Back to Projects
PlatformPrivate DeploymentQuant SystemsPrivate Codebase

Polyfokus

Prediction-market trading platform built around disciplined controls

A private trading platform for prediction markets that combines market ingestion, trader scoring, execution controls, and operator visibility in one disciplined system.

The public story is not the edge model itself. It is the architecture around it: clean data flow, conservative controls, observability, and a workflow that lets operators iterate without gambling on intuition.

Primary outcome
A high-pressure internal platform where market data, execution controls, and operator visibility all have to work together cleanly.
0.5-1s
Market Lag
Realtime chain and CLOB coverage designed for fast market reaction.
24
Ops Views
Rust TUI screens for live monitoring, bot control, and operator workflows.
3
Execution Layers
Go collectors, Python services, and Rust operations tooling working as one stack.
1
Private Ecosystem
Platform, runner, and backtesting surface packaged as a single internal product.
Platform shape

One ecosystem, not a script pile

Polyfokus is a coherent platform with collectors, storage, execution services, operator tooling, and backtesting surfaces arranged around a single operating model.

Public-safe story

The edge stays private

We can talk publicly about control surfaces, data integrity, and execution discipline without disclosing the scoring logic or strategy internals that create the actual edge.

Marketing value

Proof of systems rigor

This project signals that we can design high-pressure systems where latency, controls, and observability all matter at once.

Architecture

Built like an operations platform

The platform is split along operational boundaries so each layer can be trusted, reasoned about, and replaced independently.

Data collectors

Go services ingest markets, prices, orderbooks, and chain activity into the warehouse with pipeline freshness as a first-class concern.

PostgreSQL for configs and state
ClickHouse for market data history
Redis for live control and heartbeats

Execution services

Python services handle bot lifecycle, execution, wallet flows, and backtesting so strategy iteration does not contaminate the ingestion layer.

Runner separation for safer bot isolation
Paper and live modes
Explicit service boundaries

Operator tooling

A Rust TUI turns the system into something operators can actually run, inspect, and intervene in under pressure.

Realtime dashboards
Bot lifecycle visibility
Redis-backed status updates
Controls

Risk boundaries are part of the product

The interesting part is not just signal generation. It is the layer of gates, size controls, exposure limits, and execution policies that keep the system legible and survivable.

Spread-aware entries instead of blind mirroring
Portfolio-level exposure limits and per-market caps
Paper-to-live graduation paths instead of instant deployment
Separation between detection, approval, and execution concerns
We deliberately describe the control framework here and not the proprietary scoring internals.
Iteration loop

Backtesting is attached to the live stack

Historical simulation is not treated as a separate research toy. It exists to accelerate the feedback loop between live observations, parameter changes, and operational confidence.

Historical replay

Backtests let us test assumptions against the same ecosystem concepts used in production, rather than inventing a parallel research environment.

Operator feedback

Live dashboards, alerts, and bot control surfaces create a clear handoff between research and day-to-day operations.

Internal Systems

collin/polyfokus
collin/polyfokus-backtest-app
Polymarket-Bot

Related Systems

Smart Launcher

Need a system with hard constraints and clear operator visibility?

This is the kind of work we do best: complex systems where correctness, controls, and observability matter more than marketing fluff.