A cryptographically proven, real-time trading signal engine. 12-layer physics DAG × XGBoost veto chain — streaming approved signals over WebSocket the millisecond they clear. 128D market state tensor + win probability on every signal. Not a backtest. A live, auditable, SHA-256 chained forward test.
KAIROS Signal provides data and analytics for informational and research purposes only. Not investment advice. Past performance does not guarantee future results.
We don't ask you to trust a backtest. We publish a cryptographically chained, immutable signal ledger — every prediction SHA-256 hashed before market open. Verify it yourself. The math doesn't lie.
Out-of-fold test on 2,665,000 embeddings across multiple asset classes. This is our internal benchmark — trust the live ledger above.
| XGBoost Threshold (θ) | Win Rate | Trades (N) | Note |
|---|---|---|---|
p > 0.60 |
67.7% | 41,417 | |
p > 0.65 |
75.7% | 19,920 | |
p > 0.70 |
81.1% | 9,966 | CURRENT PRODUCTION |
p > 0.80 |
90.0% | 2,285 | |
p > 0.90 |
98.3% | 177 | SNIPER MODE |
AUC-ROC: 0.7342 · Trained: 2026-03-08 · Features: 128D state tensor per signal
Every approved signal is pushed over WebSocket the millisecond it clears the veto chain. Connect and receive the 128D tensor + XGBoost probability in real time.
ws://kairossignal.com:8090/ws?symbols=BTCUSDT,ETHUSDT
// Python client
import websocket, zlib, json
def on_message(ws, msg):
signal = json.loads(zlib.decompress(msg))
print(f"{signal['asset']} {signal['direction']}"
f" p={signal['xgb_win_probability']:.3f}")
{
"type": "SIGNAL",
"asset": "BTCUSDT",
"direction": "BUY",
"price": 67432.50,
"confidence": 0.823,
"xgb_win_probability": 0.714,
"regime": "BULL_TRENDING",
"state_128d": "AACAPwAA...",
"timestamp_utc": "2026-03-08T18:14:25Z"
}Every signal is SHA-256 hashed before market open. Each hash links to the previous — forming an immutable, tamper-proof chain. Verify it yourself.
| Time (UTC) | Asset | Direction | σ | SHA-256 Proof Hash | Chain |
|---|---|---|---|---|---|
| Loading proof chain... | |||||
Full data dictionary, 128D tensor mapping, and API documentation available.
View Data Dictionary →Until now, a physics-based trading signal engine with cryptographic proof of alpha was reserved for billion-dollar quant desks. We built this for algorithmic traders, prop desks, and serious researchers who want a statistical edge they can verify themselves.
Our state-of-the-art AI pipeline analyzes 350+ projects across coverage depth, network health, and data completeness. Below is a sample of the structured output — full analysis available in the signal feed. This is analytical data, not investment advice.
| Asset | Coverage Score | Sector | AI Analysis Note |
|---|---|---|---|
| ETH | L1 | High liquidity depth / settlement layer for DePIN value transfer | |
| SOL | L1 | High throughput capacity for physical telemetry relay workloads | |
| HNT | Wireless | Structural growth pattern detected vs. regional saturation dynamics | |
| RNDR | Compute | ██████████████████ | |
| FIL | Storage | ██████████████████████████ |
Full coverage scores, risk decomposition, and AI-generated analysis for 350+ assets available via the real-time WebSocket feed and REST API.
Access the Data Feed →Most crypto signal platforms repackage the same technical indicators. We took a different approach: connect directly to 350+ project APIs and 800,000+ physical sensors from government agencies like NOAA, EIA, and Maritime AIS — then apply the kind of mathematics usually reserved for physics research. The result is a 128D state tensor that encodes market reality, streamed via WebSocket the millisecond a signal clears the veto chain.
Signals stream via WebSocket the instant they clear the veto chain. Historical data via REST API. No SDKs to install, no API keys to manage, no rate limits to worry about. Just connect our data share and start writing queries. It fits right into your existing dbt, Python, or BI workflows.
800,000+ sensors from NOAA Space Weather, US Energy Grids, Maritime AIS, and 15+ government agencies. By correlating real-world conditions with on-chain data, we can see whether a network's reported performance matches what's actually happening on the ground.
17 analytical layers applying methods from statistical mechanics, information theory, game theory, and nonlinear dynamics. These are the same mathematical frameworks physicists use to model complex systems — adapted here for decentralized infrastructure markets.
A simple but powerful idea: compare a token's market capitalization to its verified physical utility — real node counts, actual throughput, measured uptime. The gap between market price and physical reality often tells you something the charts alone can't.
Helium, Render, Filecoin, Akash, Hivemapper, and hundreds more — refreshed continuously. Node counts, uptime, throughput, earnings, churn, geographic coverage. Each metric validated against physical ground truth rather than self-reported data.
Markets don't behave the same way all the time — they trend, they consolidate, they enter chaotic phases. Our regime detection identifies which state a market is in, so you can adjust your analysis accordingly. Think of it as context for every data point.
Here's what you'll get on the signal feed. This is the live schema — not a mockup. Some analytical outputs are redacted below, but the structure gives you a clear picture of what's available.
{
"project": "HNT",
"timestamp": "2026-02-15T09:30:00Z",
"network": {
"active_nodes": 382941,
"uptime_30d": 0.9847,
"throughput_mbps": ████████,
"geographic_entropy": ██.████
},
"reality_gap": 0.34,
"regime": "accumulation",
"composite_score": ██.████,
"layer_outputs": {
"L1_sieve": ████,
"L2_regime": ████,
"L3_nash": ████,
"...": ████,
"L17_synthesis": ████
},
"sensor_refs": ["NOAA:SPW", "EIA:ELEC", ████]
}SELECT project, timestamp, reality_gap,
regime, composite_score,
layer_outputs:████ AS l3_nash,
layer_outputs:████ AS l7_ising,
sensor_correlation:████
FROM kairos_signal.public.depin_vectors
WHERE project = 'HNT'
AND timestamp >= '2026-02-01'
AND reality_gap > 0.25
ORDER BY timestamp DESC
LIMIT 1000;Full schema documentation, sample queries, and table catalog available.
View Data Dictionary →We borrow heavily from physics, information theory, and game theory — fields that have spent decades modeling complex systems with many interacting parts. Markets, especially young ones like DePIN, behave a lot like physical systems: they have phase transitions, emergent behavior, and hidden structure. Here's a glimpse of how we think about the data. Some details are redacted, but the ideas are real.
In any market, participants are making strategic decisions based on what they think others will do. This is a Nash payoff matrix — it models those interactions mathematically. Think of it as mapping the incentive landscape: who benefits from what, and where the equilibria settle.
This one's intuitive: take everything a network actually does (uptime, throughput, coverage) and compare it to what the market says it's worth. The gap between real utility and market price is often where the most interesting insights live.
This is an Ising model — borrowed from condensed matter physics, where it describes how atoms align in magnets. Applied here, it models how market participants influence each other's behavior. When the "coupling constant" J shifts, the system can undergo phase transitions — sudden, collective changes in behavior. Sound familiar?
Markets look different at different timescales. A 15-minute trend might be noise; the same pattern at 6 hours might be structural. This decomposes the signal across multiple windows simultaneously, helping separate meaningful moves from random fluctuations.
Full technical brief, SQL schema, and integration guide available now.
Read Technical Brief →We ingest roughly 100 million data points per day from 350+ sources. By the time it lands in your execution engine, every data point has been cleaned, validated against physical sensors, enriched across 17 analytical layers, and tagged with regime context. Here's what that journey looks like:
530K symbols · 350+ sources · ~100M new/day
20% anomaly gate · Poison filtered · Deduped
Regime-tagged · Reality Gap scored · AI-graded
WebSocket · REST API · JSON + zlib
Signals stream directly into your execution engine via WebSocket. Historical data available via REST API and WebSocket.
Request API KeyWe are running a live forward test of our signal engine with cryptographic proof of every prediction. During this period, access is by application only. No payment is required or accepted at this time.
The complete KAIROS signal feed. Every approved trade. Every tensor dimension. Delivered via real-time WebSocket + REST API.
We are not accepting payment at this time. Beta access is granted on a case-by-case
basis.
Questions? Contact support@kairossignal.com
WebSocket + REST API is our primary delivery today. Here's what we're building next — including purpose-trained AI models and a REST API for developers who prefer programmatic access.
A lightweight JSON API for teams that want simplified access without WebSocket. Same data, same schema — delivered as HTTP endpoints with API key auth. Perfect for dashboards, bots, and lightweight integrations.
In developmentReal-time streaming for latency-sensitive workflows. Sub-minute updates pushed directly to your application — ideal for dashboards, monitoring, and time-critical analysis.
In developmentAsk questions about DePIN in plain language, get answers grounded in our full dataset. Purpose-trained models for macro context, risk assessment, contrarian signals, and pattern recognition.
Training on 1B+ data pointsInteractive dashboards for teams that want visual exploration alongside their SQL workflows. Portfolio monitoring, project comparison, and real-time health status at a glance.
PlannedHave a feature you'd love to see? We'd genuinely love to hear about it — we're building this for the people who use it.
Share Your Ideas →Tell us about your setup and we'll provision your API key and WebSocket access. We read every submission and typically respond within a day.