MessageEventOrderFlowImbalance

Summary

MessageEventOrderFlowImbalance is RTTA's streaming message-tape order-flow imbalance: a rolling sum of signed contributions from discrete LOB/trade events (add, cancel, trade) rather than Cont-style best-quote snapshot deltas.

Update API

result = rtta.MessageEventOrderFlowImbalance(window=50, fillna=True).update(
    event_type, side, qty
)
# result.ofi, result.event, result.signed_size

event_type: 1 = add, 2 = cancel, 3 = trade (other codes fall back to signed size only)
side: +1 = bid / buy aggressor, -1 = ask / sell aggressor
qty: non-negative event size (absolute value is used if negative)

advance(...) uses the same three inputs without returning a Python object. Multi-output batch(event_type, side, qty) returns arrays for ofi, event, and signed_size.

Theory Of Operation

Snapshot OFI (OrderFlowImbalance) reconstructs pressure from changes in best bid/ask price and size. Message-tape OFI instead consumes the exchange event stream: liquidity adds, cancels, and trades. RTTA maps each event to a signed contribution and accumulates those contributions over a rolling window of events. Positive rolling sum means net bid/buy pressure in the recent message history; negative means net ask/sell pressure.

Recurrence

Let \(e_t\in\{1,2,3\}\) be the event type, \(s_t\in\{+1,-1\}\) the side, and \(q_t\ge 0\) the quantity. Define the contribution

\[\delta_t = \begin{cases} +s_t q_t & e_t = 1 \quad\text{(add)} \\ -s_t q_t & e_t = 2 \quad\text{(cancel)} \\ +s_t q_t & e_t = 3 \quad\text{(trade)} \\ +s_t q_t & \text{otherwise} \end{cases}\]

Over a rolling window \(W_t\) of the last \(n\) contributions (window):

\[\operatorname{OFI}_t = \sum_{\tau\in W_t} \delta_\tau\]

Each update also emits the latest contribution as event and the signed size \(s_t q_t\) as signed_size.

Implementation Notes

The recurrence is implemented in src/rtta/indicator.cpp in class MessageEventOrderFlowImbalance using a RollingSumWindow. With fillna=False, ofi is NaN until the window is full.

This is a research-facing event API; it is not interchangeable with top-of-book OrderFlowImbalance without converting the feed.

Reference

Cont, Kukanov, and Stoikov, "The Price Impact of Order Book Events," arXiv:1011.6402 ( Cont-style OFI motivation; RTTA's snapshot form is OrderFlowImbalance, while this class applies the same imbalance idea to explicit message events. )
Cont, Cucuringu, et al. multi-level / LOB flow literature overview