Opening Hearing on reWARDS Committee

[Chuck]

The reWARDS Committee was established on Jan. 2022. Having operation for half a year, the reWARDS Committee kept working on a monthly budget proposal and quarterly report, which made the process more transparent and efficient.

In the past couple of weeks, I tried to make a deeper dive on the LPs incentive performance based on the data and internal community dialogues with some of the members. Although the mapping became much more clear, there are still some unanswered questions in my mind. As a community member, I think it is my duty to raise these questions openly and hope you can help me find out the answers.

Q1: What’s the basic strategy of LPs incentive plan on Ethereum? What’s the mechanism of calculating the overall stETH-ETH liquidity demands? Is there a segmentation strategy based on different expectations, like bottomline expectation(for low volatility), baseline expectation(for middle level volatility), and lifeline expectation(for critical volatility)?

Q2: How do you manage the LPs’ rewards expectations? Is there a strategy like pessimistic expectation, realistic expectation and optimistic expectation strategy?

Q3: What’s the balance mechanism between liquidity utilization and rewards demands? As the rewards increase, more liquidity may join the pools and reduce the liquidity utility and make the overall LPs’ APR kept at the same level.

Q4: Is there an LPs segmentation framework based on your data analysis?

I think these questions are important for the long-term LPs incentive operation. I hope I can be a better community contributor with your help.

@jbeezy @GrStepanov @kadmil @kai

[kadmil]

Thank you for the questions! Let me describe what the current thinking behind LP rewards on Ethereum has been and would be in the nearest future first, and then I’ll try to answer your specific questions.

What’s the main objectives of LP rewards on Ethereum have been:

1. Get stETH stakers the opportunity to exchange their assets back to ETH.

2. Get liquidity diversified enough for resilient price feeds to be possible.

3. Secure enough liquidity for potential liquidations on money markets.

As a secondary objectives, we were looking to 1) secure higher trades volume (“liquidity utilization”); 2) lower the spend without major drastic motions from our side. Basically those goals were aiming to get TVL first & ensure the exit option for ones who want to “unstake” before withdrawals are enabled.

Another thing to note, reWARDS doesn’t (and, in my opinion, shouldn’t) wield resources to fight with extreme market conditions we saw lately. Basically that means we can’t get significantly more ETH in the Curve pool with any amount of incentives we can reasonably allocate there, and don’t want to “allocate unreasonable amount of incentives” either. The market conditions are calling for change in our thinking about LP rewards.

So, with the big outflow of ETH from the (w)stETH pools & the high volatility of the market, I’d say that the focus of the reWARDS on Ethereum should be:

1. Secure enough ETH/stables liquidity for potential liquidations on money markets.

2. Ensure “price resiliency”: get the liquidity diversified enough so the amount of trades required to push the price down, say, 1% is high. Note that it’s not to target specific price, but to make the price stETH does have at the moment more stable.

3. Get liquidity diversified enough for resilient price feeds to be possible.

4. Lower the spend where possible (across all the LP incentive programs).

Again, liquidity utilization is important bonus objective here.

Hope that gives us the shared mental model.

[kadmil]

Q1: What’s the basic strategy of LPs incentive plan on Ethereum?

A1: Basically we need “at least” some multiple ETH/stable liquidity to cover liquidations of the most risky positions on money markets as the ballpark of total liquidity in the pools. On higher volatility this number would go higher due to more risk of sudden price movements, so the optimization for “price resiliency” is important as well. Not sure what you mean by “segmentation strategy” here.

Q2: How do you manage the LPs’ rewards expectations?

A2: We communicate the planned rewards amounts for the pools in monthly budgets to both 1) communicate the plans & get community feedback; 2) signal LPs what the amount of incentives would be in the next month. Should note, we don’t change the numbers drastically in single period (“no drastic motions”), so not to surprise LPs & cause massive liquidity outflow. All the communication is public — it’s exactly the budget posts & numbers you can see on AMM interfaces where reWARDS sends the incentives to.

Q3: What’s the balance mechanism between liquidity utilization and rewards demands?

A3: As I’ve mentioned, liquidity utilization is important thing we measure, but don’t promote specifically. Most of incentivized liquidity is in pools to ETH, the price there is quite stable → not much arb opportunities. The liquidity utilization has been significantly higher on pools with stables, but the liquidity amount was way lower & ~5x more expensive. For ETH pairs the benchmark is staking APR — little sense to hold liquidity in the pool if the APR is significantly lower than the staking APR one can get just by staking & holding stETH.

Q4: Is there an LPs segmentation framework based on your data analysis?

A4: We don’t segment LPs as we don’t have specific communication channels there. All the incentives are planned & distributed in the open, and I can’t tell how could we benefit from making that process more layered, tbh.

[vsh]

Kadmil’s answer is entirely correct re: stETH incentives. But one thing to note is that the majority of incentiviztion program is on Ethereum and for stETH, but it’s not the only one. E.g. stSOL incentives used to have a very different strategy of spray and pray to make stSOL widely usable in Solana defi.

[Chuck]

Appreciate your reply, which makes me more clear on the targets and base rules how the reWARDS Committee works.

Beyond what you have mentioned, there are still some unanswered questions, especially about specific mechanisms.

Q1: What’s the basic strategy of LPs incentive plan on Ethereum?
A1: Basically we need “at least” some multiple ETH/stable liquidity to cover liquidations of the most risky positions on money markets as the ballpark of total liquidity in the pools. On higher volatility this number would go higher due to more risk of sudden price movements, so the optimization for “price resiliency” is important as well. Not sure what you mean by “segmentation strategy” here.

“Segmentation strategy” means what the specific “some multiple” number is settled and what’s the variable numbers that will trigger the different level of volatility. For example, if ‘stETH to ETH average discount’ is the trigger, and ‘under 1.5%’, ‘1.5~3%’, ‘3~5%’ might use different strategies to match the demands. As you have done so much research and data monitoring in the past half a year, you might have these mechanisms to make the rewards setting more scientific.

Q2: How do you manage the LPs’ rewards expectations?
A2: We communicate the planned rewards amounts for the pools in monthly budgets to both 1) communicate the plans & get community feedback; 2) signal LPs what the amount of incentives would be in the next month. Should note, we don’t change the numbers drastically in single period (“no drastic motions”), so not to surprise LPs & cause massive liquidity outflow. All the communication is public — it’s exactly the budget posts & numbers you can see on AMM interfaces where reWARDS sends the incentives to.

Historical actions may give us some guides to the future, but things always change faster than we expected. That’s why specific mechanisms matter. LPs’ rewards expectations must have an outside reference system. It might be the Curve ‘renBTC-wBTC pool’ APR because of the same level of risks, or it might be the ‘sETH-WETH pool’ or the ‘AnkrETH pool’ APR because of the direct funds competition.

Q3: What’s the balance mechanism between liquidity utilization and rewards demands?
A3: As I’ve mentioned, liquidity utilization is important thing we measure, but don’t promote specifically. Most of incentivized liquidity is in pools to ETH, the price there is quite stable → not much arb opportunities. The liquidity utilization has been significantly higher on pools with stables, but the liquidity amount was way lower & ~5x more expensive. For ETH pairs the benchmark is staking APR — little sense to hold liquidity in the pool if the APR is significantly lower than the staking APR one can get just by staking & holding stETH.

This question is related to ‘Q2’, and tbh, I can’t support the view that ‘little sense to hold liquidity in the pool if the APR is significantly lower than the staking APR one can get just by staking & holding stETH.’. Holding stETH will always have a synthetic layer risk and stETH to ETH discount volatility risk. Provide LP gives them a chance to take less risk exposure on stETH, as well as addition return generated by transaction fees and incentive rewards. If the ‘benchmark’ means ‘cap’, that’s more reasonable.

Glad to see more discussions on these questions.

[vsh]

@Chuck

Historical actions may give us some guides to the future, but things always change faster than we expected. That’s why specific mechanisms matter. LPs’ rewards expectations must have an outside reference system. It might be the Curve ‘renBTC-wBTC pool’ APR because of the same level of risks, or it might be the ‘sETH-WETH pool’ or the ‘AnkrETH pool’ APR because of the direct funds competition.

Planning for a specific APR is futile. Market will decide how much APR are people comfortable with in that particular pool. The only one exception at scale with a target amount of APR was Anchor haha.

@Chuck

“Segmentation strategy” means what the specific “some multiple” number is settled and what’s the variable numbers that will trigger the different level of volatility. For example, if ‘stETH to ETH average discount’ is the trigger, and ‘under 1.5%’, ‘1.5~3%’, ‘3~5%’ might use different strategies to match the demands. As you have done so much research and data monitoring in the past half a year, you might have these mechanisms to make the rewards setting more scientific.

One thing to note is that we have operational resolution of about 1 month for liquidity incentives in normal weather. Having a regime where we quickly react on market shifts in a particular way is problematic - it’s more operationally exhausting, and it’s bound to attract only mercenary capital that is fast at reallocating instead of more passive LPs.

[Chuck]

What I’m interested in is the logic and mechanism behind all these monthly budget numbers. Operation is not just an art, it’s also a science, especially for multi-millions monthly budget management. As the Committee has built up many data monitors, I think their budget must be related to these data. I don’t think these mechanisms should be viewed as business secrets and let the community give feedback just based on their feelings.

I would like to see more reWARDS Committee members join in this discussion. @jbeezy @GrStepanov @kai @Ivan @McNut @adcv

[kadmil]

Mechanic followed from reWARDS solving for business objectives & correcting things which hadn’t worked. “Hadn’t worked” — have not attracted much liquidity / had low amount of trades. On top of that — lowering rewards where possible (“pool performs good enough, let’s test APR sensitivity”). Given the significant LDO price volatility, monthly cadence for those decisions doesn’t allow for “finer tuning”. At the same time, “finer tuning” = orders of magnitude more operations & comms, and those taken from all the other activities DAO should have & perform.

No “business secrets” and for the changes we provided reasoning on the monthly budget forum posts — that’s about as transparent as possible imo.

[kadmil]

As I’ve said, given the stark market turn we need to rethink the goals of incentive programs. That would affect both the goals & the “thinking framework” of the reWARDS operation. While we’re in contact with the number of teams specialising on incentive management, but without cleanly defined goals no amount of automation and insight gonna help achieving those.

[Chuck]

Ok, let’s combine these related discussions in this post.

Let’s do a little napkin math. Lido wields ~25k ETH in Treasury (the exact distribution & targeting for the funds is nuanced, but let’s skip that discussion). By the current price, it’s ~$30m. reWARDS monthly budget ~4.5m LDO, $2.6m, making the whole Treasury worth 1 year of incentives, barring all the other costs. At the same time, Lido Agent holds ~$94m worth of LDO, making this significantly better source for incentives — as long as Lido does need to provide those.

How did you work out the number ‘$2.6m’ or ‘4.5m LDO’ is needed? What if the $LDO price goes up 3X tomorrow, does that mean the whole Treasury is worth 3 years of incentives?
Considering that our incentive is targeted on steth-weth liquidity pool and there is huge volatility on LDO to WETH exchange rate, I have proposed that using WETH or stETH instead of LDO for LP incentive, which would bring more stable return to the LPs. How do you draw a conclusion that ‘making this(using LDO as incentive token) significantly better source for incentives’? Have you ever thought about whether ‘$2.6m’ is actually needed?

[kadmil]

Conclusion is “we don’t have enough ETH to cover incentives”, and “we have to manage LP expectations and not throw those away on a whim”. We were working on bringing Curve pools rewards (the biggest expense of the reWARDS budget) down, managed to cut it two-fold, LDO-wise, without having “incentive-related” disruptions of the pool.

You’re trying to solve for “let’s not sell LDOs”. I’m trying to solve for “let’s achieve overall goals while trying to sell as little LDO as possible without causing major disruptions on the way”. While — as I noted already — I’m much, much pro spending less LDO on incentives, “let’s just switch this all off” or “let’s burn the way lower reserves of ETH instead of LDOs, no matter the other costs” doesn’t cut it.

[Chuck]

So, using LDO instead of WETH as an incentive is just because that LDO is enough spending, right? What if the LDO to ETH exchange rate dropped to 50% of the current, still enough at that time?

I am not quite sure what makes you think that my goal is ’ trying to solve for “let’s not sell LDOs”', kindly tell me sir.

[Chuck]

Here is the “LDO APR” on Curve stETH-ETH pool in 2022.

Dune

Does this look like anything related to ‘a stable LP expectations’?

[Chuck]

What’s the main objectives of LP rewards on Ethereum have been:

1. Get stETH stakers the opportunity to exchange their assets back to ETH.
2. Get liquidity diversified enough for resilient price feeds to be possible.
3. Secure enough liquidity for potential liquidations on money markets.

Let us go back to the main objectives and use the data on Curve stETH-WETH pool as an example.
[17,484 stETH sold transactions since 2022-01-01]

Ok, that is the full map, which X-axis represents the amount of stETH been sold, Y-axis represents the price impact(or ‘slippage’) by specific transactions.

Wholly sh*t! Their is a sign transaction with more than 80K stETH being sold. We do not object to matching all the exchange demands, right? Let’s zoom in with the top 2% largest stETH selling excluded.

All right, we can see that set ‘under 1200 stETH exchange’ as a benchmark could cover 98% of all demands.

Let’s zoom in again with the top 5% largest stETH selling excluded.

Ok, a ‘under 400 stETH exchange’ benchmark could cover 95% of all demands.

[Data sources]
https://dune.com/queries/973761

Based on the upon data, don’t you think ‘to enable stETH holders to exchange less than 1200 stETH to ETH with less than 0.2% price impact on one transaction’ could be set as the benchmark on ordinary market volatility? The benchmark could even decrease to ‘400 stETH’ while the market volatility is low. Do you still think it is necessary to support a 600K stETH pool with 500 times larger than the benchmark?

Let’s take a look at the liquidations’ data on Aave v2.

https://dune.com/queries/974143

Not a single liquidation surpassed 1000 stETH, even in the past couple of weeks. There are only 2 among 300 liquidation transactions, surpassed 500 stETH.

Don’t tell me that you set the benchmark based on the largest stETH collateral account with 400K stETH. If so, I don’t think we can ever match that kind of demand.

All these are what I am supposed to hear from you reWARDS Committee members. If that is so called ‘“let’s not sell LDOs”’, then I really expect you guys should think about how to “let’s not sell LDOs”.

[vsh]

Targeting constant APR is a fools’ game. Stable APR is not a sign of a good market (where APR reflects price risk tolerance of LPs that changes over time).

We can ensure stable LP expectations in terms of the overall pot of incentives available to LPs, but not in terms of APR which reflects other LPs decisions.

[Chuck]

Yes, no one could dominate the APR by force. But that doesn’t mean that APR should not be used as a figure while making an incentive plan. I think that’s why so many dune dashboards have included various APR(LDO APR, Fee APR, CRV APR, Total APR etc.). All these discussions are to make a clear view on how these data were used in the rewards budget draw up process.

While in the emerging stage, like Lido on Solana, it may not be a big deal to ‘spray and pray’. Everything is worth trying. But on Ethereum, we have reached 4M $ETH staking market share and built up so many data monitors. It’s time to transform the incentives on Ethereum to a fine operation model.

[McNut]

@Chuck I greatly appreciate the data here and will review it.

I think evaluating stETH liquidity in terms of price slippage is a much more meaningful metric than TVL / volume. Ultimately, TVL / volume are just leading indicators of what price slippage may exist on an exchange.

That being said, I believe the rewards committee is hoping to mitigate the stETH / ETH price variance with higher TVL / volume, which is a justifiable cause for Lido.

However, I think it may be optimal for Lido to transition (slowly) to utilizing price slippage as a more objective metric for stETH liquidity. There are many other factors influencing the price of stETH / ETH that Lido would be hard pressed to overcome (I’m working on this analysis).

Its worth debating whether Chucks data shows that the generous LDO rewards may simply be creating excessive liquidity for one off whale txs and may not be impacting stETH/ ETH prices.

[Chuck]

I appreciate you joining this conversation. Glad to see your analysis in the coming days.

I tried to find out whether there are connections between the ‘LP total APR’ and their liquidity management behaviors.

[Tips:] Numbers more than 10K on the Y-axis are assigned as 10K for simplifying.

Dune

From the above chart, we can see that the assets(ETH and stETH) added and removed from the pool are somehow balanced under most range of total APR. Even on the lower level(3.5~4%), there is no sign of liquidity pulling out.

From this chart we can also find out that most period among the past 6 months, liquidity utility kept in quite a low level. Even in April when the total APR is in decline, the TVL(in ETH) is still on an up trend. It’s not quite clear whether these gross was mainly driven by leverage leading, or other factors.

[jbeezy]

@Chuck and @McNut These are some of the things we are starting to work on internally. As mentioned in the [reWARDS] July ‘22 Budget we are beginning an update to provide better transparency from an accounting perspective with public access to the various addresses used for incentives by our partners as well as updating the KPIs for the committee.

Some initial modeling examples we are working on:

• Amount of trades / liquidations could absorb with X% price drop
• amount of liquidations required to push stETH price 0.5%, 1%, 3% lower
• Amount of ETH in pools, relative to stETH
• Trading volume
• Live dashboard
• Models with changing incentives amounts (stretch goal)
  eth(or stables) tvl

I would welcome for either/both of you to submit a proposal if you can help with the workload or simply join the working group that is helping to build this out. We would love to solve all problems at once but we have to constantly prioritize and, every week there is another crisis in crypto so it is easier said than done. It is very easy to point out where we could improve process and we are painfully aware of most of them. The difficult part is getting extra hands to help with the lift.

• If we are overpaying for liquidity that would be amazing to reduce as a cost center.
• If we can come up with a clear strategy to convert LDO > stETH as the rewards currency, we need the operational framework for this.
• If LPs are interested in a ‘more stable’ but not fixed APR, what would that be? Can you introduce us to large LPs for customer development. Most are anonymous so this is tricky.

[carvas]

Jbeezy put it perfectly. Just as a follow-up on this, we’re aiming to have concrete outputs of this work over the next couple weeks and will be able to publish good parts here on the forum for transparency and feedback.

Also open to DM if you have other ideas!