Author Topic: Investigating the reasons for BitUSD GS  (Read 515 times)

0 Members and 1 Guest are viewing this topic.

Offline biophil

  • Hero Member
  • *****
  • Posts: 841
  • Incentives run the world
    • View Profile
    • Sign up for a Bitshares account!
  • BitShares: zebulon
Investigating the reasons for BitUSD GS
« on: December 06, 2018, 03:12:04 pm »
Greetings, Bitshares community!

A very few of you may remember me from the early days of Bitshares; back then, I was a lowly graduate student interested in game theory and cryptocurrencies.

Now, I've worked my way up to professor of Computer Science at the University of Colorado: http://cs.uccs.edu/~pbrown2/. I've been doing a bit of research on stablecoins recently, and I want to gather some perspectives on what went wrong with BitUSD. I think there are some very interesting underlying problems with stablecoins in general (see NuBits for some early examples), and collaboratively-managed stablecoins in particular (see BitUSD), and I'd like to understand these problems very clearly.

So I'd love to get some community input on the reasons that BitUSD's peg mechanism broke and it went into GS. Here are some starter questions, but of course give me your own perspectives even if they don't fit my questions at all:

1. Was BitUSD mismanaged? If so, by whom and why?
2. Was it the fault of BSIP42?
3. How will BitCNY be saved from the same fate?
4. Was the GS inevitable?

And just to be clear about my intentions here, depending on my findings, I plan to write a (possibly series of) academic research papers on the topic of how to properly incentivize stablecoins. I think the incentive issues are very deep and likely very difficult to surmount. Here's an older summary of my concerns: https://bitsharestalk.org/index.php?topic=9348.msg315532#msg315532

Offline Fox

Re: Investigating the reasons for BitUSD GS
« Reply #1 on: December 07, 2018, 02:47:40 am »
I am keenly interested in these topics and academic research thereof. I've sent a note to your University account to continue our dialog.

Cheers!
Fox
Witness: fox

Offline xeroc

  • Board Moderator
  • Hero Member
  • *****
  • Posts: 12722
  • ChainSquad GmbH
    • View Profile
    • ChainSquad GmbH
  • BitShares: xeroc
  • GitHub: xeroc
Re: Investigating the reasons for BitUSD GS
« Reply #2 on: December 07, 2018, 08:57:49 am »
1. Was BitUSD mismanaged? If so, by whom and why?
2. Was it the fault of BSIP42?
3. How will BitCNY be saved from the same fate?
4. Was the GS inevitable?
To break it down:
BSIP42 allowed witnesses to experiment more with price feeding to improve the peg.
Different algorithms have been tried which came with more risks and better peg for certain market conditions.
Market conditions changed drastically. Unfortunately, at that time, the algorithm for price feeding was such
that the peg was holding strong, at a (significantly) increased risk of global settlements.

What the algorithm basically did was move the (margin call) price feed away from market so that margin calls
(which buy at 10% above market) would only trade at current market rates. Hence the price feed was off by
10% in general (more of less). Additionally to that, a stateful feedback loop was integrated to track the *premium*
of bitUSD and move the call price even further.

The consequence was that a huge margin call aggregated during the bear market and those margin calls were
unable to buy back their bitUSD in the internal markets (not even with a 10% penalty) because the call price (feed)
was off too much.

Could a GS have been prevented without BSIP42, I am not sure. It could be because the margin calls could have
been filled more easily. Would there have been traders doing so? No one knows.

My personal conclusion (as a proxy) is to not accept call price (feed) manipulation and instead use other means
to give (and take) incentive to borrow (and short sell) new bitassets (depending on supply and demand).
Those means have been discussed back then too, but an issue in the backend that is currently been worked
on effectively prevents us from using it. Additionally to that, the bear market has continued and moving away from
BSIP42 (in particular for bitCNY) would cause an immediate global settlement, which at least is undesirable for the
vast majority of the bitCNY traders and BTS holders from China.

The ideal "solution" for me would be to sit out the bear market and as soon as collateral ratios in bitCNY pick up again
gradually reduce the impact of BSIP42 and instead replace it with (potentially a mix of) dynamic MCR and dynamic
MSSR.

bitUSD can be revived and when that happens I do not intend to support any witness that feeds a price feed that
goes more than ~2% away from fair market pricing for bitUSD.
Give BitShares a try! Use the http://testnet.bitshares.eu provided by http://bitshares.eu powered by ChainSquad GmbH

Offline bitcrab

  • Committee member
  • Hero Member
  • *
  • Posts: 1067
    • View Profile
  • BitShares: bitcrab
  • GitHub: bitcrab
Re: Investigating the reasons for BitUSD GS
« Reply #3 on: December 07, 2018, 10:10:57 am »
now bitCNY has a global settlement price of 0.2568.

the current BTS price is 0.247CNY.

this means if now witnesses feed real market price bitCNY is already in black swan status.

now most of witness add "black swan protection" logic to ensure they feed price > global settlement price*1.11

this help to avoid black swan, and now bitCNY still has a premium of about 4%.

anyway, I think this is much better than just let global settlement happen.

actually we have suggest to add same protection logic in bitUSD price feeding, but seems only little witnesses has done that.




Offline biophil

  • Hero Member
  • *****
  • Posts: 841
  • Incentives run the world
    • View Profile
    • Sign up for a Bitshares account!
  • BitShares: zebulon
Re: Investigating the reasons for BitUSD GS
« Reply #4 on: December 07, 2018, 05:27:36 pm »

anyway, I think this is much better than just let global settlement happen.


Would you say this is an accurate way to rephrase your statement: "better to lose the peg via a nonconsensus hack than to lose the peg via a GS."

I'm pretty sure that I agree with that sentiment.

That makes it sound like it's worth looking in to a major protocol update which would soften the GS nuclear option in some way.

Offline biophil

  • Hero Member
  • *****
  • Posts: 841
  • Incentives run the world
    • View Profile
    • Sign up for a Bitshares account!
  • BitShares: zebulon
Re: Investigating the reasons for BitUSD GS
« Reply #5 on: December 07, 2018, 05:41:10 pm »
1. Was BitUSD mismanaged? If so, by whom and why?
2. Was it the fault of BSIP42?
3. How will BitCNY be saved from the same fate?
4. Was the GS inevitable?
To break it down:
BSIP42 allowed witnesses to experiment more with price feeding to improve the peg.
Different algorithms have been tried which came with more risks and better peg for certain market conditions.
Market conditions changed drastically. Unfortunately, at that time, the algorithm for price feeding was such
that the peg was holding strong, at a (significantly) increased risk of global settlements.

What the algorithm basically did was move the (margin call) price feed away from market so that margin calls
(which buy at 10% above market) would only trade at current market rates. Hence the price feed was off by
10% in general (more of less). Additionally to that, a stateful feedback loop was integrated to track the *premium*
of bitUSD and move the call price even further.

The consequence was that a huge margin call aggregated during the bear market and those margin calls were
unable to buy back their bitUSD in the internal markets (not even with a 10% penalty) because the call price (feed)
was off too much.

Could a GS have been prevented without BSIP42, I am not sure. It could be because the margin calls could have
been filled more easily. Would there have been traders doing so? No one knows.

My personal conclusion (as a proxy) is to not accept call price (feed) manipulation and instead use other means
to give (and take) incentive to borrow (and short sell) new bitassets (depending on supply and demand).
Those means have been discussed back then too, but an issue in the backend that is currently been worked
on effectively prevents us from using it. Additionally to that, the bear market has continued and moving away from
BSIP42 (in particular for bitCNY) would cause an immediate global settlement, which at least is undesirable for the
vast majority of the bitCNY traders and BTS holders from China.

The ideal "solution" for me would be to sit out the bear market and as soon as collateral ratios in bitCNY pick up again
gradually reduce the impact of BSIP42 and instead replace it with (potentially a mix of) dynamic MCR and dynamic
MSSR.

bitUSD can be revived and when that happens I do not intend to support any witness that feeds a price feed that
goes more than ~2% away from fair market pricing for bitUSD.

Thanks for the writeup. A big problem stablecoins face is that risk is very hard to eliminate, and a change that reduces one risk may very easily increase some other risk. It sounds like what you're saying is that BSIP42 reduced BitAsset short-term volatility, and inadvertently (or maybe deliberately?) increased the risk of a GS. Whether the total amount of risk in the system increased or decreased isn't clear, but since GS is such a drastic operation (essentially everybody loses in a GS), anything which increases the risk of a GS is probably a bad idea.

Offline WDCBTS

  • Jr. Member
  • **
  • Posts: 22
    • View Profile
  • BitShares: blockX
Re: Investigating the reasons for BitUSD GS
« Reply #6 on: December 08, 2018, 04:11:24 am »
Greetings, Bitshares community!

A very few of you may remember me from the early days of Bitshares; back then, I was a lowly graduate student interested in game theory and cryptocurrencies.

Now, I've worked my way up to professor of Computer Science at the University of Colorado: http://cs.uccs.edu/~pbrown2/. I've been doing a bit of research on stablecoins recently, and I want to gather some perspectives on what went wrong with BitUSD. I think there are some very interesting underlying problems with stablecoins in general (see NuBits for some early examples), and collaboratively-managed stablecoins in particular (see BitUSD), and I'd like to understand these problems very clearly.

So I'd love to get some community input on the reasons that BitUSD's peg mechanism broke and it went into GS. Here are some starter questions, but of course give me your own perspectives even if they don't fit my questions at all:

1. Was BitUSD mismanaged? If so, by whom and why?
2. Was it the fault of BSIP42?
3. How will BitCNY be saved from the same fate?
4. Was the GS inevitable?

And just to be clear about my intentions here, depending on my findings, I plan to write a (possibly series of) academic research papers on the topic of how to properly incentivize stablecoins. I think the incentive issues are very deep and likely very difficult to surmount. Here's an older summary of my concerns: https://bitsharestalk.org/index.php?topic=9348.msg315532#msg315532


People who borrow money in a bull market don't want to pay back money in a bear market. Price control makes the settlement price higher than the market price. I think BTS needs laws to restrain these borrowers.


Offline WDCBTS

  • Jr. Member
  • **
  • Posts: 22
    • View Profile
  • BitShares: blockX
Re: Investigating the reasons for BitUSD GS
« Reply #7 on: December 08, 2018, 04:19:17 am »
now bitCNY has a global settlement price of 0.2568.

the current BTS price is 0.247CNY.

this means if now witnesses feed real market price bitCNY is already in black swan status.

now most of witness add "black swan protection" logic to ensure they feed price > global settlement price*1.11

this help to avoid black swan, and now bitCNY still has a premium of about 4%.

anyway, I think this is much better than just let global settlement happen.

actually we have suggest to add same protection logic in bitUSD price feeding, but seems only little witnesses has done that.

Feed price control also needs to restrain broken promises borrower at the same time

Offline Victor118

Re: Investigating the reasons for BitUSD GS
« Reply #8 on: December 08, 2018, 09:53:22 am »
Sorry for my english, but what about the implementation of Bancor protocol ?
It seems very interesting to improve liquidity without borrowers

Offline bitcrab

  • Committee member
  • Hero Member
  • *
  • Posts: 1067
    • View Profile
  • BitShares: bitcrab
  • GitHub: bitcrab
Re: Investigating the reasons for BitUSD GS
« Reply #9 on: December 08, 2018, 04:02:44 pm »

anyway, I think this is much better than just let global settlement happen.


Would you say this is an accurate way to rephrase your statement: "better to lose the peg via a nonconsensus hack than to lose the peg via a GS."

I'm pretty sure that I agree with that sentiment.

That makes it sound like it's worth looking in to a major protocol update which would soften the GS nuclear option in some way.

I don't think "nonconsensus hack" is a suitable name for that.

there is already a lot of discussion on an update handle method on black swan, most of key opinion leads agree to adopt some new methods to handle that without stopping borrowing feature, such as what are mentioned here. https://bitsharestalk.org/index.php?topic=27522.0

but as you know, to do such update in a public blockchain need long time to go through the discussion->BSIP draft->voting->develop and test->hard fork process. when the real risk is nearby, I think it's good for the witnesses to feed the black swan protection price, as it is based on public discussion, not private manipulation.

and you can see bitCNY and bitUSD are in different status now, although neither of them peg well, but bitCNY has a 2-6% premium, however bitUSD has a 30% discount.
« Last Edit: December 09, 2018, 01:05:04 pm by bitcrab »

Offline JonnyB

  • Hero Member
  • *****
  • Posts: 636
    • View Profile
    • twitter.com/jonnybitcoin
Re: Investigating the reasons for BitUSD GS
« Reply #10 on: December 08, 2018, 11:26:05 pm »
Global settlement happens when collateral is not maintained by bitasset shorters.

Bitshares has a built mechanism to punish them to maintain their collateral and also incentivise would be margin call buyers.

The bitasset shorter is punished by selling his collateral at a 10% discount.
The margin call purchaser is incentivised to buy BTS at a 10% discount from feed price.

This system is called MSSR and is currently set at 10%

All we need to do is increase it to 20% and bit asset shorters are more likely to maintain their collateral.


I run the @bitshares twitter handle
twitter.com/bitshares

Offline Cryptick1

Re: Investigating the reasons for BitUSD GS
« Reply #11 on: December 10, 2018, 03:17:08 am »
In my opinion, the the Global Settlement on BitUSD was unavoidable. It was mainly (more than 50%) the direct result of the major drop in BitCoin. Global Settlement in BitUSD was collateral damage. (Note: Bitcoin fell from near $20,000 in January to $3,500, this is a fall of more than 85%. There are not many systems that can handle such a fall.)

For perspective, most Wall Street, traditional banks, and other such financial infrastructure would also collapse if put under those kinds of stresses. In part, it is sort of amazing the system was as dynamic as it was and lasted as long as it did, under those circumstances.

A lot of the other comments here are from some very knowledgeable people.


Opinion…take with extra salt...
I would also like to add that there are certain, “games that were played” with magins. Many of these were about maxing out margins. I understand why some of the actions were taken, but -particularly from an academic point- I do not think there were necessarily good or beneficial to the system as a whole. To start off with many big players maxed out magin to the absolute max -particularly on the way up. So people would create USD, then buy BTS with it, then use that BTS to create more bitUSD and repeat the cycle.  (This is sort of like borrowing money on your credit cards to buy a house and then mortgaging the house.)  While we have a official 1.75 asset requirement, borrowed assets allowed one to effectively get greater leverage than the system was originally intended for because borrowed money was used as collateral. When there was a turn in the market, it was not long before these highly leveraged individuals were suddenly being called. They did not have much protection.

I think there were also games played were money was moved between accounts. (I am not trying to blame people here, there is no problem with managing money in different accounts.) That said, it can be smart, and beneficial, but can also hide certain actions.

I think many that played the game, may be less likely to fully explain it public. They do no want their strategies understood. I think there are some aspects to what happens they would prefer to keep secret.

 That said, as markets enter a long downtrend, there was an advantage to exit certain positions sooner rather than later. A margined position, can be used like a stop loss order, where a margined position will absorb market liquidity. If you had an account, that was getting margin called you could absorb all available liquidity. Taking this liquidity to exit positions while prices was high, was more advantageous than holding onto said positions until they fell. By exiting quickly, one could save profits to buy back in later at a lower price. Many people did not like this constantly margin call walls this activity put up.

With everyone margining on the way up, prices could rise very quickly in a feedback loop. With potentially less than completely liquid markets, and markets rising, and people that constantly margined accounts to the max could push huge upswings in the BTS price very quickly. This though could result in the last person to buy, -taking on too much margin- and losing his position. Whales, would then eat up the small fish. It was a cycle that seemed to happen again and again. It is dangerous if one did not know what they were doing.

Being an open blockchain, it could then add a completely new level of game, where people can see each other’s positions and purposefully squeeze them. All evidence of past transactions is recorded in the blockchain. I think it would provide on interesting research project for some student at some time.

Playing with margin can be very dangerous. Some things are going to happen. There are lots of levels to the activity that takes place. 

Offline Victor118

Re: Investigating the reasons for BitUSD GS
« Reply #12 on: December 10, 2018, 03:21:47 pm »
BitUSD needs borrowers to exist, but borrowers do not care about demand, it means that the supply is not correlated with demand, a demand that increases does not incentivize borrowing and therefore increase supply