ICHI pool on Bancor v3 for a specified trial period.

AGAINST:
Do not implement recommendations from Almanak for the ICHI pool on Bancor v3.

History & Context

Almanak is currently collaborating with the Bancor DAO to reduce the deficit of the v3 protocol. The current approach focuses on the four biggest pools on Bancor (ETH, LINK, DAI, WBTC), where both the swapping fee & the trading liquidity are managed by Almanak. Since the start on November 14th, fees per volume have doubled across these pools whilst holding the same volume as comparable pools in the same period.

Over the first month of implementation, Almanak was able to contribute to the following results (first monthly report here):

  • Protocol Deficit: 1-month decrease of 16%, circa $5M
  • Swapping Fee: Overall increase of the fees to support the revenue while striking the balance with required arbitrage volume
  • Almanak Pools Performance: Ratio Fee/Volume is above non-Almanak pools despite observed decrease in volume across the market

ICHI has been part of the Bancor journey since whitelisting ICHI Legacy on Bancor v2.1 (followed by co-investment), followed by the migration of the liquidity onto v3 and transferring to ICHI.farm. Both proposals have been voted for with a large majority, cementing the partnership between the two protocols and led to the ICHI <> BNT pool on ICHI with additional xICHI incentives.


Fig. 1: Historical deficit of ICHI pool on Bancor v3

With the stop of impermanent loss protection on Bancor v3 right after the approval of the ICHI liquidity pool, an imminent deficit took over the pool. The current deficit compared to staked funds stands at 35.79% (December 16th 2022).

The foundation to reduce the deficit over time is twofold and follows the outline currently executed for the other pools on Bancor:

  • Change and monitor the swapping fees to attract volume without extracting value from the pool from arbitrage
  • Reduce the trading liquidity (deposits being available on-chain and prone to value extraction) if necessary to shield current funds from further deficit.

Especially in a volatile market environment with changing directions, Almanak provided a solid foundation to its controlled pools, lifting fees collected by trades without butchering volume on pools. Especially taking on an additional pool such as ICHI will contribute to the closing deficit and validate the current track record.

Appendix

Methodology & Outline

A detailed overview of the methodology how fees and trading liquidity are optimized can be found here (as a governance post) or here (as a report). Only an abstract will be made available in this forum.


Fig. 2: Flowchart of a simulation run

We leverage our agent-based simulation platform to identify the best possible sets of parameters to shorten the path to recovery of the protocol. We distinguish three relevant types of agents for the Protocol: liquidity providers, traders and arbitrageurs. Prior to an optimization run, simulation agents are being re-trained and adjusted based on the latest data from the past months. We perform overfitting tests to validate the calibration.

To measure the efficacy of the simulation and agent-based modeling solution, we run walk forward optimization and tests which not only validate our approach but help refine hyperparameters.

Finally, we select the optimal pool swapping fees and on-curve trading liquidity that maximize the pool revenue and minimize pool exposure to deficit.

As for the other assets within Bancor, this will also apply to the pool containing ICHI tokens. We include the pool with its LPs and traders & arbitrageurs to interact with the pool throughout the simulation. Note that due to the smaller size and the limited volume, the ICHI pool is optimized separately from the larger pools to avoid any suboptimal solution-finding for the smaller pool. The update frequency stays the same as for the current pools.

Note that the reporting can be posted on the Bancor governance, yet we are open to transfer reporting onto any platform the ICHI DAO is most comfortable with. An example report currently published on the Bancor governance forum can be found here.

The battle-tested methodology developed by Almanak focuses on state-of-the-art concepts and tools to explore the open solution space and extract the optimal solution to each individual pool. By tailoring the different components of the simulation environment and the optimization to the risk profile of each asset, the proposed parameter set is adjusted to reflect the current state of the market as well as the future risk borne by the pool.

Analysis

Based on the parameter set available for the other assets on managed Bancor v3 pools, Almanak optimizes the mentioned risk parameters within the same optimization environment, yet changing the agent behavior to adjust the actions and respective reaction of the pool in regards to any potential market scenario.


Fig. 3: Asset correlation within the Bancor protocol covered by Almanak & ICHI

As a first iteration, we execute a first iteration of optimizations to check all parameter sets and their performance on a plethora of price trajectories. Note that for the sake of simplicity, we focus on profitability only, which is seen as the main driver to reduce a pool deficit. We additionally highlight the deficit saved from the different protocol recommendations.


Fig. 4: Heatmap of monthly ICHI pool profits of averaged parameter sets

The initial observation coming from the optimization is to initially reduce the trading liquidity by a bigger margin compared to the other pools. A potential reason for this is the strong correlation between ICHI and BNT, leading to an even more impactful deficit growth once the asset price decreases. “Storing” that share of the trading liquidity helps to limit the impact on the pool deficit.

Trading fees are also impacted by the optimization, where trading fees are slightly higher compared to the swapping fee currently on-chain. This is mainly driven by getting higher value extraction from each trade, as those are limited on the platform. The combination of reducing the amplitude of potential asset value decline and raising the fees per volume unit as a pair are the plan to reduce the deficit over time.

To conclude, Almanak opts to reduce the trading liquidity of the ICHI pool on Bancor to limit the impact of further declines in token value and therefore the additional deficit. Additionally, swapping fees will be slightly increased to bring additional revenue to further minimize the current deficit of the pool.

Implementation & Conclusion

Implementation of the recommendations are upheld by the multi-signatories of the Bancor protocol. Since the recommendations are pushed at the same time as for the other pools serviced, implementation time stays the same. Gas costs for changes in swapping fees and/or trading liquidity needs to be covered by the ICHI DAO.

Weekly updated recommendations can help to stimulate trade volume and incrementally accrue more profits over an expected volume to limit the deficit on a respective pool. As the current results of the other main pools covered by Almanak have positive outlook, we highlighted a similar, yet constrained impact on the ICHI pool on Bancor v3.

4 Likes

I’m all for it since the ICHI DAO is covering the costs

4 Likes

I plan to vote no (unless this proposal changes).

It seems like everyone wins but the Bancor DAO.

ICHI gets Almanak to manage their pool
Almanak has a new client in ICHI
The Bancor DAO gets to… pay gas…

Unless someone is buying and burning some BNT. I vote no.

2 Likes

I think you don’t understand it completely. Almanak will generate extra fees from this ICHI pool which in turn will buy more BNT(90% of fees goes to buying BNT,10% of fees goes to users).It’s exactly what you said you want.

1 Like

That’s a goal. Not a promise.

Why should our DAO cover all the gas here and not ask anything in return?

1 Like

Hi @eldude, Philipp from Almanak here.

Thanks for joining the conversation and voicing your concerns. We value the interaction with the Bancor community and are happy to clarify the points raised.

  • As @mh89spd already highlighted, the interests of ICHI token holders on Bancor are aligned with the overarching goal of Bancor: reducing the deficit. With more swapping fees earned and further deficit reduced, both the protocol and the pool benefit. Regarding historic performance, the current results of the ongoing trial of recommendations provided by Almanak can be found here.

  • As mentioned in the proposal above, gas fees will be covered by the Bancor DAO. The current estimate for additional gas costs has to be verified by the Bancor team but is estimated minor so far.

Let me know if you have any further questions or comments!

1 Like

As noted above, unless someone is willing to buy/burn some pre-defined amount of BNT per week, my vote is no.

2 Likes

perhaps if you can show that almanak’s work created a significant fee boost, this can be a valid vote.
so far it seems that the first month ended up with many changes and zero impact (not to say a possible negative one).

given that i would suggest to vote against this and let the fee structure on the ICHI pool stay as is.

actually, if the ICHI dao thinks a different fee might be more relevant, perhaps the vote can be to perform a one time change to a new fee structure.

2 Likes

Do we have any kind of formal proof that changing the fee portion according to Almanak’s recommendation actually generates more revenues (accumulated fees) vs keeping the fee portion as is?

Increasing the fee portion increases the fee per traded amount, but decreases the overall trade volume, hence potentially decreases the accumulated fees.

Decreasing the fee portion increases the overall trade volume, but decreases the fee per traded amount, hence potentially decreases the accumulated fees.

And there doesn’t seem to be any way of knowing in advance whether a pool will have accumulated more fees under the current fee portion, or under the Almanak’s recommend fee portion.

AB testing might be used as some sort of empirical (and retrospective) evidence, which may allow to determine after a few months whether or not it is worth relying on Almanak’s recommendation.

But unless the results of that are unambiguously conclusive, they would be a rather weak justification for continuing to apply Almanak’s recommendation.

1 Like

Hey @barakman @teamAsaf @eldude

Thanks for your comments, we deeply appreciate your input into the conversation!

I would suggest moving the discussion on the Almanak Performance to the report conversation so we can apply your feedback and suggestions:

Answering Asaf’s and Barak’s questions, there is formal verification of the performance: 2022_12_14_Bancor_Almanak_Update - Google Slides

I highly encourage you to check the slide: Almanak Fee Performance and Almanak Pools Performance.
We are comparing Almanak-managed pools vs non-managed in terms of the fees generated, normalized by the volume. Note that Almanak outperforms other benchmark pools on additional fees generated without cutting the volume. Unfortunately, AB testing is challenging and we agreed previously with the Bancor R&D team that it is not the best way forward.

The purpose of the proposal:

ICHI is among the top 5 biggest pools on Bancor, therefore adding it to the optimization trial period should be beneficial for Bancor DAO. As @mh89spd mentioned above, the increase in the pool volume will positively affect the BNT price due to an increase in fees which will affect more BNTs to be bought from the open market.

We are hearing your concerns about additional gas costs, Almanak has been told that the gas cost is a minor cost for the foundation. However, we can come back to ICHI and ask for this cost reimbursement as well. Alternatively, at the end of each month, Almanak can reimburse the foundation for the gas cost of ICHI updates. We anticipate that the gas cost of the TL and fees update will be around $250 per month per pool.

Looking forward to hearing your further feedback!

3 Likes

almanak slides do not clearly indicate an increase in fees over any test pool.
this on its own shows that they are unable to support the claim of improving the collected fees without hurting the volume.

perhaps i am naive thinking that recommendations like this can provide any additional value
“new swapping fee 0.0791% (from 0.5000%)”

1 Like

Hi @teamAsaf,
thanks for voicing your concerns so critically. We value your honesty and hope to further clarify the impact Almanak has on Bancor pools.
Furthermore, your feedback helps to additionally create more digestible reports and post clearer results.Let us dive into your last post

almanak slides do not clearly indicate an increase in fees over any test pool.

We tried to map this as visually appealing as possible but failed to do so. Specifically slide 7 of the report (check Ratio Fees/Volume for both clusters & the follow-up slide for volume) was made to cover this. Additionally, bear in mind that it is tough to spot it with the current market conditions. The effect will be amplified when market picks up. Since less volume is currently spent on DEXes, none of the exchanges can generate more fees to any comparable period, as fees depend on volume.

this on its own shows that they are unable to support the claim of improving the collected fees without hurting the volume.

Overall DEX market volume had a drawdown of 56% over the past 60 days. When comparing this to the loss in volume over the Bancor pools, we note a loss of 60.5% for non-Almanak pools and 65.4% for Almanak pools of Bancor. Note that we aggregate the Bancor numbers over the initial observation period (November 14th until December 11th) and average the numbers to compare the impact. We can therefore assume that Almanak was able to maintain overall market volume of the Bancor pools relative to the DEX average.
We now can have a look at the fee revenue per volume unit to check whether despite the lower volume more fees have been aggregated. We again compare Almanak pools and other pools with fees/volume on a daily basis. Both groups have seen an increase when looking at the pre-trial period of 90 days and the ongoing trial period. Other pools have 1.45x the fees per volume, whereas Almanak pools boasted 2.30x in comparison. All data is retrieved from the Bancor Analytics Dashboard under Daily Fees and Daily Volume.

perhaps i am naive thinking that recommendations like this can provide any additional value
“new swapping fee 0.0791% (from 0.5000%)”

We are happy to outline the idea of attracting volume to generate more fees vs. restricting value extraction through arbitrage, which is one of our key mantras in the trial period. Could you clarify your point made here? Do you think the swapping fees are too low/improbable?

Let us know whether you have any additional questions regarding this proposal or the general methodology in the other proposal here!

2 Likes

In general and not specific to the ICHI pool, I think we should wait until the test period is over and we have a final report before introducing new changes / increasing the scope.
I believe there’s about 6 more weeks for the 3 months testing period report to be out. I suggest waiting before introducing more overhead.

4 Likes

Hey @yudi thanks for the comment.

I think what you are suggesting is reasonable. Over the next 6 weeks, we will focus on analysing the performance of the currently optimising pools, so after the testing period, we would be able to provide a report without any additional noise. If everything is fine after the testing period, we can have further discussions on introducing changes.

Let’s part the proposal then for now and come back to it around 15th of February.

4 Likes

i believe this is a great suggestion.
before we add more stuff into a running test, lets validate the test results and let it run to the fullest.

2 Likes