Quant research instrumentPaper onlyHuman decision
Research the signal. Respect the risk.
Qantrive runs strategy research end to end: features, signal hypotheses, a risk gate, out-of-sample and walk-forward validation, and a lifecycle that cannot promote itself past paper. Below is one real run, computed by the engine when this page was built.
01 · Signal
BUY confidence 83/100
RSI 18.6 oversold + MACD bullish crossover. Bar 189, stop set. One of 32 hypotheses in this run.
02 · Risk gate
PAPER APPROVED at default limits
Same signal, stressed book: BLOCKED by RULE-RISK-01, RULE-RISK-06, RULE-RISK-09, RULE-RISK-12.
03 · Validation
REJECT 3 critical failures
- Profit Factor 0.79 < 1.0 — REJECT
- Trade count 8 < 10 — statistically meaningless
- Out-of-sample return is negative
04 · Lifecycle
rejected next: backtesting
Model notes: none, display-only. The verdict came from the deterministic engine.
Deterministic sample series from the engine's own fixtures. Not market data, not a recommendation. Profit figures are deliberately not shown.
Risk field · default limits, enforced in code
- max risk per position
- 1%
- max portfolio heat
- 10%
- max open positions
- 5
- max correlation
- 0.70
- daily loss stop
- -3%
- drawdown stop
- -20%
01 · Regime
Context is classified before a signal is read.
Every symbol is placed in one of five regimes. The rules are checked in order and the first match decides, so a volatile market is never mistaken for a trend.
- 1HIGH VOLATILITY
ATR(14) ≥ 2.0% of closeChecked first, whatever the trend. - 2RANGE BOUND
ADX(14) < 20No meaningful directional trend. - 3RISK OFF
20-bar momentum < -5%Trending, with sharp negative momentum. - 4TRENDING BULL
EMA50 at or above EMA200Trend present, positive spread. - 5TRENDING BEAR
EMA50 below EMA200Trend present, negative spread.
02 · Validation
10 strategies, one series. 7 rejected.
Each strategy in the catalog is run on the same SPX sample series and judged against the same gates. A strategy that passes reaches paper, which is as far as anything automated goes. The series is synthetic, so this shows how the gates behave, not which strategy works.
| Strategy | Family | Signals | Lifecycle |
|---|---|---|---|
| RSI + MACD TrendProfit Factor 0.79 < 1.0 — REJECT | momentum | 32 | rejected |
| RSI Mean ReversionProfit Factor 0.77 < 1.0 — REJECT | mean reversion | 108 | rejected |
| VWAP ReclaimTrade count 7 < 10 — statistically meaningless | mean reversion | 10 | rejected |
| EMA 12/26 CrossProfit Factor 0.90 < 1.0 — REJECT | trend | 6 | rejected |
| Bollinger Band Breakout | breakout | 35 | paper trading |
| Donchian Channel Breakout | breakout | 66 | paper trading |
| ATR Volatility Bands | breakout | 126 | paper trading |
| Trend PullbackTrade count 0 < 10 — statistically meaningless | trend | 0 | rejected |
| Volume Spike ReversalTrade count 0 < 10 — statistically meaningless | volume | 0 | rejected |
| Momentum ContinuationProfit Factor 0.78 < 1.0 — REJECT | momentum | 9 | rejected |
03 · Lifecycle
Nothing automated goes past paper.
The discovery lifecycle is an allowlist. From paper trading the only automated move is to disabled. Anything beyond paper needs an explicit human decision outside the automated system.
- idea→ backtesting · disabled
- backtesting→ validation · rejected · disabled
- validation→ paper trading · rejected · disabled
- paper trading→ disabled
- rejected→ backtesting
- disabled→ backtesting
Product truth
Research software, not a promise of returns.
Qantrive does not present simulated or paper results as investment performance. Live trading is off: hard-coded false in the research engine and off by default in the backend. The live broker providers are placeholders with no API client. Nothing here is financial advice.
- Live trading in the engine
- Off · hard-coded
- Live broker connection
- None · placeholders only
- Highest automated state
- paper trading
- Model commentary
- Display-only