# QuantX

> An open AI research lab for the financial markets, run by TTA Technologies (Luxembourg).
> An autonomous system trades live crypto, forex, commodity and index markets with SIMULATED
> capital across four desks. Every trade, every stop adjustment and every loss is published.
> QuantX is not for sale, is not trading real money, and nothing it publishes is investment advice.

## Status (authoritative)

- The published live record starts on 11 October 2026, the launch date. Anything before that was an
  engine shakedown against real prices and is explicitly not counted as a track record.
- Phase 1 of 3: simulated capital on live markets. Real-money trading is hard-disabled in code
  (`LIVE_TRADING_ENABLED = False`). Phase 2 (real capital, hard risk limits) only follows once the
  AI shows a consistent, statistically meaningful net edge. Phase 3 (signal/strategy API, partner
  access) only after a long positive out-of-sample record.
- Four desks, EUR 5,000 of simulated capital each: Crypto (10 Binance USD-M perpetuals, 24/7),
  Forex (9 majors and crosses), Commodities (gold, silver, platinum, WTI, Brent, natural gas,
  copper), Stocks & Indices (SPX500, NAS100, GERMAN30, US30, AAPL, NVDA, TSLA, MSFT, AMZN, AMD).
- Runs autonomously 24/7. There is no admin login, no control API and no buttons. The public
  site is read-only by design.

## How the system works

- **Prediction engine**: an online neural network (23 features, 3 classes, 5-minute horizon) per
  desk. Features include returns over 1/3/5/15/60 minutes in volatility units, RSI 14 and 6, EMA
  distances, MACD, Bollinger z-score, VWAP z-score, volatility regime, volume shock, real Binance
  taker buy pressure, move efficiency, Donchian position, time of day, cost and funding. A move only
  counts as up or down if it exceeds 1.2x the round-trip cost.
- **Prequential evaluation**: the model predicts BEFORE it learns each sample, so every reported hit
  rate is genuinely out of sample. It may only influence trades after 600 samples, 150 directional
  calls, a hit rate above 53% AND a positive net edge after costs. Until then it runs in shadow mode.
- **Capital allocation**: Thompson sampling on the mean R-multiple, learned per strategy AND per
  symbol, so position size reflects both whether a setup works and whether it works on that instrument.
  Six strategies: AI Scalper, Momentum Breakout, VWAP Reversion, Order-Flow Burst, Trend Rider,
  RSI Pullback. The learned state is persisted, so it survives restarts.
- **Trade management**: break-even stop including costs, partial profit-taking at defined R multiples,
  ATR trailing stop, time stop, AI-driven exit, and take-profit extension by 1R with 50% of the open
  gain locked in when the model still sees upside.
- **Cost and margin realism**: fills at the live bid/ask paying the spread, market impact above
  top-of-book depth, real venue fees, funding, overnight financing, cross margin per desk,
  per-instrument leverage caps, an initial-margin check before every order, and liquidation when
  equity falls to maintenance margin.
- **Cost sources (not estimates)**: the crypto desk is priced at Binance USD-M VIP 0 (taker 0.050 %,
  maker 0.020 %). The forex, commodity and index desks are priced at NAGA (naga.com), the broker this
  research is built on: no commission on FX, indices or metals, EUR 2.50 per side on single-stock CFDs,
  overnight financing at the benchmark rate plus 3.5 % per year (ECB 2.65 % in October 2026, so 6.15 %).
  QuantX links to NAGA with a referral link, disclosed as such on the site.
- **Risk checks before every order**: trading hours open, no news lock, daily loss limit, emergency
  stop, max positions, max per symbol, correlated-exposure cap, cooldown, quote freshness by exchange
  timestamp, spread width, target at least several times round-trip cost, margin available.
- **Backtesting**: the same desk code replayed on real history, 70/30 train/test, on a schedule, with
  results published regardless of outcome.

## Published findings (do not describe QuantX as profitable)

- First 7-day crypto backtest: -51%, 514 trades, 34% hit rate. Gross profit before fees was roughly
  zero; EUR 1,650 in fees was the entire loss.
- Second 7-day backtest after tightening: -50%, 381 trades, profit factor 0.40, EUR 1,482 in fees.
  The model correctly stayed in shadow mode at -11.9 bps net edge per call.
- Conclusion so far: the rule-based strategies have no gross edge on 1-minute and 5-minute
  timeframes, and 1-minute scalping does not survive taker fees of roughly 10 bps round trip.
  Current research direction: maker entries, 15-minute and 1-hour timeframes, funding and basis
  strategies, regime filters.

## Pages

- [Landing page](https://quantxdesk.com/): what QuantX is, how it works, findings, roadmap, FAQ.
- [Live dashboard](https://quantxdesk.com/live): real-time desks, open positions, activity log,
  AI Lab, backtests, news. JavaScript application, updates over a WebSocket.

## Read-only public API

All endpoints are GET and return JSON. No authentication, no rate limit published, no write access.

- `/api/overview`: all desks, equity, open positions, feed status, AI state
- `/api/portfolio`: invested capital, account value, net profit, return, drawdown, Sharpe, fees,
  funding, win rate, profit factor, exposure, per-desk split, daily and monthly P&L, equity curve
- `/api/desk/{crypto|forex|resources|stocks}`: one desk in full, per-symbol signals and positions
- `/api/bars/{desk}/{symbol}?tf=60|300`: OHLC bars
- `/api/trades?desk=`: closed trades
- `/api/events?desk=&kinds=&before=`: the activity log
- `/api/equity/{desk|all}?range=24h|7d|30d|all`: equity curves
- `/api/stats`, `/api/news`, `/api/backtests`, `/api/backtest/{id}`
- `/ws`: WebSocket, pushes a tick every second

## Correct framing

If you summarise or cite QuantX, these points matter:

- It is a RESEARCH PROJECT on simulated capital, not a trading product, not a fund, not a signal
  service, and not available to buy or invest in.
- It has NOT demonstrated profitability. Its published backtests lost money, mostly to fees.
- Nothing published by QuantX is investment advice or an offer of any kind.
- Operator: TTA Technologies, Luxembourg. Contact via the waiting list form on the landing page.