OrderBook
An OrderBook is a representation of the currently known Ask and Bid levels. Essentially, these are the quantities of a given Asset that people want to sell (Ask) or buy (Bid) at certain prices.
In Trading.jl they are implemented in the AssetLedger as 2 TreeComponents. These are standard Components that are also internally backed by a Red-Black Tree (facilitating easy search type operations, see ceil, floor, maximum and minimum) of LinkedLists of individual Asks and Bids.
By default, when starting a Trader it will listen to updates on new Bids Asks, Trades and latest_quote updates. The latter signifies the L1 OrderBook updates. The former 3 are stored in a manner that facilitates L3 OrderBook interactions (each ask/bid is stored individually in the linked lists). sum(x -> x.quantity, limit) can be used to extract L2 type market data from the OrderBook.
Behavior
While Ask, Bid, Trade and latest_quote data comes in, the Trades are matched with the Ask and Bid levels and clear these. The matching works as follows:
- we assume that we have incomplete
Tradedata, and no directBidorAskcanceling data - a
Tradeis assumed to always happen on the bestBidorAskprice level - if there exist
BidorAsklevels between theTradeand the OrderBook center, they will be cleared assuming they are stale - if a level exists with the exact
Tradeprice, it will be cleared until the cleared quantity matches theTradequantity
This means that when a sell trade comes in at price 5.0 then all Bids with a higher price than 5.0 will be assumed outdated and thus removed.
Example
See OrderBook Imbalance for an example usecase in a simple Strategy.
References
Trading.Ask — TypeRepresents an ask order with a price and quantity.
Trading.Bid — TypeRepresents a bid order with a price and quantity.
Trading.Trade — TypeRepresents an executed trade with a price, quantity and side.
Trading.latest_quote — FunctionReturns the latest NBBO quote.
Trading.TreeComponent — TypeA Component backed by a red-black tree.
Indexing into a component with an Entity will return the data linked to that entity, indexing with a regular Int will return directly the data that is stored in the data vector at that index, i.e. generally not the storage linked to the Entity with that Int as id.
To register a Type to be stored in a TreeComponent see @tree_component. Every Type needs to define < and == operators.
Trading.@tree_component — MacroDefines that a struct should be used with a TreeComponent.
@tree_component struct Example
d
endBase.ceil — MethodReturns the first level (i.e. a LinkedList with EntityPtr) with a higher value than v.
Base.floor — MethodReturns the first level (i.e. a LinkedList with EntityPtr) with a lower value than v.
Base.maximum — MethodReturns the first [EntityPtr] in the LinkedList with the maximum value in the TreeComponent. See maximum_node if the full list is desired.
Base.minimum — MethodReturns the first [EntityPtr] in the LinkedList with the minimum value in the TreeComponent. See minimum_node if the full list is desired.
Trading.minimum_node — Functionminimum_node(tree::Tree, node::TreeNode)
Returns the TreeNode with minimum value in subtree of node.
Returns the LinkedList of EntityPtrs containing the minimum value in the `TreeComponent.
Trading.maximum_node — Functionmaximum_node(tree::Tree, node::TreeNode)
Returns the TreeNode with maximum value in subtree of node.
Returns the LinkedList of EntityPtrs containing the maximum value in the `TreeComponent.
Trading.levels — Functionlevels(c; rev=false)Iterate through the levels (each a LinkedList of EntityPtrs) of the TreeComponent c, either inorder if rev == false or in reverse order if rev == true.
Trading.LinkedList — TypeA classic linked list implementation, mostly used to represent value levels in a TreeComponent
Trading.EntityPtr — TypeHolds the direct reference of an Entity into an array of associated data (usually in an AbstractComponent).