About | Nova Group

The idea

Nova Group has been studying financial markets since 2020. Our central idea is one sentence: price is a consequence of the market's flow of trades and orders — not its cause. Accept that and the work becomes clear: you can't understand price until you understand the orders behind it.

A large part of those years went into exactly that — collecting and cleaning the trades and orders in the market's depth at levels 1, 2 and 3, across venues from CME to centralised and decentralised crypto exchanges and the Tehran Stock Exchange. On that data we build machine-learning, deep-learning and reinforcement-learning models to reach a causal understanding of order flow, and to forecast price from it.

Our method

Every project we build stands on these same three layers, from the bottom up. The lower layer goes nowhere without the one above it — and the upper layer does not exist without the one below.

1

An independent platform, built per project

For each project we build its own data and execution platform from the ground up, independent of any off-the-shelf tool.
2

A deep order-execution simulator

The pillar of our work. An ordinary backtest assumes every order fills instantly at the quoted price — an assumption that flatters the numbers. We rebuild the venue's execution engine: orders queue, wait their turn, fill in part or stay at the back of the queue. This is the rule in every one of our projects, and its form changes with the market: in an order-driven market, rebuilding the queue and the execution of orders; in a liquidity pool, placing capital in the rebuilt pool and working out its fees and losses.
3

Analytics, charts and predictive models

We start from raw data: the flow of trades, orders and price changes. Statistical and analytical work takes that flow apart and turns it into features that can explain the market's next move. On those features we train machine-learning, deep-learning and reinforcement-learning models, and what comes out of that path is a prediction algorithm.

Trade and order flow Statistical analysis Model training ML · DL · RL Prediction algorithm

Charts and analytical tools are built in this layer too, so that every step of the path can be seen and checked. This is the last layer, not the first — without the two beneath it, no number means anything.