Skip to content

Architecture

StrategyBase class relations

All strategy classes are derived from the StrategyBase class, which is derived from the TimeIterator class.

The concrete strategy classes included with Hummingbot, including ArbitrageStrategy, CrossExchangeMarketMakingStrategy, and PureMarketMakingStrategy - are all child classes of StrategyBase.

Each StrategyBase object may be managing multiple ConnectorBase and WalletBase objects.

How It Works

All strategy modules are child classes of TimeIterator, which is called via c_tick() every second.

What this means is, a running strategy module is called every second via its c_tick() method to check on the markets and wallets, and decide whether it should perform any trades or not. One way to think about it is that a strategy module acts like it's watching a movie frame-by-frame via c_tick(), and reacts to what it sees in real time via trading actions.

If you're reading or writing a strategy module, the c_tick() function should be treated as the entry point of a strategy module. If you're reading a strategy module's code, c_tick() should be where you start. If you're writing a new strategy module, c_tick() is also going to where you start writing the important bits of your strategy.

Connectors

Each StrategyBase object may be associated with multiple connectors.

  • cdef c_add_markets(self, list markets)

    Associates a list of ConnectorBase objects to this StrategyBase object.

  • cdef c_remove_markets(self, list markets)

    Disassociates a list of ConnectorBase objects from this StrategyBase object.

  • active_markets property

    List of ConnectorBase objects currently associated with this StrategyBase object.

Market Event Interfaces

The StrategyBase class comes with a set of interface functions for handling market events from associated ConnectorBase objects, which may be overridded by child classes to receive and process market events.

The event interface functions are as follows:

  • cdef c_did_create_buy_order(self, object order_created_event)

    A buy order has been created. Argument is a BuyOrderCreatedEvent object.

  • cdef c_did_create_sell_order(self, object order_created_event)

    A sell order has been created. Argument is a SellOrderCreatedEvent object.

  • cdef c_did_fill_order(self, object order_filled_event)

    An order has been filled in the market. Argument is a OrderFilledEvent object.

  • cdef c_did_fail_order(self, object order_failed_event)

    An order has failed in the market. Argument is a MarketOrderFailureEvent object.

  • cdef c_did_cancel_order(self, object cancelled_event)

    An order has been cancelled. Argument is a OrderCancelledEvent object.

  • cdef c_did_expire_order(self, object expired_event)

    An order has expired. Argument is a OrderExpiredEvent object.

  • cdef c_did_complete_buy_order(self, object order_completed_event)

    A buy order has been completely filled. Argument is a BuyOrderCompletedEvent object.

  • cdef c_did_complete_sell_order(self, object order_completed_event)

    A sell order has been completely filled. Argument is a SellOrderCompletedEvent object.

Creating and Cancelling Orders

StrategyBase includes pre-defined logic for creating and cancelling trading orders - which are the primary ways for a strategy to interact with associated markets.

It is highly encouraged to use these functions to create and remove orders, rather than calling functions like c_buy() and c_sell() in ConnectorBase objects directly - since the functions from StrategyBase provides order tracking functionalities as well.

Place order

cdef str c_buy_with_specific_market(self, 
                                    object market_trading_pair_tuple,
                                    object amount,
                                    object order_type = *,
                                    object price = *,
                                    double expiration_seconds = *
                                    )

cdef str c_sell_with_specific_market(self,
                                     object market_trading_pair_tuple,
                                     object amount,
                                     object order_type = *,
                                     object price = *,
                                     double expiration_seconds = *
                                     )

Creates a buy or a sell order in the market specified by market_trading_pair_tupleand returns the order ID string.

Arguments

  • market_trading_pair_tuple: a MarketTradingPairTuple object specifying the ConnectorBase object and trading pair to create the order for.
  • amount: a Decimal object, specifying the order size in terms of the base asset.
  • order_type: an optional OrderType enum specifying the order type. Default value is OrderType.MARKET.
  • price: an optional Decimal object, specifying the price for a limit order. This parameter is ignored if order_type is not OrderType.LIMIT or OrderType.LIMIT_MAKER.
  • expiration_seconds: an optional number, which specifies how long a limit should automatically expire. This is only used by Ethereum-based decentralized exchanges like Radar Relay where active order cancellation costs gas. By default, passive cancellation via expiration is used on these exchanges.

Cancel order

 c_cancel_order(self, object market_pair, str order_id)

Cancels an active order from a market.

Arguments

  • market_pair: a MarketTradingPairTuple object specifying the ConnectorBase object and the trading pair to cancel order from.
  • order_id: Order ID string returned from a previous call to order creation functions above.

Order Tracking

Each StrategyBase object comes with an internal attribute _sb_order_tracker, which is an OrderTracker object. The OrderTracker object is responsible for tracking all active and in-flight orders created by the StrategyBase object, and also all in-flight order cancels.

StrategyBase and order tracker

When writing or modifying a strategy module, you can use the built-in OrderTracker object to query the active or in-flight orders / cancels you currently have. It's useful for preventing issuing duplicate orders or order cancels.

Below are some of the user functions or properties under OrderTracker that you can use:

  • active_maker_orders property

    Returns a list of still active limit orders, with their market object.

    Return type: List[Tuple[ConnectorBase, LimitOrder]]

  • market_pair_to_active_orders property

    Returns a dictionary mapping from market trading pair tuples to lists of active limit orders.

    Return type: Dict[MarketTradingPairTuple, List[LimitOrder]]

  • active_bids property

    Returns a list of active limit bid orders, with their market object.

    Return type: List[Tuple[ConnectorBase, LimitOrder]]

  • active_asks property

    Returns a list of active limit ask orders, with their market object.

    Return type: List[Tuple[ConnectorBase, LimitOrder]]

  • tracked_taker_orders property

    Returns a list of in-flight or active market orders, with their market object. This is useful for decentralized exchanges where market orders may take a minute to settle due to block delay.

    Return type: List[Tuple[ConnectorBase, MarketOrder]]

  • in_flight_cancels property

    Returns a dictionary of order IDs that are being cancelled.

    Return type: Dict[str, float]

Fee Accounting

OrderCandidate

An order proposal created by a strategy.

A BudgetChecker takes an OrderCandidate and fills in fees to be paid for this particular order (in pre-defined tokens). Then checks if, after accounting for the fees, the account balances allow for a placement of this order, and if not, adjusts the order amount accordingly.

Fees can be payable in the base tokens, quote tokens, or 3rd party tokens.

In general, if an order opens a position, fees will be charged as an additional cost of the order, and if an order closes a position, fees will be deducted from the returns. A few exchanges however don't follow this principle. Strategies however don't have to handle this and should rely on the BudgetChecker obtained from the exchange to calculate fees the correct way.

Once the candidate order has been sized by the BudgetChecker, the strategy can examine the sized order to get more information such as OrderCandidate.collateral_dict to get a dictionary of the costs associated with the order, or OrderCandidate.potential_returns to get an idea of the token and amount of the returns associated with the order.

BudgetChecker

Is intended to be a single universal solution for fee accounting and checking feasibility of OrderCandidates.

Provides utilities for strategies to check the potential impact of OrderCandidates on user account balances. Mainly used to determine if sufficient balances are available to place a set of strategy-proposed orders.

It can work with a single OrderCandidate or a list of OrderCandidates. In case of multiple OrderCandidates the BudgetChecker verfies if the set of orders as a whole is feasible.

The BudgetChecker also locks in collateral required for orders and adjusts collateral available for future OrderCandidates.

  • reset_locked_collateral()
  • adjust_candidates()
  • adjust_candidate_and_lock_available_collateral()
  • adjust_candidate()
  • populate_collateral_entries()

Example

budget_checker = market_info.market.budget_checker
order_candidate = OrderCandidate(
    trading_pair=market_info.trading_pair,
    is_maker=False,
    order_type=OrderType.LIMIT,
    order_side=TradeType.BUY,
    amount=order_amount,
    price=order_price,
)

adjusted_candidate_order = budget_checker.adjust_candidate(order_candidate, all_or_none=True)

if adjusted_candidate_order.amount < order_amount:
    # Order cannot be placed
else:
    # Order can be placed