Closed tamlyn10 closed 1 year ago
I think option 3 is a no go because of the potential surprising impact of boosting it on the LPs risk/inventory.
I also agree with you that this isn't a priority short term and think we should avoid making a decision on this until we have data from testnet including liquidity monitoring.
The option not listed is to enforce a minimum target stake scaling factor that makes the scenario impossible. I think (though memory is rusty so may be wrong) that if target stake scaling factor is at least 1 then, as everything gets multiplied by the risk factor, you'd always be able to cover the max OI in the window (which is at least equal to the current OI) or you'd be in an auction anyway, so I am not sure that this scenario requires special logic. But let's see.
Yeah, I deliberately didn't include the option you mentioned because:
target_stake_scaling_factor
to >1, because the taraget stake may be low as a result of the mark price dropping or the rf being low due to other parameters.
target_stake = mark_price x max_oi x target_stake_scaling_factor x rf
Regarding my option 3 I agree it's not great.. but, market makers of last resort are a mechanism we could play with more and lots of ways to work that in. Definitely not worth complicating stuff now tho, of course.
We should probably change the mechanism to "closeout as much as the order book allows and confiscate the margin".
This will tip the market into liquidity auction (because it will eat any limit orders that LP commitment is pegged to). Once the auction ends we go back to reevaluate the margin position of that party: most likely the margin balance will be zero (of course MTM in another market can provide balance) so we repeat "closeout as much as the order book allows and confiscate the margin".
If the MTM inflow into general from another market was big enough then perhaps the party gets to keep the rest of their position - fine, I can live with that.
In our position resolution spec, we say:
We had to put this in because we didn't have liquidity monitoring or auctions but I think this is problematic for a few reasons:
tau
level set in the risk model, it may mean theeffective-tau
is >risk-model-tau
which means the statistical "tuning" of the system is off (this mostly affects the insurance pool and the probability of socialised loss).Proposed alternative approaches:
Other ideas? cc @davidsiska-vega @barnabee @witgaw
Note, I don't think this is a priority in the short term.