BSIP: 0033
Title: Maker Orders With Better Prices Take Precedence
Author: Abit More <https://github.com/abitmore>
Status: Installed
Type: Protocol
Created: 2018-02-17
Discussion: https://github.com/bitshares/bitshares-core/issues/625,
https://github.com/bitshares/bitshares-core/issues/453
Replaces: -
Worker: 1.14.94
Currently, in BitShares, under certian circumstances, a taker order may be matched with a maker order which is not "on the top" of the order book. This behavior is unexpected and irritating for users.
This BSIP proposes a principle: always match taker orders with maker orders with the best prices (aka on the top of the order book) first.
As expected, when matching taker limit orders with maker limit orders, the maker limit orders with better prices will always be matched first.
For example, if trader A has a limit order selling 100 BTS at 0.1 USD per BTS, trader B has a limit order selling 100 BTS at 0.09 USD per BTS, which means B's limit order has a better price for the opposite side to buy. Now if trader C placed an order that buys 10 BTS at 0.105 USD per BTS, B's limit order will take precedence, A's limit order won't be matched.
However, when there are (maker) margin call orders in the market which have met the requirements that to be matched (able to be margin called), they always take precedence over the (maker) limit orders on the same side, no matter whether the limit orders provided better price.
For example, if trader A's margin call order is selling 100 BTS at no less than 0.1 USD per BTS, trader B has a limit order selling 100 BTS at 0.09 USD per BTS, which means B's limit order has a better price for the opposite side to buy. Now if trader C placed an order that buys 10 BTS at 0.105 USD per BTS, A's margin call order will take precedence, B's limit order won't be matched. That means C is forced to "buy high" when have chance to "buy low", which is unexpected.
Users have been confused by this behavior, as discussed in bitshares-core issue #625 and other threads.
Another scenario is described in Bitshares-core issue #453 that sometimes a taker margin call order may be matched with a maker limit order which is not on the top of the order book. This can be seen as a bug.
Always matching taker order (e.g. a buy) with a maker order which offered best price (aka lowest ask), is a simpler rule which most users would understand easily.
There is a parameter in price feed named MSSR, which stands for "maximum short
squeeze ratio". Maker price of margin call orders is MSSP, which stands for
"maximum short squeeze price", is calculated as feed_price / MSSR
.
Note: feed_price
here is in terms of debt/collateral, aka "how much debt per
collateral".
That said, the price that margin call orders are offering is MSSP. The prices those limit orders are offering are the limit prices.
When placing a limit (e.g. buy) order with a price beyond the lowest sell, the order is expected to "walk the book", matching each order on the opposite side in turn, at that order's price, until the new limit order is completely filled, or there is no more sell order matching its price.
To meet the expectation,
- firstly, we need to match the limit buy order with the limit sell orders whose prices are lower than MSSP and prices can match the new order;
- then, if the new limit buy order hasn't been completely filled, match it with the margin calls if MSSP can match the new order's price;
- then, if the new limit buy order still hasn't been completely filled, match it with the rest sell orders until it's completely filled or no more sell order matching its price
New limit order is being processed in apply_order(...)
function of database
class.
Currently, in the function, firstly call orders will be checked and matched. After that, limit orders on the opposite side will be checked and matched.
Need to change the logic to:
- firstly, sort the limit orders on the opposite by
price
, best price first, end at MSSP; check them one by one, callsmatch(...)
function until the return value is not2
which means the new order is completely filled; - if reach the end (MSSP), which means the new order is still unfilled,
call
check_call_orders(..)
function or an equivalent; - check if
new_order_object
still exist, if yes, redo step one but set the maximum possible price of the market as end price.
For bitshares-core
issue #453,
in check_call_orders(...)
function of database
class,
iterator limit_itr
will move forward when variable filled_limit
is true
.
filled_limit
will be set to true
when a limit order get completely filled.
However, since filled_limit
is declared out of the while
block,
it doesn't get reset to false
after limit_itr
moved forward. That means
after the first limit order get completedly filled, filled_limit
will always
be true
, so limit_itr
will always move forward no matter whether current
limit order got completedly filled, so a taker call order may match
with a limit order that is not on the top of the order book.
To fix this, need to change the code to make sure limit_itr
always references
the limit order on the top of the order book when dereferencing.
[to be added if any]
[to be added if any]
This document is placed in the public domain.