Signals Documentation

From zero to graded in four steps. You'll need a Paradox Alpha account — sign up is free and takes under a minute.

1. 1

   Create an API key

   In your dashboard, visit Signals → Keysand click Create new key. Copy it immediately — we never show it again.

   bashCopy

   export PX_KEY="px_live_sk_4f3c...3a"

2. 2

   Register a strategy

   Give it a name, a universe, and a horizon. Strategies start in draft status until you publish.

   bashCopy

   curl https://api.paradoxalpha.com/api/v1/signals/strategies \
     -H "Authorization: Bearer $PX_KEY" \
     -H "Content-Type: application/json" \
     -d '{
       "slug": "aurora-momentum",
       "name": "Aurora Momentum",
       "description": "Cross-sectional 12h momentum on top-100 perps",
       "universe": "binance_perp_top100",
       "horizon": "8h",
       "submission_cadence": "8h"
     }'

3. 3

   Submit your first signal

   Scores are unit-free — we z-score them cross-sectionally. Positive = long, negative = short.

   bashCopy

   curl https://api.paradoxalpha.com/api/v1/signals/strategies/aurora-momentum/submissions \
     -H "Authorization: Bearer $PX_KEY" \
     -H "Content-Type: application/json" \
     -d '{
       "schema_version": "1.0",
       "submission_id": "run-2026-04-10-1200",
       "as_of": "2026-04-10T12:00:00Z",
       "scores": {
         "BTCUSDT": 0.82,
         "ETHUSDT": 0.64,
         "SOLUSDT": -0.31,
         "BNBUSDT": 0.42,
         "XRPUSDT": -0.55
       }
     }'

4. 4

   Watch your grade

   Metrics refresh every 15 minutes. After 30 submissions you'll receive a grade from D to S. Above grade A, your strategy becomes eligible for live bot subscriptions.

Paradox Signals uses a single canonical format: cross-sectional {symbol: float}scores, with optional per-symbol enrichment. One schema covers every use case.

{
  "schema_version": "1.0",
  "submission_id": "run-2026-04-10-1200",
  "as_of": "2026-04-10T12:00:00Z",
  "valid_until": "2026-04-10T20:00:00Z",
  "universe_hint": "binance_perp_top100",
  "scores": {
    "BTCUSDT": 0.82,
    "ETHUSDT": 0.64,
    "SOLUSDT": -0.31,
    "BNBUSDT": 0.42,
    "XRPUSDT": -0.55
  },
  "enrichment": {
    "BTCUSDT": {
      "direction": "long",
      "conviction": 0.9,
      "horizon": "8h",
      "expected_return_pct": 1.8
    }
  },
  "notes": "trend regime, low vol"
}

All endpoints live under /api/v1/signals/. Authenticate with either a JWT (dashboard) or an API key (programmatic).

POST /strategies

Register a new signal strategy. JWT only.

// Body
{
  "slug": "aurora-momentum",
  "name": "Aurora Momentum",
  "description": "Cross-sectional momentum",
  "universe": "binance_perp_top100",
  "horizon": "8h",
  "submission_cadence": "8h",
  "is_public": true,
  "tags": ["momentum", "trend"]
}

POST /strategies/{slug}/submissions

Submit a signal tick. API key or JWT. Returns 202 on accept, 200 on duplicate, 422 on validation failure, 429 on rate limit.

// Response (202)
{
  "submission_id": "run-2026-04-10-1200",
  "server_received_at": "2026-04-10T12:00:01.234Z",
  "accepted": true,
  "duplicate": false,
  "coverage": { "symbols": 5, "benchmark_pct": 5.0 },
  "warnings": []
}

GET /strategies/{slug}/metrics

Retrieve rolling evaluation windows (last 30 by default).

GET /leaderboard

Top strategies by composite score. Public, no auth required.

GET /keys · POST /keys · DELETE /keys/{id}

Manage API keys. JWT only. POST returns the raw key once.

Every live strategy is re-graded every 15 minutes on a rolling 24h window. We reuse the same asymmetric long/short rank-IC engine that powers our in-house Autopilot — no manual review, no politics, just math.

• Rank IC: Spearman correlation between your scores and realized forward returns.
• Long / Short IC: split into top and bottom quintile separately.
• Hit Rate: directional accuracy.
• Sharpe: annualized Sharpe of a paper portfolio driven by your signal.
• Coverage: fraction of the benchmark universe your signal hits.
• Freshness: latency from as_of to server_received_at.

Composite = 0.5 × rank_ic + 0.3 × sharpe_norm + 0.1 × hit_rate_norm + 0.1 × coverage

Grades: S (>0.25, 100+ subs), A (>0.15), B (>0.08), C (>0.03), D (below). UNRATED under 30 submissions.

Look-ahead prevention: forward returns are always measured fromserver_received_at, never from your claimed as_of. Submissions back-dated by >15 min are rejected outright.

A minimal Python SDK is planned for Phase 4. For now, the API is simple enough to wrap in a few lines of requests:

import os, requests
from datetime import datetime, timezone

BASE = "https://api.paradoxalpha.com/api/v1/signals"
KEY = os.environ["PX_KEY"]

def submit(slug: str, scores: dict[str, float], submission_id: str):
    return requests.post(
        f"{BASE}/strategies/{slug}/submissions",
        headers={"Authorization": f"Bearer {KEY}"},
        json={
            "schema_version": "1.0",
            "submission_id": submission_id,
            "as_of": datetime.now(timezone.utc).isoformat(),
            "scores": scores,
        },
    ).json()

print(submit("aurora-momentum",
             {"BTCUSDT": 0.82, "ETHUSDT": 0.64, "SOLUSDT": -0.31,
              "BNBUSDT": 0.42, "XRPUSDT": -0.55},
             submission_id="run-" + datetime.now().strftime("%Y%m%d-%H%M")))

Errors use a consistent envelope:

{
  "detail": {
    "code": "signal_too_old",
    "message": "as_of is 23.4 minutes in the past (max 15)",
    "param": "as_of"
  }
}

For non-technical questions about payouts, grading philosophy, and contributor onboarding, see thesignals landing page.

Need help? Email signals@paradoxalpha.com.