In this section we will utilise our multi-agent simulation engine (called IPOR Test Framework, ITF for short) to backtest the impact of model assumptions and different market conditions on the PnL of the pools.

ITF is a simulation engine for Interest Rate Swaps.

Why did we build the IFT market simulator? The primary motivation was to offer a robust testing and validation platform, allowing our team and other traders to confidently assess and refine trading strategies before taking them live. The IFT simulator enables backtesting on historical data, giving us insights into how algorithms would have performed historically. This tool is invaluable in understanding the potential risks, as it lets us evaluate metrics such as maximum drawdown, volatility, and more under various market scenarios. Notably, it offers both the perspectives of market makers and traders. After initial tests, strategies can undergo refinement based on the simulator's feedback, ensuring optimal performance. We also recognize the importance of stress testing. We can identify potential vulnerabilities by exposing algorithms to extreme or unlikely market conditions. Such comprehensive testing, coupled with successful simulation outcomes, can bolster confidence in a strategy. Moreover, the simulator provides insights into the potential market impact of trades, especially in less liquid markets, making us aware of factors like slippage. On the technical front, it ensures our trading system's infrastructure is free from software glitches that could hamper trade execution in real scenarios. Beyond its technical benefits, the IFT simulator has become an educational tool. Our interns have benefited immensely, gaining hands-on experience without the risks of live trading. Furthermore, simulations offer a cost-effective alternative to live trades, saving us from potential hefty gas fees.

This section is split into two parts. In the first part, we will briefly describe the agents utilised for the simulations. In the second part, we will look at some of the PnL plots and AMM quoting policies resulting from the Base Spread Model and Demand Model.

Agents

The agents utilized belong to one of the four families of strategies:

For reference, below is a flowchart of the Global Points Time Window agent

Global Points Time Window agent flowchart

Global Points Time Window agent flowchart

Agents overview

Global Points Time Window: This agent opens a receive fix IRS contract on the time window maximum on its way to the minimum, while on the minimum point, this agent opens a pay fix. The duration is chosen to be the optimum duration. See the image below for an example where the maximum and minimum are dotted in colors.

Global Points Time Window. The green dots represents contract OPEN while red dots are contract CLOSE

Global Points Time Window. The green dots represents contract OPEN while red dots are contract CLOSE