I agree with
@tonyk that we need to limit our discussion to the USD:BTS market, i.e. prices indicating how many BTS you need to pay for 1 USD, e.g. 250 BTS/USD.
This way shorts are those who actually have a short position, i.e. they have sold their USD and to close their collateral position they will need to buy it back in the future.
So BM's original drawing in OP, which is based on the opposite view, should not be part of the discussion, as it just adds to the confusion.
And I agree with
@tbone that we need to differentiate between these two prices (and situations):
(1) Margin Call Trigger Price (MCTP), i.e. the price at which a margin call is triggered, bearing in mind that the trigger event occurs when MCTP gets above Your Personal Call Price (YPCP).
(2) Actual Buy-Back Price (ABBP), i.e. the price at which the actual buy-back is executed by the system once the margin call is triggered.
We absolutely need to distinguish between these two situations, as otherwise we end up having a paradox illustrated by the order-book shown below, an event which actually took place in October and put the shorts in an unacceptable price trap.

(Now I wonder if the situation above was actually considered by BM when he was designing the SQP concept - is it a bug or a feature?)
Regarding (1), i.e. MCTP:I agree with
@Xeldal that basing the MCTP on the lowest ask price (from the internal market) is a bad idea, as in a thin market the lowest ask cannot be really trusted.
BM's way to alleviate this weakness, achieved by including an additional restriction (i.e. the lowest ask needs to be below SQP to become a valid margin-call trigger) does not really help because:
- if we define SQP to be far from the feed (e.g. 50% above the feed) it does not force the longs to offer a decent exit price for the shorts when triggering a margin call,
- if we define SQP to be close to the feed (e.g. only 10% above the feed, as we have now) we could just as well rely on the feed itself, as having SQP at this low level does not make margin calls execution much more efficient.
This is how I propose to solve this problem: let's define MCTP as
the lowest of these two:
(a) the external market's feed price (or a little above the feed, like 5% above the feed).
(b) the internal market's lowest ask price
This solution will protect us from being dependent entirely on the external feed (in case there are some technical issues with it). For a margin call to be triggered,
both the lowest ask
and the feed need to be above YPCP.
And once a margin call is triggered (i.e. MCTP gets above YPCP), ABBP needs to be established and this has nothing to do with MCTP - it can be much higher than MCTP, the only restriction is that it needs to be equal or above YPCP.
Regarding (2), i.e. ABBP:I think I understand BM's intention to avoid those kinds of market limbos that we had in 0.x. A margin call should have a real chance to be executed immediately, even in a thin market. And this is where something like SQP (defined even as high as 50% above the feed) might make sense. It can render ABBP very bad for a short position holder but it will, at the same time, offer a predictable limit for the damage incurred. And that's the reality of shorting in a an ill-liquid market: when you allow yourself to be margin-called, the consequences can be very painful. But still for the whole market it's better to execute a margin call successfully (even at a very bad price for the shorter) than have this kind of limbo that we had in 0.x.
EDIT: To sum up, my conclusion is this: currently we have a serious flaw in the margin-call rules but this is not because SQP is such a bad concept.
The problem is that
SQP is currently being misused, because it acts as
both a margin-call trigger level
and a margin-call execution level.
These two situations should be treated separately (as
@tbone pointed it out) and SQP should only be used for the margin-call execution part,
not the trigger part.
For the trigger part we should rely on the feed (or something very close to the feed), plus the lowest ask level as an extra safety feature.