How Systemic Trading works

What it is

Systemic Trading is a quant workbench for systematic traders. It brings three tools together in one interface so you never leave to move a strategy forward — from a raw backtest to a tradable, prop-firm-ready signal.

The three modules

Evidence over gut feel

The guiding idea runs through every module: you always see the range of possible outcomes and an honest verdict — never a single cherry-picked curve. Nothing here promises profit; it tells you how much of your result is edge and how much is luck.

Create your account

1. 1Open /get-started and complete the short onboarding — pick your experience, markets and a primary goal so the app tailors its defaults.
2. 2Sign in with email/password, Google or Apple. Everything runs in the browser, behind a single login — nothing to install.
3. 3You land on the app launcher at /apps. Switch between the three modules any time from the top bar.

Bring in your data

Two kinds of data feed the platform:

Backtests

An MT5 Strategy-Tester report (HTML) — drop it on the dashboard or the Import page. Trades, equity and key metrics are parsed and versioned in the Vault.

Price data

Daily price CSVs per symbol under Seasonality → Data (date + OHLC). This powers the Seasonality Scanner and the real market-regime analysis.

Import & Vault

Drop your MT5 HTML report and every strategy is stored, versioned and comparable in the Vault. Robustness and prop-firm checks run automatically on import.

Monte Carlo

Reshuffles and resamples your trades thousands of times to produce a confidence band around your equity — so you see luck vs. edge. The charts show the equity fan plus the full return and max-drawdown distributions with P5 / P50 / P95 markers.

Walk-forward

Rolling in-/out-of-sample folds reveal overfitting: a strategy that only works in-sample shows here. You get walk-forward efficiency per fold, in-sample vs out-of-sample return, and a stitched equity strip.

3D robustness surface

Visualizes performance across a parameter grid — is your result on a fragile ridge or a stable plateau? A plateau survives small parameter changes; a spike is likely a fit.

Regime analysis (real market data)

Analyze buckets your trades by the regime the ASSET itself was in — pulled from the OHLCV Data Storage, not inferred from your P&L. Trend comes from the asset's price vs. its own moving average (Bull/Bear); volatility from its realised volatility (Low/High).

How it resolves the symbol

The traded symbol is mapped to a stored symbol automatically (broker suffixes like .s / -Cs are stripped, e.g. GER40.s → GER40).

What you see

Realized equity by regime, a Bull/Bear × volatility performance table, per-trade regime tags and a regime-aware Monte Carlo. A banner shows the resolved symbol, bar count and coverage.

If no data is stored

When the asset has no price data yet, Analyze falls back to a trade-derived regime and tells you which symbol to import under Seasonality → Data.

Prop-firm EV

Simulates the daily-loss and max-drawdown rules of common firms and estimates your payout probability and expected value — so you know a challenge is worth attempting before you pay.

Risk re-scaling & Report

Risk re-scaling

Scales position size mathematically correctly to a target risk, so results across strategies are comparable on the same footing.

Report Developer

Export a clean, multi-page PDF report for any analyzed strategy — cover, KPIs, equity, distributions and verdict.

Data

Upload daily price CSVs per symbol (date + OHLC) under Seasonality → Data. Everything in this module — and the Strategy-Quant regime analysis — works on this stored data. Symbols are kept uppercase (e.g. GER40, EURUSD) at the D1 timeframe.

Explorer

Analyze a date window Seasonax-style: average curve, individual years, hit rate, average return and p-value. The optimizer scans ±N days around your window for a stronger one, and you can set the entry/exit side per pattern.

Bias map

The former Trading Assistant, now a tab here: it shows, per asset, whether it's seasonally bullish, bearish or neutral right now — sorted by strength, as a cards or spectrum view, with a monthly heatmap and 10/15/20-year cycles. Use it as a directional filter you time technically in TradingView.

Confidence A–D

Bundles hit-edge, p-value, sample size and consistency into one rating so you know how reliable a bias or window is. Recompute over the last 10 / 15 / 20 years to check it holds across cycles.

Events & Vision AI

Events

Test event-driven patterns to the exact day — Easter, Christmas, moon phases, Mercury retrograde — plus election-cycle filters. Turn any event window straight into a building block.

Vision AI

A strict, data-driven pass that surfaces the most valid seasonalities with honest warnings — plus a calendar/events vision that reads screenshots. No sugar-coating.

Validation, EA Lab & Calendar

Validation

Re-check candidates one by one (chart, individual years, KPIs) and mark them validated.

EA Lab & export

Combine building blocks into a portfolio strategy, check the trade distribution and a year filter, then export a MetaTrader 5 EA with broker symbol mapping (e.g. PU Prime).

Calendar + Telegram

Validated patterns land on a calendar; an optional AI agent pings you on Telegram before and on the day a pattern starts.

Two experts: MQL5 & Pine v6

An AI strategy builder with a dedicated expert per target platform — MQL5 (MetaTrader 5 Expert Advisors) and Pine Script v6 (TradingView) — each with strict compile-safety rules. Describe your idea in plain English or compose it from the vetted idea library (entries, filters, exits, risk blocks).

Backtesting (Python lab)

A dedicated Backtesting page runs a Python strategy on decades of data with a high-performance engine — pandas, plotly and qualitative market data included. Grid-optimize parameters, then export the result straight into Strategy Quant for the full robustness analysis.

Library

A dedicated Library page collects your saved strategies and reusable building blocks, so you can start a new idea from something proven instead of a blank page.

Editor, attachments & vision

Real editor

Generated files open in a coding workspace — line numbers, in-place editing, a collapsible history sidebar, Copy and Download (.mq5 / .pine), and 'To chat' to send an edited file back to the expert.

Attachments

Attach specs, trade logs or data files (.txt, .md, .csv, .json, .mq5, .pine — up to 300 KB) as context.

Visual answers

Ask for a pattern or setup and the Copilot renders it as an inline chart — candlesticks or lines with annotations — right in the conversation.

History, models & usage

Past chats live in the sidebar (per-language, stored locally). Pick the model per request — Sonnet (balanced), Opus (most capable) or Haiku (fastest) — and track requests and token usage. A rate limit keeps things fair.

What's included

Every plan ships with market data — no separate vendor, no API keys, no egress fees: 20+ years of daily CFD history, tick data (L1 quotes & trades) and OHLCV bars you can resample from seconds to monthly, across 40+ markets (FX, indices, metals, energy, crypto).

Importing your own data

Under Seasonality → Data you upload a daily-bar CSV per symbol. The parser is tolerant — it accepts QuantDataManager, Dukascopy ('Gmt time,…') and MT5 exports, with a date/datetime column plus Open/High/Low/Close.

Symbol mapping

Broker symbols differ from base symbols. The EA export maps base names to your broker (e.g. PU Prime: FX/metals/indices .s, softs/copper -Cs, DAX = GER40.s), and the regime analysis maps the other way to find your stored data. You import and analyze on clean base symbols.

A typical workflow

1. 1Have an edge idea — a mean-reversion system, a seasonal window, or a directional read for an asset.
2. 2Prove it: backtest in MT5 and import to Strategy Quant; run Monte Carlo and walk-forward. Keep only what survives.
3. 3Add seasonality: find or confirm a calendar window in the Seasonality Scanner; validate it and export an EA, or send it to the calendar.
4. 4Check the daily bias each morning in the bias map for your assets.
5. 5Execute within the rules: on a prop-firm account, the built-in daily-loss and max-drawdown checks keep you inside the limits.

Security & data

Authentication

Every app route sits behind login (email/password, Google or Apple via Supabase Auth). API requests carry a short-lived bearer token verified on every call.

Transport & headers

All traffic is HTTPS with HSTS. A strict Content-Security-Policy, frame-ancestors 'none' (no clickjacking), nosniff and a restrictive Permissions-Policy ship on every response.

API hardening

The backend validates every payload with strict schemas, rate-limits sensitive endpoints (e.g. the Copilot), returns structured errors with trace IDs and never echoes secrets.

Your data & secrets

Backtests and price data stay in your account's database. API keys (Anthropic, Telegram) live server-side as environment variables — never sent to or stored in the browser. Providers (Supabase, Vercel, Railway) process data under DPAs.

FAQ & troubleshooting

Which file formats are supported?

Backtests: MT5 Strategy-Tester HTML reports (UTF-16 is fine). Price data: daily CSV with a date/time column and Open/High/Low/Close.

Do I need to code?

No. Import, analysis and EA export run through the UI. The Copilot's Python lab is there if you want to go deeper.

My seasonality / regime pages are empty

You haven't imported price data yet. Go to Seasonality → Data and upload daily CSVs per symbol — the same data powers the regime analysis.

Is it suitable for prop firms?

Yes — daily-loss limit, max drawdown and payout EV are built in and simulated directly in the backtest.