Tier 4 Fund Management Discussion

[Nagalim]

“If either group has <4% of their respective marketcap, shareholders grant up to 5%.”

This is right, I forgot to add that.

I’m not sure about the limits given in 2.) and 3.).

I edited to add ratios below a certain total liquidity. I’m so down to talk more about these limits, this is what I wanted to get to: the meat of the multisig.

As far as time windows are concerned, I think that should be mitigated by use of multisig and human awareness. I also left out mention of velocity, i.e. how many funds are withdrawn from the multisig address how quickly to deal with the problem. For example, if the buy side is low for an hour, perhaps $100/hour are used whereas if it’s been a day of broken pegs we ramp up to $1,000 an hour.

[masterOfDisaster]

I’m so down to talk more about these limits, this is what I wanted to get to: the meat of the multisig.

Sorry for being a pain in the ass, but I sincerely only want to have the best start for the next steps and limits play a role for that.

As far as time windows are concerned, I think that should be mitigated by use of multisig and human awareness. I also left out mention of velocity, i.e. how many funds are withdrawn from the multisig address how quickly to deal with the problem. For example, if the buy side is low for an hour, perhaps $100/hour are used whereas if it’s been a day of broken pegs we ramp up to $1,000 an hour

I agree that some flexibility helps in cases where static rules would hinder effective mitigation.
On the other hand that puts a lot more responsibility on the shoulders of the fund keepers. Over time and based on experience some borders should be defined.
I bet that shareholders will do that as soon as they feel in the position for that - which is not at the start of this operation.

What to do with park rates? Is that a useful recommendation for shareholders?

Interpretation of the liquidity tiers, a waterfall model, triggers, metrics and actions

if the (average sum of tier 1 to tier 3 in the last time frame x) sell side liquidity is bigger than buy side, NSR holders are expected to raise parking interest above 0%.

Or did you leave it out, because it has nothing to do with management of funds?

[Nagalim]

I don’t believe T5 should have to do with T4 or T6 in the same way that T4 and T6 are independent of each other. T1-3 follows waterfall, T4-6 are independent reserves.

[mhps]

I don’t believe T5 should have to do with T4 or T6 in the same way that T4 and T6 are independent of each other.

T5 is conceptually the same thing – spending money to control liquidity. It just has different time-scale and way better decentralization and way lower operational complexity and cost. It’s the preferred tool over NSR grant for mid-short term (days to weeks) buy side support.

[dysconnect]

The catch is that parking itself isn’t a very good mechanism that connects its utility to the network and cost as pointed out by @Benjamin.

I have been vocally cautious with T1 liquidity and not a big fan of reliance on parking rates specifically for defending the peg. That said I don’t want to throw away T5 at the current state either, because it’s still cheaper than Tier 1, which has its own other problems.

While I talked about replacing the parking mechanism with deferred transactions I have been too busy for some time to write and think carefully about it…

[mhps]

The catch is that parking itself isn’t a very good mechanism that connects its utility to the network and cost as pointed out by @Benjamin.

His arguement was that parking in its current practise currently can cause an open end obligation to pay interest in the future. If it is only used for mid-short term (days to weeks) buy side support, for example by a motion to never pay interest for parking longer than a month, then the risk is gone and the benefit stays.

[Nagalim]

We set up a multisig NSR address (T6). This group is given 5% of the NSR supply (42mil NSR) with the instructions to sell a portion of them for NBT if buy side liquidity in T1-3 goes below $80k while sell side is above $100k (or if buy is <40% of total and total is <$100k)

What do people think about this? I think this would be a good first step.

Perhaps we should say that the group should use an increasing market velocity of $5/hour for every hour the above conditions are met. So that would mean that 10 hours of that condition will result in $225 of NSR going to market. 24 hours would result in $1,500.

[masterOfDisaster]

I don’t want to throw away T5 at the current state either, because it’s still cheaper than Tier 1

His arguement was that parking in its current practise currently can cause an open end obligation to pay interest in the future.

Paying for tier 1 liquidity is an open end obligation to pay as well. Liquidity needs to be paid for immediately, parking rates in the future.
Raising parking rates only in case of danger for the peg is less reliable than activity on tier 4 and tier 6.
Is parking rate really more expensive than paying for tier 1 (sell side) liquidity?

NSR holders can find out how many NBT get parked for x% per month and compare it with compensation for tier 1 liquidity.
If you pay 10% per month for $x (tier 1 sell side) liquidity, you have 1.1 times $x at the end of the month - the same is true for paying 10% interest for 1 month parking.

My gut feeling tells me that in the end “emergency short-term insanely spiked interest” is more expensive than offering decent interest continuously and not necessarily more expensive than compensation for tier 1 liquidity.

This is a rather extended side note, but I wanted to shed light on my perception of it.
My understanding is that

• parking rates (tier 5) should have a mid-term to long-term effect,
• NSR sales (tier 6) a short-term to mid-term effect and
• BTC sale (tier 4) an immediate effect on reducing the sell side.

If you remove NBT in excess from the market (paid for liquidity or interest) by exchanging them for NSR in a continuous process, you likely get more NBT for the NSR compared to selling NSR in a situation in which the peg is under pressure; which means it’s cheaper for Nu.
Selling NSR in case the peg is threatened is still possible.
Maybe my thinking is flawed, though.

We set up a multisig NSR address (T6). This group is given 5% of the NSR supply (42mil NSR) with the instructions to sell a portion of them for NBT if buy side liquidity in T1-3 goes below $80k while sell side is above $100k (or if buy is <40% of total and total is <$100k)

What do people think about this? I think this would be a good first step.

I agree.

Dealing with an hourly increasing market velocity might be the hardest part of that, because it would require frequent activity of the custodians. I doubt that is possible.

Or do you mean selling NSR once a day with the resulting $ of the formula $5/h? That should be possible, especially if different multi signature addresses are used. For managing only several hundred to a few thousand $ value 2-of-x or 3-of-x should be fine. The big amounts should require more custodians and a majority of total signers.

And once again I think how beautiful it would be to have seeded auctions for that, send NSR once a day to the auction, put a current status (=NSR market rate) and the remaining time of the auction on the auction website or on nubits.com and exchange NBT for NSR that way!

The ability to bring tier 6 funds to market needs to be delegated to custodians (the future tier 4 fund managers?). Shareholders can grant NBT or NSR, but the process to sell them is not ready yet.
This would be a start in the right direction.

@Nagalim’s proposal is simple and efficient.

What do the other potential tier 4 fund managers think of that?

[Nagalim]

Right, it wouldn’t require a sell every hour, just a nominal way to calculate how much to bring to market, whether that occurs once per day or whatever. The precise numbers can be fuzzy because the signers are human and can account to the best of their ability and a broken peg has no strict definition. There will need to be a strict protocol, of course, whereby funds are brought to market.

Seeded auctions will easily take advantage of these rules, we don’t need to worry about how that will fit in now.

[mhps]

Paying for tier 1 liquidity is an open end obligation to pay as well.

You get it wrong. If you set 1 year interest to 120% pa (10% per mo), the protocol will pay 10% per mo after 11 months to those who parked. Park rate can have 1 year 5 year 10 year terms… There is no way to stop paying interest to those who have parked short of a hardfork. That long term open end obligation was what @Benjamin was talking about. If you set liquidity interest to 10% per mo, you can stop it with a motion any time.

Is parking rate really more expensive than paying for tier 1 (sell side) liquidity?

No. Park interest is like 1/10 of T1 cost. See posts starting from this in “Park rates are our method for short term peg maintenance”.

parking rates (tier 5) should have a mid-term to long-term effect,

No park rate should be used only for mid and short term to avoid obligation in the future. Specifically rate longer than 1 month shouldn’t be offered.

NSR sales (tier 6) a short-term to mid-term effect and

Burnt NBT are by definition effective forever. Once its burnt it’s gone.

[masterOfDisaster]

If you set liquidity interest to 10% per mo, you can stop it with a motion any time.

I might have explained it badly. Exactly that was what I had in mind: parking interest for short periods of parking, but which are continuously offered.
Nu currently has per minute compensation for ALP and per month compensation for MLP.
Why not offer parking interest for one month parking period?

Specifically rate longer than 1 month shouldn’t be offered.

I agree that the parking period should be short and in case the incentive needs to be increased an increase of interest rate should be the means.

Burnt NBT are by definition effective forever. Once its burnt it’s gone.

And once again I might have chosen the wrong words. I’m aware of the lasting effect of burned NBT.
By categorizing

• parking rates (tier 5) as mid-term to long-term,
• NSR sales (tier 6) as short-term to mid-term effect and
• BTC sale (tier 4) as immediate

I didn’t mean how long the effect lasts, but instead the speed with that an action is effective on sell side reduction.

Raising parking interest takes some time to kick in.
But offering paring interest for one month parking and removing 10,000 NBT from sell side this way might be cheaper (per month) than paying compensation for reducing tier 1 sell side by 10,000 NBT or increasing buy side by 10,000 NBT value.

For the dilution of NSR (that might be necessary to remove NBT from the market) it plays no role whether the NBT entered the market for paying parking interest or tier 1 liquidity compensation.
I say Nu doesn’t know what’s cheaper at the moment and should try to find out.

Why not offering 5% monthly parking interest (with one month duration) if tier 1 liquidity compensation is double of that rate?

[Nagalim]

Pretty please can we stop talking about T5 here?

<80k buy and >100k sell activates T6, as does >60% sell when total liquidity is <100k, also if nbt is being sold for <$0.9. Acceleration is 5nbt/hr^2.

Are these parameters acceptable?

[mhps]

Pretty please can we stop talking about T5 here?

It’s just to understand the full picture or why are you talking about T6?

Are these parameters acceptable?

I think trading volume (how fast the wall is being eaten) should be considered.

[Nagalim]

T6 is multisig, T5 isn’t. I’d be happy to talk about T4 too, it will follow very similar rules.

The trading volume is the 5nbt/hr^2. Maybe we should do 2 nbt/hr^2? That would mean $600 in the first 24 hours and $1,752 in the second 24 hour period.

For T4, we can do:
<20k buy and >40k sell activates T4,
as does >75% sell when total liquidity is <50k,
also if nbt is being sold for <$0.9.
Acceleration is 5nbt/hr^2.

Maybe we should have a maximum velocity of $5,000/day or something. I’m also trying to think of a more smooth trigger, like:
<25% when total is <50k
<30% when total is <70k
<35% when total is <100k
<40% when total is <140k
<45% when total is <190k

I would have an absolute nerdgasm if we could use this equation:
Threshold = 50%tanh(x/150,000)
Where x is total network liquidity
And tanh is the hyperbolic tangent, a complex function used in relativity, among other things.
http://wolframalpha.com/input/?i=50tanh%28x%2F150%2C000%29&x=0&y=0

[Nagalim]

Threshold % = 40(1-20*arctan(x/20)/x) where x is the total liquidity in kNBT:

So the way it would work is that the signers would take note of when the threshold is passed and start counting in 2 nbt/hr^2. Once the signers are obligated to spend some unit of btc (5 btc?) a signer will make a tx and pass it around for everyone to sign.

[mhps]

That tan() looks too much like a hack.

How is this for a model, in layman understandable terms: if buy or side is going to reach 25% total liquidity in 24 hours, fire fighting mode is on.

To realize the above, I suggest that at any moment T0, use a 3rd order polynomial (a t³ + b t² + c t + d) to fit liquidity data in the last 48 hours, sampled in two hours interval so there are 24 data paires (liquidity, t), where t is time between sample time and T0, in hours.

The predicted liquidity in 24hrs is
a 24³ + b 24² + c 24 + d

The good things about the method is that

1. it takes 48 hours data so randome spikes will be mostly ignored;
2. it has up to the third order terms so it accounts for constant, speed, acceleration, and change rate of acceleration. No higher terms are used, to increase robustness of the fit;
3. It is updated at any time so the curve is adapting latest conditions

uncertain aspects:
3rd order polynmial may not act as fast. The exact form of the function can be improved.

To implement a script can be written to do the calculation based on liquidity data from a client by anyone. Nu website can have the data published in realtime.

[Nagalim]

So the issue with a hard rate (25%) is what happens if we have 500,000 nbt of liquidity? Then we get into the situation where we’re hundreds of kNBT in debt before we start putting out the fire. With more liquidity in the system, our threshold needs to be tighter.

Arctan is a sigmoid function. We can use other sigmoid functions that look prettier, but this one was my favorite.

I like the fitting, but in my mind the signers are basically just going to play things by ear anyway. But I’m not wholly against curve fitting, it could work if the sampling rate is reasonable. How do you feel about the 2 nbt/hr^2 acceleration? Does the curve fitting model use something else to determine the size of the Tx (in btc) to sign?

[mhps]

Curve fitting scales. I don’t know magic numbers such as 2 nbt/hr^2 are good, and if they are, they will still be good tomorrow.

I haven’t thought about tx to sign. Can be an amount enough to postpone hitting the 25% threshold for 48 hrs.

Combining the two paragraphs above, if we want to remove the 25% magic number, we can implement a “servo loop” that injects or removes fund according to a continuous feed back of liquidity status. The more off-balance the buy/sell sides are, the stronger fund injection/removal becomes to remove imblance. Just to achieve balance it can be quite insensitive to the form of the feed back function as long as the feedback is negative.

What you have proposed can be a step toward that directiion but it has many magic numbers. The curve fitting approach with a high thresold (e.g. 40%) plus continuous “postponing 48 hrs” (evaluated and applied if needed every day) is adaptive and conceptually easy to understand. (The 40% will be a soft parameter that is there to adjust gas/brake pedal dead gap.)

[Nagalim]

My approach uses that same 40% as one of its magic numbers. The other magic number I use is 20kNBT which is a scaling factor for the arctan. The final magic number (2nbt/hr^2) is totally up for debate for me. However, I really don’t understand how your method gets around the 2nbt/hr^2, which is a statement about market velocity and how much to bring to market how quickly.

So your fitting model is like this (tell me if I’m wrong):
We fit to find out if the trends are such that the threshold will be passed in the next 48 hours. If so, bring to market the amount required to make it so that the market will reach a balanced peg in 48 hours instead.

If there is a single big sell event, how would your method handle that? Wouldn’t it throw off the fitting so much that the numbers all go haywire and we bring way too much money to market?

[dysconnect]

Ultimately this is what we should aim for:

Based on a stochastic model, be able to calculate a confidence interval of NBT liquidity for the next 48 hours.

These are concrete measurements we should take:

1. The covariance between non-NuBot NBT price, NuBot NBT price and BTC price.
2. The covariance between NBT walls and BTC price
3. BTC volatility

For a start one should use a Gaussian distribution to model the ups and downs and do a simulation. Then perhaps we can consider fat-tail distributions for more robustness. Ultimately it boils down to the NBT flow in the system so things like constant parking rates or tiered transactions are quite important.

Rare events like “sudden sell-offs” are ultimately what we want to mitigate using tier 4 so they don’t really have to be modelled in the same way.