Aegis Protocol - The Risk Doctrine That Refuses To Lose Twice
Aegis is the risk doctrine of the Glavior stack - a hard, deterministic refusal system that vetoes trades before they exist. R:R floors, extension blocks, exhaustion gates, correlation caps, and a permanent memory of losing patterns.
Aegis - The Risk Doctrine
Aegis is the protocol responsible for one thing and one thing only: **refusing trades the rest of the stack would otherwise take.** It is not a risk dashboard. It is not a warning system. It is a hard, deterministic veto layer that runs before any order is allowed to be constructed.
In the Glavior architecture, every approved trade is a trade Aegis was unable to kill. That asymmetry is the entire point.
Quick Specification
- **Protocol Code:** AEG
- **Role:** Risk doctrine · Deterministic veto layer
- **Status:** Live · Production
- **Position In Stack:** Upstream of order construction, downstream of perception (Genesis)
- **Outputs:** `APPROVED` · `BLOCKED(<reason_code>)` · `THROTTLED(<size_multiplier>)`
Mandate - The Burden Of Proof Is On The Trade
Most retail risk systems are permissive by default and restrictive on request. They let any signal through unless the user has manually configured a guardrail. Aegis inverts the contract.
Every candidate trade is treated as guilty until proven innocent. It must clear every Aegis gate or it does not exist. There is no "soft warning". There is no "are you sure?". Either Aegis returns `APPROVED` or the order is never constructed.
[INSIGHT] The cost of a missed trade is bounded. The cost of a taken bad trade is not. Aegis is calibrated against that asymmetry, not against trade frequency.
The Veto Stack - Six Hard Gates
Aegis evaluates six independent gates in strict order. The first rejection short-circuits the rest - the system never has to "balance" reasons to block.
Gate 1 - Risk:Reward Floor
Any trade whose planned R:R is below **2.0 after expected spread and slippage** is blocked. There is no override. A 1.8 R:R setup with "amazing" structure is still a blocked trade.
This single rule eliminates the majority of statistically losing systems that survive on win rate.
Gate 2 - Extension Block
If price is more than **~3% extended from VWAP** in the direction of the intended trade, entry is invalidated. The system's literal status output is: *"Entry Invalid - Extended. Wait for pullback."*
This is the anti-chase clause. It is the single most expensive rule for impatient users and the single most valuable rule for the equity curve.
Gate 3 - Exhaustion Gate
A short with 15-minute RSI below 30, or a long with 15-minute RSI above 70, is blocked pending cool-off. Aegis does not let the stack short into oversold or long into overbought, no matter how clean the structure looks.
Gate 4 - Correlation Cap
Aegis maintains a live map of active positions and pending signals. If a new candidate trade would push total directional exposure above the correlation cap (for example, three concurrent BTC-correlated longs), the trade is **throttled** to a smaller size multiplier or blocked entirely.
This is the layer that prevents "diversified" portfolios that are actually one bet wearing four tickers.
Gate 5 - Intent Conflict
Aegis reads the Genesis intent layer. A long on an altcoin while BTC is dumping > 0.5% in 1h is a hard block. A short on an altcoin while BTC is pumping > 0.5% in 1h is a hard block. No exceptions, no overrides.
[MISTAKE] "But the alt looks ready." The alt always looks ready. The King decides whether the court trades. Intent conflict is the most expensive lesson in crypto, and Aegis enforces it by contract.
Gate 6 - Pattern Memory
Aegis keeps a permanent, immutable log of every losing setup the stack has ever taken - the regime, the structure, the microstructure, the intent. If a new candidate matches a previously-failed pattern with high similarity and the regime has not meaningfully shifted, Aegis throttles or blocks.
This is not adaptive learning in the opaque sense. It is a published memory of mistakes that the system is forbidden to repeat without an explicit doctrine change.
The Reason Codes Are Public
Every Aegis block is emitted with a structured reason code. There is no "trade rejected" without a why.
```json { "protocol": "AEG", "decision": "BLOCKED", "reason_code": "EXTENSION_BLOCK", "detail": "price 3.42% above session VWAP", "candidate_fingerprint": "<sha256>", "emitted_at": "<ISO-8601>" } ```
That payload is written to the same append-only ledger that holds approved signals. Aegis blocks are part of the public record, not buried in a private log.
Throttling - The Middle Ground
Not every risk concern is a binary block. When a candidate is structurally valid but contextually weak - for example, a clean setup during a high-funding regime - Aegis returns `THROTTLED(0.5)` and downstream sizing is automatically halved.
Throttling is the system's way of saying *"this trade is allowed to exist, but it is not allowed to be expensive."*
What Aegis Refuses To Be
To set expectations honestly:
- **Aegis is not a trailing stop.** Stops are a Zenith and Paladin concern. Aegis is a pre-trade veto layer.
- **Aegis is not a sentiment filter.** It does not block trades because the news is scary. It blocks trades because the structural risk math is unfavourable.
- **Aegis is not adjustable per user.** The doctrine is platform-wide. There is no "aggressive mode" toggle that disables R:R floors. The institutional version of this product is the only version.
The Asymmetry Aegis Buys
A model with a 55% win rate and a 2.0 R:R floor is enormously profitable over time. A model with the same win rate and no R:R floor is a coin flip. The entire purpose of Aegis is to defend that asymmetry on every single decision, forever.
[AI] Aegis does not exist to be liked in the moment. It exists to be respected on the equity curve.
The Bottom Line
Aegis is the protocol that turns Glavior from "a system that signals" into "a system that **refuses**." Every approved trade is a trade that survived six independent attempts to kill it, and every rejection is published with a reason. That is what risk doctrine is supposed to look like.