Case Study — Algo Platform Integration

Integrating Tradency's RoboX with TOCOM

A FIX adapter bridging NTT Data's RoboX algo-trading engine to Japanese futures markets — solving non-FIFO position referencing, order routing, risk checks and back-office position management for a platform that does no position-keeping of its own. Delivered 2017–2018.

Executive Summary

NTT Data engaged Touch-Fire to build an adapter interfacing the RoboX algo-trading engine (from Tradency) to the exchange. RoboX feeds orders into the Touch-Fire system, which routes them to the venue and returns every order and fill message back to RoboX — with TOCOM as the initial market, and the Tokyo Financial Exchange to potentially follow.

The defining challenge: RoboX runs many independent strategies that can hold long and short positions in the same instrument simultaneously, and it performs no position management of its own. A simple FIFO model was therefore impossible. Touch-Fire designed a per-position referencing scheme and took on full back-office position-keeping, while feeding RoboX exactly the order, fill, position and collateral data it needed.

Technical Architecture

RoboX sends orders into a dedicated Touch-Fire FIX RoboX adapter. From there they flow through the hub to the venue adapters exactly like any other client order, before being routed to the exchange.

  • Fill-or-Kill orders: all strategy orders are sent FOK to minimise resting working orders and partial fills.
  • Unified symbols → deliveries: RoboX trades "unified" contracts and does not manage individual futures deliveries or rollover, so the adapter maps each incoming symbol to the appropriate far-month delivery.
  • Risk on the way through: the hub performs risk checks and updates cash and net-liquidity for every order before it reaches the market.

Because a single account can hold multiple simultaneous long and short positions opened by different strategies, positions cannot be offset by FIFO. Touch-Fire implemented an explicit position-reference model:

  • Opening orders carry a blank position reference; the adapter assigns one and returns it on the fill — using the venue execution ID, an incrementing integer that is also visible to downstream back-office feeds.
  • Closing orders carry the original position reference, so the exact opening position being closed is unambiguous — with the instrument looked up from the referenced opening fill rather than the (possibly expired) symbol.
  • Multiple fills at different prices on both opening and closing orders are handled and reported per execution.

Since RoboX keeps no positions, the Touch-Fire back office performs all bookings and then feeds RoboX the data it needs to display state:

  • Position reports after each fill — including the margin requirement for opening fills, and the opening/closing detail for offset (closed-out) positions.
  • Collateral reports — cash outstanding (deposits − withdrawals + P&L) and used margin (total initial required margin).
  • FIX position requests/reports so RoboX can download open and closed positions on restart, filtered by date and status.

RoboX is not directly connected to the broker, so all broker information flows through Touch-Fire over the most appropriate transport for each data type (latency is not critical here):

Broker → Touch-Fire (file / REST) → RoboX (FIX / REST)
  • New account information from the broker via file, passed to RoboX via a REST request.
  • Cash deposit/withdrawal updates via file, relayed to RoboX over FIX collateral messages.
  • Account-lock and prohibited-trading status held at Touch-Fire and not forwarded.

Results & Impact

RoboX strategies trade Japanese futures through a clean FIX interface, with Touch-Fire handling delivery mapping, routing and risk transparently. An explicit, execution-ID-based position-reference model lets independent strategies open and close overlapping long/short positions safely on a single account. Touch-Fire provided the position-keeping, collateral and reporting that the algo platform deliberately leaves out — a complete operational layer beneath the strategy engine.

Related Capabilities

Our pluggable parent/child algo engine — SOR, iceberg, stop and event-driven strategies. Sub-millisecond matching and multi-venue order routing. Position keeping, collateral, margin and post-trade reporting.

Have an Algo Platform to Connect?

Touch-Fire builds the adapters, routing, risk and back office that turn a strategy engine into a market-ready trading operation. Talk to us about your integration.

Contact us about algo integration.