The Problem With Most Tools
Retail traders typically work with RSI, MACD, moving averages, and maybe a Fear & Greed Index. These tools are not wrong — but they all come from the same place: price action. When they agree, it usually just means the price has been moving in one direction for a while. It's not independent confirmation. It's the same signal dressed differently.
Institutional traders solve this by pulling from genuinely uncorrelated sources — on-chain data, options positioning, macro liquidity, prediction markets, exchange flows. They look at what the market is doing across multiple dimensions at once. That's an expensive, full-time operation.
ODIN was built to do that work automatically, for a retail price.
The Conviction Score
ODIN's output is a single percentage: the Conviction Score. It measures how strongly ODIN's 16 independent market pillars agree on a direction. A score of 30% means weak, fragmented consensus — the kind of noise that produces false signals. A score of 75% means technical, macro, on-chain, sentiment, and derivatives signals are all pointing the same way simultaneously. That's a different category of signal entirely.
Why consensus matters: Any single indicator can produce a false signal. When 12 of 16 completely independent data sources align, the probability of that being noise collapses. The Conviction Score is a measure of signal quality, not just direction.
The score updates continuously as new market data arrives. It accounts for macro events — an FOMC meeting or CPI print can suppress conviction even when technical signals are strong, because professional traders step back during scheduled uncertainty. ODIN models that.
How the Score Is Built
Each pillar is scored independently, then combined through a weighted model. Here is how the process works:
15 Pillars Scored in Parallel
Each pillar pulls from its own data source and scores independently on a fixed scale. Technical Analysis, On-Chain Metrics, Polymarket, Exchange Flows — none of these share inputs. A bull signal from one cannot inflate another.
ML-Adaptive Weighting
Not all pillars are equally predictive in every market regime. ODIN runs logistic regression on recent backtest windows to continuously reweight pillars based on what has actually been working. In high-volatility regimes, macro and volatility pillars carry more weight. In trending markets, technical and on-chain carry more.
Macro Event Adjustment
Tier 1 macro events (FOMC, CPI, NFP, GDP) apply a conviction suppression factor in the 12-hour window around the event. ODIN does not try to predict Fed decisions — it steps back, like a professional would.
Volatility Gate
When the ATR percentile exceeds the 85th percentile — meaning current volatility is in the top 15% of its historical range — ODIN blocks bullish signals entirely. Entries in extreme volatility regimes have historically poor risk/reward.
Walk-Forward Validation
The backtesting engine uses walk-forward methodology: it trains on one historical window, then evaluates on the next window it has never seen. This is the institutional standard for preventing overfitting — common in quantitative funds, rare in retail tools.
Data Sources
ODIN pulls from 10+ independent data providers. No single source is trusted exclusively — where possible, signals cross-reference multiple inputs for the same pillar.
Technology Stack
ODIN is built on Python 3 and Flask, with a vanilla JavaScript frontend. The signal engine runs in a background thread, caching API responses to keep data fresh without hammering external rate limits. Signal computation runs across a 16-worker thread pool — all 16 pillars calculated in parallel.
The ML prediction engine uses a Gradient Boosting Classifier trained on 14 OHLCV features with no look-ahead bias. It generates 24h, 7-day, and 30-day direction forecasts with 1-sigma and 2-sigma confidence bands. Models auto-retrain every 7 days on fresh data. Models are validated using held-out test sets to reduce look-ahead bias.
The AI Analyst is powered by Claude (Anthropic). It receives the full current signal payload — all 16 pillar scores, conviction, regime, macro events — and can answer plain-English questions about what the market is doing and why.
The platform is served as a Progressive Web App (PWA) — installable on any device, with a service worker cache for offline access to the last-loaded signal state.
Built By
ODIN was designed and built by Renault Franklin — Certified Bitcoin Expert (Blockchain Council) with 8 years of hands-on experience in digital assets, trading, and Bitcoin education through RFBitcoin.com.
Every signal, pillar weight, and feature in ODIN reflects real trading decisions made over years of market participation — not academic theory. The platform was built because the tools Renault wanted to use didn't exist at a retail price point.
See It in Action
Open the live dashboard — free tier available with no credit card required.
Open Dashboard See Plans & PricingWant the full technical picture? Read the methodology →