An automated multi-layer crypto futures signal engine — built and privately run since early 2025.
Aminoid Alpha scans a curated watchlist of Binance USDT-M perpetual futures every 15 minutes, scores each coin across 22 technical, structural, and market-microstructure indicators, and — only when a strict multi-layer gate is cleared — pushes a fully-formed trade signal (direction, entry zone, stop-loss, two take-profit levels, leverage guidance, and an AI-generated reasoning summary) straight to a private alert channel.
This repository is documentation, not source code. It exists so that traders, quants, and developers can understand what the system does and why it works, without exposing the exact engine that runs it. If you want access to live signals, see Getting Access below.
- At a Glance
- Track Record
- How It Works
- What's Documented Here
- What's Not Public
- Getting Access
- Risk Disclaimer
| Markets covered | 32 curated coins, Binance USDT-M perpetual futures |
| Scan interval | Every 15 minutes (watchlist split into two alternating halves — each coin re-scanned roughly every 30 minutes) |
| Indicators | 22, across 5 independent analytical layers |
| Timeframes used | 15m, 1H, 4H (multi-timeframe confluence) |
| Scoring model | Independent long/short scoring out of 11.0, not a single biased vote |
| Signal gate | Multi-condition — score threshold, layer-participation minimum, momentum/reversal confirmation, disagreement cap |
| Output per signal | Direction, entry zone, stop-loss, TP1, TP2, risk/reward, leverage suggestion, confidence score, AI-written reasoning |
| Delivery | Real-time private alert channel |
| Live since | ~6 months of continuous private operation and iteration |
| Status | Actively maintained — indicators, gate logic, and risk rules are still being refined |
Over approximately 6 months of live, manual execution (I place every trade myself based on the signals — this is not a fully autonomous bot placing orders), Aminoid Alpha has produced:
- 8,000+ completed trades
- ~90% win rate
These figures are self-reported from my own trading activity and have not yet been independently audited. I'm not asking anyone to take that on faith — proof of performance and a free trial are available on request (see Getting Access). The best way to evaluate any signal system is to watch it in real time, so I'd rather show you than just tell you.
A structured, timestamped forward log is being maintained going forward specifically so performance can be verified independently over time, rather than taken solely on past word.
At a high level, every 15-minute cycle runs through five stages:
- Data Collection — Pulls fresh 15m / 1H / 4H candles, funding rate history, and open interest for each coin in the active watchlist half, plus a BTC-wide market context read (BTC's own trend and BTC dominance movement).
- Indicator Computation — Calculates all 22 indicators across 5 layers (Trend, Reversal, Momentum, Entry Timing, Microstructure) for every coin, independently for both the long and short case.
- Scoring & Gating — Tallies an independent long score and short score per coin. A signal only survives if it clears a minimum score, gets confirmation from at least 3 of the 5 layers, has genuine momentum or an early-reversal cluster behind it, and isn't undermined by too many conflicting indicators.
- AI Reasoning Layer — Every signal that clears the gate is handed — as pre-computed, already-final numbers — to an LLM whose only job is to write a clear, human-readable explanation of why the setup qualified. The AI never sees raw price data and never sets the score, direction, or risk levels itself; every number in the final alert is deterministic, calculated by the engine before the AI is ever called.
- Delivery & Deduplication — A price-aware cooldown makes sure the same underlying move isn't re-announced as a "new" signal a few hours later just because a lagging indicator re-triggered. Only genuinely new setups are pushed out.
See ARCHITECTURE.md for the full pipeline breakdown and INDICATORS.md for every indicator used, and SCORING.md for how the scoring and gating philosophy works.
This repo is intentionally detailed about concepts, structure, and philosophy:
ARCHITECTURE.md— the full data pipeline, the 5-layer model, the AI reasoning layer, watchlist design, dedup logicINDICATORS.md— all 22 indicators, what each measures, and which layer/timeframe it belongs toSCORING.md— how long/short scoring works, why it's independent rather than vote-based, what the multi-layer gate checks, how leverage and confidence are derivedSAMPLE_SIGNAL.md— a real (sanitized) example of what a signal looks like end-to-endFAQ.md— access, cost, risk, and common questions
To keep this system effective for the people who use it, the following are deliberately not published:
- Source code / implementation
- Exact scoring weights, thresholds, and gate constants
- Exact formulas for stop-loss / take-profit placement
- The specific automation platform and infrastructure it runs on
- API keys, credentials, or any operational internals
This isn't about hiding how trading signals work as a category — that's well-understood, public knowledge in the industry. It's about not handing away a working, tuned implementation that took six months of live iteration to get right.
Aminoid Alpha isn't sold as a packaged product or open-source tool — access is granted directly.
Contact me on Telegram: @AminMunir
I'm happy to:
- Share proof of past performance
- Offer a free trial period so you can evaluate signal quality yourself
- Answer questions about how it fits your trading style before you commit to anything
Aminoid Alpha generates technical trading signals for informational purposes. It is not financial advice. Cryptocurrency futures trading, especially with leverage, carries substantial risk of loss and is not suitable for everyone. Past performance — including the figures stated in this repository — does not guarantee future results. Always do your own research and only trade with capital you can afford to lose.
This repository will be updated as the system evolves. Star/watch it to follow along.