Proposal: Implement Almanak’s daily trading fee and trading liquidity recommendations

** Expected on snapshot on October 9th **

Bancor DAO Almanak Proposal #1 -


Almanak, a revenue - optimization & risk management platform for DeFi & Gaming, has created a Protocol Economy Assessment Report for Bancor DAO.

Almanak would like to propose an ongoing long-term collaboration with the Bancor DAO, where Almanak would deliver regular actionable, data-based recommendations on protocol parameters. Approval of this proposal would grant the Bancor DAO multisig the permissions to implement daily fee and trading liquidity updates recommended by Almanak without a DAO vote.


Implement Almanak’s daily trading fee and trading liquidity recommendations for the ETH, DAI, LINK, and wBTC pools over the course of 3 months (ending December 31st, 2022) without a Bancor DAO vote and to be executed by the Bancor DAO.


Do not implement recommendations from Almanak.


Almanak is an institutional risk management & revenue-optimization platform, engaging in performance-based, long-term partnerships through the provision of data-driven recommendations and C-level intelligence to DAOs and management teams of blockchain-based projects.

Almanak uses agent-based modeling to simulate and optimize decentralized finance protocols towards sustainable growth, competitiveness and innovation. The intelligence it provides delivers economic security, while simultaneously maximizing key profitability metrics.

Learn more about Almanak by following it on Twitter and Medium.

Almanak proposes a collaboration with Bancor DAO to design, deploy and maintain a customized protocol risk management framework that will ensure the protocol’s sustainable growth and competitiveness. The goal of this proposal is to present Almanak’s engagement offer towards the DAO.

Bancor V3 Case Discussion

Bancor DAO recently disabled BNT distribution due to an emergency action taken to preserve the health of the protocol and users’ funds. Almanak has looked into the potential optimization levers of the Protocol in order to mitigate the potential risk of realized and unrealized vault deficits, and to maximize the Protocol’s revenues. Such optimization levers are (i) on-curve trading liquidity (TL) size, (ii) dynamic swapping and withdrawal fees, (iii) external liquidity protection pools sizes, and (iv) dynamic cool-down periods, .

Recognizing the changes within the Protocol, Almanak has shifted its focus towards an optimization plan for Bancor V3 to achieve the Protocol sustainability.

Outline for a Proposed Optimization Plan

The main objective under the current conditions should be to decrease the deficit of the already existing pools. Bearing in mind the recent Protocol update, we have focused on the analysis and optimization of the following elements:


  • Maximization of the Protocol’s revenue (protocol capital efficiency).
  • Minimization of the Protocol’s exposure to unrealized vault deficits.
  • Minimization of the recovery time of the Protocol’s deficit (protocol reactiveness).


  • Pool swapping fees - optimization of the fees per pool (dynamic or static) to maximize the Protocol’s revenues from trading.
  • Pool depths - optimization of the on-curve trading liquidity, in order to minimize protocol’s exposure to further deficit.

Almanak proposes to execute the recovery efforts, while being mindful about maintaining a secure and stable transition towards the protocol’s new state.

Scope of work

  • Provide recommendations on optimizations to the each token pool’s trading fees and available trading liquidity over an initial period of 3 months on four pools : ETH, LINK, WBTC and DAI
  • use the legacy split approved Bancor DAO (90:10) between fees used for the Vortex burner and TKN LPs.

Scientific Methodology

We leverage our agent-based simulation platform to identify the best possible sets of parameters to shorten the path to recovery of the protocol. Based on the 24h-volatility, Almanak’s platform simulates 7000 price trajectories to measure how agents behave in each of the scenarios.

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 protocol revenue and minimize pool exposure to deficit. The results are being presented and updated every 24 hours to act re-actively upon market conditions.

Simulation Results and Optimization Recommendations

Almanak conducted optimization simulations for four main Bancor pools (ETH, LINK, WBTC and DAI). As of the time of simulation (August 2022), these four pools represent an aggregated 67% of USD volume and 45% of transactions for WITH trades, and respectively 43% and 71% for FOR trades.

The scope can be extended to more pools as required by Bancor DAO. For illustration, below are presented the results charts for the ETH pool as well as the Daily Revenue and Burning Ratio comparative analysis for the aggregated four pools.

ETH Pool

Chart 1: ETH Pool Unrealized Vault Deficit

Vault deficit and represents the current delta between the value of assets deposited and staked by liquidity providers and the actual value available within the vault. Note that the difference between the two values was compensated via BNT distribution prior to 6/19.

To conclude, reducing the TL of a pool helps to reduce the impact on IL. Transferring funds out of TL into the vault secures LP assets, which in a first instance limits the protocol deficit. Secondly and acting with a bigger impact, these assets were only subtle to market changes as long as they were available within the trading pool. Not being part of TL anymore, LPs will need less compensation, as more assets are available.

Lastly, the following trend can be observed: the smaller the weighted TL size is, the higher the impact on the Protocol’s deficit. To do so, the Protocol would have to limit its TL in a volatile environment drastically to reach a smaller deficit.

Chart 2: ETH Pool Profit (net revenue minus realized vault deficit)

By the shape of the dark coloured cells one can observe a general tendency towards smaller pool depths and higher trading fees.

Extremes observed at the top right and bottom left corners represent highly volatile market scenarios which can be schematically split in two opposite configurations:

  • (i) when a market breakdown is imminent: profits can be maintained with high swapping fees and smaller TL.
  • (ii) when no crash is present: profits are generated through high TL and low swapping fees.

These configurations being extreme, they are less frequent in the sampled simulation distribution, therefore also indicating less robust optimization solutions. Indeed, sets at the bottom left represent less than 0.01% each on average and the top right is 0.05%. The most common and robust solutions are located in the center of the heatmap, with an average representative share of 1.2%.

Daily Revenue

Chart 3: Comparison of Daily Protocol Revenue – 5-Day Moving Average

The cumulative Daily Revenues over 30 days is USD 551,552 and USD 469,777 for the Optimized and Non-Optimized solutions respectively.

When moving to a high-level view, revenue on a daily scale is an easier path to compare two different solutions and their impact on a particular metric. In this case, Almanak’s solution outperforms the current Bancor solution on most days. The main reason behind this is the limited TL following a volatile market. While the current Bancor solution has always historically kept all tokens on-curve available for trading , the introduced risk levers help deploy funds tactically based on market conditions and produce lower slippage and better execution in these times.

Burning Ratio

Chart 4: Comparison of Daily vBNT Burn – 5-Day Moving Average

The cumulative vBNT Burn over 30 days is USD 1,180,113 and USD 1,004,115 for the Optimized and Non-Optimized solutions respectively.

As the vBNT burn is pegged to a fixed split of the incoming revenue (Chart 3), a high correlation of the burn of vBNT tokens with the latter can be observed.

Note that in V3 90% are used to buy BNT but they are not yet swapped for vBNT burning (not implemented at the time of writing. The protocol is currently collecting this “V3 BNT” in a separate wallet. Next chart below is showing the accumulation with and without Almanak optimization.

Chart 5: Accumulated V3 vBNT

The accumulated “V3 vBNT” over 30 days is USD 1,180,113 and USD 1,004,115 for the Optimized and Non-Optimized solutions respectively.


Overall, the current assessment focuses on what parameter combination limits the Protocol deficit, while increasing the profitability over the mentioned observation time. The main result brought forward by the Almanak’s protocol risk management framework is that the protocol profitability is a function of volatility and arbitrage volume, which can be actively managed through concerted tuning of the pools’ dynamic swapping fees and trading liquidity.


Initial Phase – On-boarding and Build-out

In an initial phase, Almanak would provide recommendations for LINK, WBTC, DAI and ETH pools over a 3-month period.

Recommendation would be provided daily on the following parameters:

  • Optimal trading fees per pool.
  • Optimal on-curve trading liquidity and strategy to shrink/grow the pool.

Execution of the Recommendations

Almanak will provide recommendations to Bancor DAO on the optimal trading liquidity and swapping fees size on a daily basis. We advocate daily recommendation as it brings better efficiency thanks to reactivity to past 24h market conditions. The main challenge is the implementation of the fee updates through the Bancor DAO’s weekly voting procedure that usually spans over 72h and would therefore significantly tamper the efficiency of the recommendations. We are seeking approval from the Bancor DAO to bypass the required weekly voting procedure and instead implement daily trading fee and trading liquidity recommendations for the ETH, DAI, LINK, and wBTC pools over the course of three months.

Almanak will collaborate with the Bancor Protocol to update swapping fees and trading liquidity on a daily basis. The recommendation will be reported and displayed daily on a bespoke Bancor dashboard (see illustration below), designed by Almanak. The ETA for such a dashboard is 3 months, pending Bancor DAO’s feedback and expectations. In the meantime until the dashboard is ready Almanak will be putting our recommendations in Bancor’s Discourse governance forums for the community to be kept up to date.

Further Collaboration

After the 3-month initial phase and assuming positive feedback from Bancor DAO, Almanak would provide new proposals extending the services with new features. Almanak envisions a transparent, performance - based, long-term collaboration aimed at supporting Bancor DAO with key governance decisions with regards to protocol economy and profitability improvements.


Thank you for all the research you’ve done and time you’ve dedicated to Bancor v3. A revenue - optimization & risk management platform is definitely something we can benefit from considering the current state of things. I’m looking forward to hearing from the rest of the Bancor DAO on tomorrow’s call and moving forward with this proposal. :clap: :muscle:


Posting the YouTube link from our DAO Discussion:


Action creates information

There was a story from Brian Armstrong when he was on Lex Fridman’s podcast a couple of months ago that I’m reminded of. In initially building Coinbase, he said it was like standing at the bottom of a foggy mountain, where you can see the peak but have no idea how to get there. In this case, action creates information. By walking forward a few feet, more of the path ahead is revealed and you can continue on. Sometimes, that reveals that you went the wrong way and you had to backtrack, but that’s still useful information.

The incentives are aligned here, as Michael and his team want to build the brand, which directly translates into a better Bancor. We benefit from a more efficient & profitable protocol (while alleviating the deficit) and Almanak gets to approach other DAO’s with the positive results. If there was a DAO that needed these services, Bancor is the perfect test case. The downside is also limited, as even outside of the three-month trial, we can pull the ripcord and be right back where we were.

I believe the three-month trial has an upside bias in that we all learn more through action, and can adjust properly, i.e., at the risk of sounding pollyannaish, I think through allowing Almanak to work they may discover more profitable and efficient avenues for Bancor beyond what the initial simulations are showing. This is mentioned in part on page 18 of their proposal, in that the more live information produced by running their products, it becomes more robust over time. From a layman’s standpoint, it makes sense that they also learn the effects of the daily recommendations themselves on trading activity.

The larger and less quantifiable aspect to this that excites me is the alleviation of the core contributors of Bancor to focus mindshare on things that they deem most important. What Almanak is proposing in essence a low-cost option of a third-party team directly focused on both the near-term deficit reduction, which is a dual benefit of increased profitability & efficiency, and the long-term sustainability of the protocol. What I want to highlight here is that it opens space for new creative ideas from the core contributors and the increased likelihood of innovative and creative improvements to the protocol.

Overall, this is a win-win scenario in that it shifts focus to a larger and motivated team to make Bancor better. I can’t emphasize enough how important it is to free up focus to specific groups, as the Bancor core contributors are allowed to focus on the most efficient goals, while Michael’s team has a clear goal of making the protocol better. There’s so much upside potential here that isn’t obvious to just simply alleviating the deficit. Great things can come about by giving people more room for focus and creativity.

I’m all for this.


I always appreciate your posts @PaperStreetCapital :white_heart:

I couldn’t personally agree more with this statement and the weight it carries:

It’s so important incentives are aligned, and I agree with your view; they are. Through bettering Bancor’s current position, Almanak is simply seeking the opportunity to prove themselves and show us that they’re competent and capable of providing us a sort of relief we haven’t been able to experience yet.

In response to this:

I’d love to see how this would play out. The charts provided in the proposal showed strength, and if new avenues were to be discovered throughout the 3 month trial period, it could potentially provide that much more of a positive impact on the current state of the protocol.


This is extremely valuable imo as well. Freeing Mark, Mike, Nick, etc to focus on the new AMM model, and have a group of experts running simulations & making recommendations, sharing the same goal as the Bancor DAO/Community, brings immense value. We are at a point now where, imo, the deficit can be approached from different angles, and I believe any potential upside outweighs any potential downside, especially seeing that the DAO can simply vote to stop the implementation of these recommendations should we deem them counterproductive or harmful.

These are all strictly my own personal opinions. I would love to hear the those of other DAO members as well…


One quick question. On the DAO Discussion call, we spoke about specific parameters for the 4 pools. Would you provide us with an idea of what those parameters might look like?

1 Like

Hi Jen, thanks for picking this up from yesterday!

Parameters are the ones that have been identified as “Protocol levers”, as in key drivers to optimize the objective function of maximizing revenue / minimizing deficit.

They are of two types:

  1. Trading fees: fee level expressed in bp charged per trade on the pool
  2. Pool depth: on-curve liquidity level expressed in % of the vault value

For #1 Trading fees, we can actually be more granular by even providing the updates at the level of the volatility factor and the arbitrage volume factor that are entering the trading fee formula.

As discussed over the call, we have built a Telegram bot available for anyone to track the latest recommendations. That bot can push the updates in whichever frequency, form and granularity chosen by the DAO. That could look like : “ETH Pool Update: trading fee 40 bps with a trading liquidity of 83% of the vault value”.

Hope that answers your question. All the best.


I can’t agree more and highly appreciate your answer.

In our ethos is to work on a performance basis therefore this 3 months period is crucial for us to measure our impact on the Bancor ecosystem and provide a suitable report.

We are also here for long term, we want to stay with Bancor DAO as long as possible and became an active DAO member. Having access to all data we process and knowledge we are more than happy to leverage it and participate in DAO proposals, because Bancor success is our success.

I am glad that you noticed the opportunity that now the core Bancor team can focus fully on innovation. I would like to mention that we also have in our roadmap sandbox platform that would allow protocols to test their new solutions in a close to reality environment in user friendly way. Currently the product requires case by case on-boarding but can be used if needed as well.


I am fully in favor of this proposal. I appreciate the amount of effort that those at Almanak have put into this proposal, and think that they could prove to be a valuable resource for Bancor moving forward.

I’m especially fond of the level of transparency described here, and think that the community would appreciate this level of visibility into constant protocol parameter changes.

This gives the Bancor team more time to focus on other things pertaining to the protocol, and offsets time that would have otherwise been spent investigating these various parameters of Bancor.

Thank you to the Almanak team for putting all of this together, and I am sincerely rooting for you. Your success is our success. I’d like to see you show the world what you can do.


Would like to see this proposal put to vote on Sunday, September 25th.


Hey @dirtyfrenchman - I fully support your enthusiasm! That being said, as it’s a complex topic and an important step for the protocol, I would encourage all contributors to take the time to go over the report and not hesitate to ask questions and challenge what we are proposing. We will all benefit from an enlightened vote process. All the best.


Thank you so much for your explanation. The Telegram bot is already built then? If the proposal were to pass, this would be available immediately for the DAO to follow?

1 Like

Thanks for the detailed proposal. I am very supportive FOR.

How will we evaluate the success and hence failure of the 3 month trial period?

For example, what if these recommendations have increased protocol revenue, relative to doing nothing,. but also increased the deficit exposure? And conversely, what if these recommendations have decreased protocol revenue but decreased the deficit?


I think this proposal is broadly helpful as an initial trial. Let’s bring it to snapshot as soon as possible.

1 Like

Hi @cube – Many thanks for your feedback and question.

As of now, there are two ways we plan to assess the performance our of recommendations:

  1. Impact on profitability, defined as the ratio of Revenue to Deficit: “success” will be deemed if the latter has increased over 3m by a rate of growth higher than the average rate observed prior to Almanak launch. Since it’s a ratio, we may also need look into the behaviour of the numerator and the denominator to check whether they evolve in the right direction on average over 3months.

  2. Comparison of “Almanak fee” to a “Benchmark fee”: Almanak fee is the recommended fee. Benchmark fee is the “best” trading fee that Bancor would be able to charge in order for the pool to remain sustainable, as in properly balanced, without consideration of trading liquidity levels. The notion of “best” is then to be understood as “given the market volatility and the external liquidity required to arbitrate the pool price discrepancy”. When computing the benchmark fee volume is assumed to be the actual observed volume. Doing so, we can assess the capacity of Almanak fee to adapt to the market “cost of liquidity”. If the Almanak fee is diverging too much from Benchmark over a given period of time, then a performance breach could be triggered.

There’s a third way, still at a research stage for now, which consists in simulating a whole environment of Bancor without Almanak recommendations and compare its profitability and protocol levers with the production ones (i.e. integrating Almanak’s recommendations).
This third avenue is objectively the most complete one although arguably the most complex one as well. In particular, simulating the volume that would prevail in a world without Almanak is a challenging task, although not impossible if model bias is properly managed. More research is required today.

1 Like

Yes Telegram bot is operational and can be parameterized to be up and running day one. :slight_smile:

1 Like

What potential risks are there to the protocols funds if this proposal is passed?

I’m trying to wrap my head around the idea behind this proposal, and it seems like this is something that bancor should explore. So I am in favor of the proposal, but I would also like to be aware of any potential risks, if any.

( I am in no way insinuating that Almanak will act maliciously with the example below, I just want to understand the risks )

Is it possible for a situation to occur where the protocol potentially opens itself up to vulnerabilities that can be exploited by a nefarious actor, if Almanak is given the rights to move funds/increase/decrease liquidity in a pool?

or are there any potential risks that the community / DAO as a whole should be aware of?

Hi @Jindo and thank you very much for your question.

Security of the protocol is major topic that we consider very seriously at Almanak.
As far as Almanak is concerned, it requires Almanak to follow and operate by the protocol security guidelines and framework already in place. I invite the owners of the latter to comment specifically on this if need be.

That being said, several comments can be made here in regard to your concern.

  1. Almanak will “only” provide recommendations, Bancor DAO MSIG will continue to own the execution of the changes. Almanak will never be able to move any funds. So in effect, Almanak is not adding supplementary risk nor increasing the protocol attack surface.
  2. Almanak’s recommendations on Trading Liquidity level will not imply nor require a different smart contract update process from the current one. Again Almanak is not adding supplementary risk nor increasing the protocol attack surface.
  3. At any point in time, Bancor DAO can vote to unwind the optimization service and recover to original situation.

I hope this covers your interrogations.
All the best.

thank you for getting back to me,

What I gathered from the call, is that there is a possibility where the DAO may vote on a proposal that, if passed, will grant Almanak the right to make changes to the trading liquidity levels as they see fit (for maximizing profitability etc…)

Assume that this vote takes place, and is passed.

If liquidity in a specific pool is either decreased or increased, is it possible, in any scenario, where a nefarious actor is able to capitalize through the changes in liquidity levels, and gamify the protocol and more easily extract value /TKNS from the protocol?

I don’t know the in’s and out’s of AMMs, so I apologize if my question comes off as ignorant. I’ve simply been in this space long enough to witness enough exploits in defi via bugs, flash loans, new implementations (i.e. altering liquidity levels which bancor has never done before). So even though I am enthusiastic about Bancor working with Almanak, and I have high hopes for a successful relationship, I also want to be made aware of any potential risks before placing my vote.

Hey @Jindo
Yes, you gathered correctly about the vote.
That’s a great question I am happy you ask. So let me break it down for you.

Short answer:
We integrate the effect of changing on-curve liquidity on the trading activity by adjusting the trading fee accordingly, thereby also monitoring the arbitrage volume. Since the exploit you are referring to essentially put in play arbitrage profit seeking, we consider that our recommendations will optimally limit this risk.**

In the current state of our optimization analysis (as it appears in the report), the protocol levers are trading liquidity (TL) and trading fees (TF). They are treated in combination by the optimizer so that the effect of increasing/decreasing TL is controlled for in the recommended TF level. This is done by monitoring the volatility and the arbitrage volume.

To act swiftly, the optimizer checks if and by how much the volatility has changed over the past 24 hours and adapt the trading liquidity accordingly. Any change in the TL is therefore linked to the level of volatility which also enters into the TF function. Additionally, as we want to control for the effect of TL change on trading activity (and trading behaviours like the one underlined in the question), the optimizer monitors the volume and the revenue share of arbitrageurs that are acting on prices discrepancies. Based on how large the share is and the recurrence of the trades, TF will adjust and impact the number of incoming trades (the higher the fees, the lower that number).

A remark on MEV:
During our assessment we used a dataset comprised of 30days (over July-Aug 2022, (dataset here)). This dataset didn’t show any clear sign of MEV (see here for a definition). Therefore, MEV is not taken into account in the results presented in the report. However, should we identify proof of MEV we will incorporate a factor to control for the effect on profitability.