Prediction markets can suffer a black swan

[pmconrad]

@abitmore noticed that when a price feed for a prediction market is published, the usual black swan conditions are checked. The black swan checks disregard the is_prediction_market flag, and if the conditions are met, global settlement is triggered for the prediction market.

This behaviour does not make sense:

• PM tokens are always backed by the same number of backing tokens. A PM will be settled in the range [0,1], which means there is always sufficient collateral available.
• The actual feed price and MCR seem to be otherwise unused in a PM, so there is no reason to act upon them.

(This has happened on mainnet and testnet, which caused the recent problems.)

• http://docs.bitshares.eu/bitshares/tutorials/pm-close-manual.html should also be corrected, as right now it indicates that a price feed must be published for settling a PM.
• the pm-create-manual document needs to be revised as well

[abitmore]

I thought it's a feature, so not only the issuer but also specified feed producers can global settle the PM.

[pmconrad]

Hm. But the black swan settles at the ratio of the least collateralized short, which is always 1:1 in a PM. So this is a rather limited feature.

[abitmore]

Agreed. Since we don't have a specification document, we can only guess. We may need to redesign this feature.

[abitmore]

Need a BSIP.

[abitmore]

After done more investigation, I agree this is a bug. We can fix it by disabling price feeds for prediction markets, don't need a BSIP.

The feature that global settle a prediction market via price feeds is debatable and deserves a BSIP.

By the way, need a better business model for the asset creator. Because the token can worth zero in the end, percentage market fee is not enough. This can be discussed in a BSIP as well.

[abitmore]

Sorry for accidentally sent out an unfinished comment.

I decided to postpone this as it's still like a uncompleted feature.

• When the outcome of a PM is 1, the issuer can global settle with 1, or the price feed producers can feed 1 (which will trigger a black swan event then global settle at 1), the result is same.
• When the outcome of a PM is 0, the issuer can global settle with 0, but
  • the price feed producers are unable to feed 0 (which is a bit complicated to fix),
  • even if they feed a very small price e.g. 0.0001, the market won't be global settled since a small price won't trigger a black swan event.

The final fix would be:

• allow and only allow feeding 0 or 1
• when there are enough feeds, global settle at median feed price.

[abitmore]

Updated the documents according to current situation, may need to update again after the issue is fixed.

• https://github.com/bitshares/docs.bitshares.org/blob/master/source/bitshares/tutorials/pm-create-manual.rst
• https://github.com/bitshares/docs.bitshares.org/blob/master/source/bitshares/tutorials/pm-close-manual.rst

[jmjatlanta]

The final fix would be:

* allow and only allow feeding 0 or 1

* when there are enough feeds, global settle at median feed price.

Q1: To feed 0, which you say is not easy, are you speaking of permitting 0 on one side of the equation ( removing validation in price::validate() ) and then cleaning up code to assure no divide by zero problems? Or perhaps you have in mind a way to feed 0 another way?

Q2: For your second bullet point, are you suggesting that having the required number of feeds present in a prediction market should automatically trigger global settlement?

[abitmore]

@jmjatlanta I think it's better to discuss in bitshares/bsips#42 atm.

A2Q1: IIRC, the former.
A2Q2: something like that, perhaps using a new operation is better.

[pmconrad]

The final fix would be:

• allow and only allow feeding 0 or 1
• when there are enough feeds, global settle at median feed price.

No. bitshares/bsips#42 discusses a protocol change. This ticket here is about a bugfix.
The pricefeed should not load to global settlement on PMs, that's all.

[abitmore]

Fixed by #2019.