Developer Update #7 - New trading strategies and more

TL;DR

• New trading strategies - opt-in/out per strategy, JSON-defined strategies with 70+ parameters, and a new Signal Strength (1–5) score.

• Liquidity-trap protection v2 now live platform-wide; internal tests showed >90% reduction in agents entering positions in liquidity traps.

• Manual position closure from the UI with real-time P/L estimates at close.

• Webhooks (beta) for real-time trade signals and agent transactions so you can wire Nexgent to Telegram, X, or your own applications.

• More polish, safeguards, and groundwork for the upcoming ML-driven strategy adjustments.

🧩 New trading strategies - Opt-In/Out + JSON Strategies

We’ve rebuilt the foundations so strategies are modular and fast to iterate.

New: Strategy Selection UI

You can now opt in/out per base strategy from the Signals configuration screen. This makes it easy to A/B test setups across agents.

Strategy Specs: JSON, 70+ Parameters

Strategies are defined as simple JSON templates composed of:

• metrics (volume, price change, buy/sell ratio, whale participation, liquidity deltas, etc.)

• operators (>=, <=, crossesAbove, slope>, etc.)

• windows (e.g., 1m/5m/15m buckets + aggregations)

• weights (for scoring & strength)

Example

{
  "id": "whale_volume_spike",
  "name": "Whale Volume Spike Detection",
  "description": "Detects when whale trading volume increases significantly",
  "template": "volume_spike",
  "conditions": [
    { "metric": "changes.whaleVolume", "operator": ">=", "value": 1.5, "label": "Whale volume increase >= 150%" },
    { "metric": "current.whale.whaleVolumePercent", "operator": ">=", "value": 0.20, "label": "Whale volume >= 20% of total" }
  ],
  "logic": "AND",
  "enabled": true
}

This is the groundwork for the Strategy Manager, where you’ll be able to create and iterate on your own base strategies (through a no-code UI).

Current Strategies (refreshed and tuned)

• Volume + Buy/Sell Ratio Surge - simultaneous surges in volume and buy bias

• Price Reversal Detection - contrarian entries when price dips but buy pressure rises

• Stealth Trending Detection - momentum/trending algorithm for early movers

• Breakout Confirmation - structure + momentum confirmation with multi-stage checks

Early staging results: with the new strategy engine and parameter tuning we observed a win-rate lift from low-60s to >80% over the last 48 hours in the staging environment.

Pro tip: The best combo right now is Breakout Confirmation + Volume/Buy-Sell Surge + Stealth Trending Detection. Price Reversal Detection remains the riskiest (and currently the worst performer) - treat it as an optional, high-beta layer.

📈 Signal Strength (v1): 1–5 Confidence Score

Every emitted signal includes a 1–5 strength rating based on:

1. Excess over threshold per condition,

2. Weighted importance of each condition type, and

3. Breadth of confirmation (how many conditions passed).

Conceptual sketch

For each condition c:
  excess_c = (observed_c - threshold_c) / max(threshold_c, ε)
  score   += weight_c * clamp(excess_c, 0, cap)

strength = clamp( round(normalize(score, bands=[1..5])) , 1, 5 )

This immediately improves selectivity. In a near-term release, you’ll be able to set “Minimum Strength to Act” per agent (e.g., only act on 4–5). Longer-term, ML-driven confidence will supersede this heuristic.

🧪 Liquidity-Trap Protection

We’ve shipped a major under-the-hood upgrade that reduced trap incidents by >90% in testing and is now deployed platform-wide. It’s applied regardless of risk appetite.

We saw agents “winning” through traps under certain assumptions - not reflective of harsh on-chain reality. Details will stay private to avoid gaming, but the effect is simple: fewer bad fills, more honest exits.

🛠️ Manual Position Closure (New)

Live trading demands emergency brakes. You can now manually close any open position from:

• the main dashboard (Live Agent Trades), or

• the Open Trade detail modal.

You’ll see a live P/L estimate and % change at the moment you click close.

⚠️ Execution note: Trades execute at the market rate. Actual outcomes may differ from the estimate due to slippage and execution latency (network + routing). The estimate isn’t a guarantee.

We’ll be using this heavily during live tests to validate exits and route sanity.

🔔 Webhooks (Beta) - Connect Nexgent to Your Stack

You can now plug Nexgent AI into external systems for real-time notifications without polling.

Available events (initial):

• tradeSignals (new signal emitted)

• Agent transactions (fills/closures)

Example payload (trade signal)

{
  "event": "tradeSignals",
  "timestamp": "2025-08-09T07:46:57.274Z",
  "agentId": "585d1502-0f60-4f56-a47b-f694d6eafc2a",
  "data": {
    "id": 10597,
    "token_address": "9JErKCoJX7nmMsYvicoZfEdTdW1qiF9oeocHWyKLmoon",
    "token_symbol": "NEXGENT",
    "price_at_signal": 0.0002574,
    "created_at": "2025-08-09T07:46:57.206Z",
    "activation_reason": "Significant volume increase (123.52%) and Significant increase in trade count (115.38%)"
  }
}

Use cases: Push alerts to Telegram or X, trigger internal risk dashboards, or pipe signals into your own execution bots. Documentation will include recommended patterns for reliability and idempotency. We’ll expand the event catalog soon.

🧭 What’s Next — ML, No-Code Strategy Builder, More Events

• Machine Learning Strategy Tuning: continuous model ingest of trade outcomes to auto-adjust thresholds, weights, and windows. Think: learning from every near-miss and over-fit.

• No-Code Strategy Builder: define conditions, windows, and weights via UI.

• Agent Controls: per-agent minimum signal strength, tighter daily risk caps, and scenario-based slippage ceilings.

• Live trading: Results from the live trading tests will be made public as they come through

Final Thoughts

The theme this week is trust and control: better strategies, stronger safeguards, and clearer levers for you to shape agent behavior - all while we take the careful first steps into live execution.

If you build on top of the webhooks, or experiment with new strategy mixes, tell us what you find. Your feedback directly shapes the next set of releases.

— The Nexgent Team