πŸ‘ ML Ensemble Active

Always Awake.
Always Watching Your Back.

Your AI trading assistant that never sleeps. 4 base models + 1 meta-learner working together to protect your capital and find your edge.

Vigil caught the bear market before it hit. Will you be ready for the next one?

Create Free Account β†’
88.8%
Backtest Win Rate (WF-CV)
4+1
Base Models + Meta-Learner
107K+
Hours of Training Data
19.6
Backtest Sharpe Ratio

From November 2021 to December 2022, BTC fell from $68,601 to $15,643 β€” a 73% crash. A $10,000 buy-and-hold portfolio would have become $2,700. Vigil's ensemble detected the crash, blocked all entries, and kept the full $10,000 intact.

What Vigil Sees Right Now

Delayed snapshot. Create a free account to see model status.

Market Sentiment Loading...
SELL -- Confidence: -- -- BUY

πŸ”’ Sign up free to see more details

Last updated: --

What You'll Receive

Real signals. Real data. Delivered to Telegram and Discord in seconds.

🟒 BUY SIGNAL β€” BTC
Entry: --
Stop Loss: --
Take Profit 1: --
Take Profit 2: --
Confidence: 92%
Model Agreement: 100%
Bucket: Bitcoin Foundation
⚠️ Not financial advice. This is an example signal format, not a live recommendation.

Live Performance

Two sources of truth: rigorous backtesting and real paper trading. Both transparent.

πŸ“Š Backtest Results Walk-Forward Cross-Validation on 107,936 hourly candles (2013–April 2026)
88.8%
Win Rate
19.6
Sharpe Ratio
13.2%
Max Drawdown

Methodology: Walk-forward cross-validation β€” the model trains on past data and is tested on future data it has never seen, simulating real trading conditions. Think of it as a "time machine test" that prevents the model from cheating by peeking at future prices. Tested on BTC/USD hourly data from Gemini exchange, Jan 2013 – April 2026 (107,936 candles). Includes 0.1% slippage + 0.35% Gemini taker fee on every trade. 5-fold temporal splits for XGBoost, 3-fold for LSTM/RF/Transformer.

πŸ›‘οΈ Bear Market Protection Validated against the 2022 crypto winter (Nov 2021 – Dec 2022)
+73%
Outperformance vs Buy & Hold
0%
Capital Lost
100%
Entries Blocked
10,124
Candles Analyzed
97.2%
Time in Bear Mode
$0
Drawdown

Test period: Nov 1, 2021 – Dec 31, 2022 (14 months). BTC fell from $68,601 to $15,643. On a $10,000 starting portfolio, buy-and-hold would have ended at ~$2,700. Vigil's ensemble was bearish on 97.2% of the 10,124 hourly candles, blocked all entries, and preserved the full $10,000. The bear bounce detector also identified 64 flash crash opportunities (5%+ drops with RSI below 30) with a 53% win rate using 3% take-profit / 3% stop-loss.

⚠️ Past performance does not guarantee future results. Backtest results may not reflect real trading conditions. Trading cryptocurrency involves substantial risk of loss.

How Vigil Works

4 base models + 1 meta-learner. One mission: watch the market so you don't have to.

01

Vigil Watches

Every hour, Vigil analyzes 13+ years of real BTC price data from Gemini exchange β€” over 107,000 data points across 14 market indicators.

02

Models Vote

4 different AI models each make their own prediction independently. A 5th "referee" model combines their votes into one clear signal. When they all agree, you know it's serious.

03

You Get Alerted

You receive a signal with the exact entry price, stop-loss, take-profit levels, and a confidence score showing how much the models agree β€” delivered to Telegram or Discord in seconds.

πŸ“Š NEW: Interactive Vigil Chart

See BUY/SELL signals, regime detection, and model agreement directly on a live BTC chart.

Sign up to access β†’

The Ensemble

🌲 XGBoost v2

Spots trends in technical indicators. Great at finding patterns in structured data.

🧠 LSTM v1

Remembers price sequences. Understands that what happened in the last 24 hours matters.

🌳 Random Forest v1

200 decision trees voting together. Reduces the chance of any single model being wrong.

⚑ Transformer v1

The same AI architecture behind ChatGPT, adapted for price prediction. Finds complex patterns.

🎯 Meta-Learner (the 5th model)

A LogisticRegression model that learns which base models to trust in which conditions. It combines all 4 votes into one calibrated probability β€” the final signal you receive.

All models are tested using "walk-forward" validation β€” meaning they're trained on past data and tested on future data they've never seen, just like real trading. This prevents overfitting (the #1 reason most trading bots fail in production).

πŸ“Š Interactive Vigil Chart

See BUY/SELL signals, bear regime detection, and model agreement directly on a live BTC chart.

BTC/USD STRONG BEAR
Prob Down: 95.1% Confidence: 90.2% Agreement: 100% Mode: ensemble
1D 5D 1M 3M All
BUY | 75% SELL | 100% SL TP1 STRONG BEAR
Vigil AI ENSEMBLE
Model Votes
βœ“ XGBoost βœ“ LSTM βœ“ RF βœ“ Transformer βœ“ Meta-Learner
Real ML signals on a real chart

BUY/SELL markers, regime shading, model votes, SL/TP levels, and time range controls.

Sign Up to Access β†’

⚠️ Static preview for informational purposes only β€” not live data. Actual chart uses real ML ensemble signals.

Pricing

Start free. Upgrade when Vigil earns your trust.

Monthly Annual Save 20%

Observer

$0/mo
  • BTC sentiment direction (1h delay)
  • Basic performance dashboard
  • Bear regime alerts
  • Basic chart (delayed data)
  • Exact probabilities
  • Model agreement scores
  • SL/TP levels
  • Altcoin trade signals
  • Full interactive chart
Join Free

Guardian

$149/mo
  • Everything in Sentinel
  • Full chart + API data export
  • REST API access
  • Model attribution analytics
  • Kelly sizing parameters
  • Full on-chain data (flows, whale tx)
  • Feature drift reports
  • Shadow log analytics
  • Priority support
Get Early Access
πŸ”’ Lifetime Access β€” $499 one-time

Guardian-level access, forever. Limited to the first 50 members.

Reserve Your Spot β†’

The ML ensemble is trained on BTC data. Altcoin signals (ETH, SOL, etc.) use technical indicators gated by the BTC ensemble β€” when the ensemble is bearish, altcoin entries are blocked. This BTC-first approach protects capital across all assets during downturns.

Built by an Engineer, Not a Guru

No hype. No guaranteed returns. Just rigorous ML and transparent results.

Vigil is built by a software engineer with professional experience in cloud infrastructure and machine learning at AWS. Unlike signal groups that rely on gut feeling, Vigil uses a 4+1 model stacking ensemble β€” the same approach used by quantitative hedge funds, but made accessible for retail traders.

Every signal includes a model agreement score so you can see exactly how confident the system is. When models disagree, Vigil tightens stop-losses to protect you. When they all agree strongly, it lets winners run. During extreme bear conditions, Vigil automatically closes positions β€” because sometimes the best trade is no trade.

We publish performance transparently, including losses. If Vigil underperforms, you'll know. That's how trust is built.

☁️ AWS Infrastructure

Production systems at scale

πŸ€– Machine Learning

Ensemble methods, walk-forward CV

πŸ“Š Quantitative Trading

Backtesting, risk management, slippage modeling

πŸ”’ Security First

Encrypted secrets, JWT auth, audit logging

⚠️ Risk Disclosure

  • Trading cryptocurrency involves substantial risk of loss. You may lose some or all of your invested capital.
  • Past performance does not guarantee future results. Signals are for informational purposes only.
  • Vigil does not provide financial advice, investment advice, or recommendations to buy or sell.
  • ML models can and do produce incorrect predictions. Market conditions change.
  • Only trade with money you can afford to lose. Always use stop-loss orders.
  • You are solely responsible for your own trading decisions.

Let Vigil Watch Your Back

Create a free account. Early subscribers get special pricing.

Create Free Account β†’

Free Observer tier. No credit card required.

⚠️ Vigil is not a registered investment advisor. All content is for informational purposes only. Not financial advice.