Launch Hummingbot Miner


Info on developing TWAP

The TWAP strategy is a common algorithmic execution strategy used for splitting up large orders over time. Specifically, the TWAP strategy helps traders minimize slippage when buying or selling large orders. These features make the strategy more useful to traders and will help when creating future, more complex strategy:

  • Incrementing / maintaining states over clock ticks
  • Quantizing (rounding down to nearest tradable value) order size
  • Dividing an order into segments
  • Incorporating time delays between segmented orders


The TWAP strategy divides a large user order into chunks according to the following user configurations:

  • Total order size
  • number of individual orders
  • time delay between orders

Figure 1: Processing orders

The orders are then split into tradable (quantized) amounts and executed sequentially with the indicated time delay between orders. There is no time delay before the first order. Because only one order is placed in a clock tick, a state machine must emit multiple orders over different clock ticks. To see the executed orders, type history into the command prompt.


Here are the additional user-configurable parameters for the TWAP strategy (fields are added to config_map file):

  • time_delay : Change the question to ask for the number of seconds to delay each individual order. (e.g., How many seconds do you want to wait between each order?)
  • num_individual_orders : a new field added to the config map. It should ask for the number of individual orders that an order should be split up into. (e.g., Into how many individual orders do you want to split this order?)


The TWAP strategy logic is trying to split a large order into smaller ones over time. It does that by maintaining important information about the state when processing orders by adding state variables.

Custom state variables can be added to the strategy by setting variables in the __init__ function.

  • self._quantity_remaining : Indicates the number of orders left to be placed as individual orders. This state variable is updated after each order is placed and persisted until the order is processed.
  • self._first_order : Indicates whether the current individual order is the first order.

Figure 2: Placing orders

TWAP processes orders when there is a remaining order quantity & the specified time_delay has passed. Specifically, some of the key elements in utilizing the remaining order quantity and time_delay are detailed below:

  • If self._quantity_remaining is greater than 0, place an order
  • If self._first_order is true, we want to place an order as soon as self._current_timestamp > self._previous_timestamp we don't have a time delay before the first order
  • If it isn't the first order, check that self._current_timestamp > self._previous_timestamp + self._time_delay
  • Once an order is placed, update self._quantity_remaining by subtracting the amount of the order just placed curr_order_amount : Either (total order amount)/(number of orders) or self._quantity_remaining depending on which is smaller
Edit on GitHub