Algorithmic Trading Platform

experimental

Quantitative futures research and execution platform

2024 to present

A futures research and execution engine. Find edge, prove it, trade it. Rinse and repeat.

TypeScript

Nuxt

Vue

Python

PostgreSQL

Supabase

Redis

BullMQ

DuckDB

Parquet

> the problem

Most trading strategies lose money. The ones that don't usually have weaker edge than they appear. I wanted to prove it conclusively, test as many strategies as I could think of, find out which ones actually work, and have the system ready to trade the winners live.

> the approach

The platform is a pipeline. Data from live feeds and 50+ years of historical archives flows through layered indicator engines across any number of timeframes. Strategies get backtested, replayed against real price and volume data, and scored using detailed performance metrics and risk analysis. The key decision: the research engine and the live trading engine are the same code. A strategy that survives testing can be deployed live without rebuilding anything. The strategy library grows constantly. Any futures market can be plugged in and tested.

> the lessons

The system has confirmed what I suspected: most strategies don't hold up under scrutiny. But a few do. Replay validation has been the single most valuable feature; it catches strategies that look great in backtesting but break down against real execution dynamics. The platform is live trading today. It works. It proves results. And it keeps searching. It is very challenging to find edges in the market. Thus the research replay engine is incredibly valuable.

> system architecture - a brief overview for those curious

Quantitative Futures Research & Execution Platform

System Architecture

Frontend

Trading Cockpit

Live action cards & confluence gaugesInteractive candlestick chartsReal-time data freshness monitoring

Research Lab

Strategy backtesting with SSE progressOHLC + indicator visualizationEquity curves & trade heatmaps

System Monitor

Indicator warmup & health statusAlgorithm managementJob queue oversight

▾

Nuxt + Vue + ECharts

▾

Data Sources

DxFeed

Live WebSocket streamFutures & equities1-minute resolution

Databento

Historical tick dataParquet exports50+ years depth

Supabase

PostgreSQLOHLCV + indicatorsAuth & RLS

TastyTrade

Broker APIOrder executionPosition mgmt

▾

ingest & normalize

▾

Indicator Engine

OHLCV Service

10 timeframes from 1m to 1WRolling buffers per marketLive & historical modes

Technical Indicators

6 indicator families computedacross all timeframes per bar

MomentumVolatilityTrend

Signal Matrix

Cross-timeframe confluenceReal-time status per market

▾

signals & confluence

▾

Natural Cycles Correlation Engine

Geocosmic Signal Research

Swiss Ephemeris (planetary positions & aspects)Solar & lunar cycle correlation analysisAspect filter discovery (e.g. Venus 60° on ES)Multi-decade statistical validation

PythonSwiss EphemerisPandas

Cycle Signal Database

Pre-computed planetary aspect windowsPer-market correlation tablesRegime-aware filter configurationsJSON configs consumed by strategy engine

PostgreSQLJSON configsParquet

▾

filter & enhance

▾

Strategy Engine

Strategy Registry

Pluggable strategy architectureRuntime registration via side-effect importsShared interface: live, backtest, replay

RAF Multi-TFSqueeze CompoundSolar LunarDonchianCRDSession Sweep

SimulationEngine

Per-algorithm instance, pooled & reusedN-timeframe confluence scoringDynamic position sizing via risk tiersEngine-managed trailing stopBracket orders: entry + target + stop

▾

execute

▾

Execution Modes

Live Trading

DxFeed WS → Indicators
→ Signal Generator
→ TastyTrade API
→ Position tracking
→ Supabase persistence
→ Discord notifications

BullMQRedisSSE

Backtesting

Parquet / Supabase
→ BacktestRunner
→ In-memory engine
→ MFE Calculator
→ SSE progress stream

DuckDBParquetMinute

Replay Validation

Walk-forward config
→ Live engine path
→ 1m bar injection
→ Broker Simulator fills
→ Ground-truth P&L

CLIBroker Simulator1m Resolution

infrastructure

Job Processing

BullMQ queue orchestrationTradingWorker (concurrency=1)Auto-recovery on startup

Broker Simulator

TastyTrade-compatible mockPrice-aware fill simulationBracket order lifecycle

Walk-Forward Optimizer

Parameter sweep & rankingOut-of-sample validationRisk-return scoring

VueNuxtTypeScriptNitroPostgreSQLSupabaseRedisBullMQDuckDBParquetEChartsVitestBun

> project/algorithmic-trading-platform[experimental]