Developer Update #8 - Dollar-Cost Averaging (DCA)

At Nexgent, our mission is to make trading agents as adaptive and resilient as possible. Markets are rarely linear, and volatility is often where the best opportunities lie - if handled correctly. That’s why we’ve been working on introducing Dollar-Cost Averaging (DCA) logic into our trading agents.

This is still very much a work in progress, but we wanted to share what’s live in our test environment today, how it’s already impacting trade behaviour, and where we see this feature evolving over the next few iterations.

Why Dollar-Cost Averaging Matters for Agents

DCA is a time-tested trading approach: instead of betting everything on a single entry, you add to positions when prices move against you, lowering your average entry price and improving recovery potential.

In traditional trading, humans do this manually. We want our agents to execute this systematically, with discipline, guardrails, and eventually, AI-driven adaptability.

With the first iteration, agents will:

• Detect when a position has dropped by a defined threshold (currently -17%).

• Add additional entries of the same or configurable size.

• Enforce cooldowns and minimum position ages to prevent over-exposure.

The result: more resilient positions that can recover faster when the market turns.

Current Technical Implementation

Here’s the baseline configuration we’re running right now in our test environment:

dca: {
  enabled: true,               // Enable/disable DCA feature
  trigger: -17,                // % drop from average purchase price that triggers DCA
  purchaseSizeMultiplier: 1.0, // Relative size (1.0 = 100% of original buy)
  maxDcaPerPosition: 2,        // Maximum number of DCA entries per position
  cooldownMs: 120000,          // 2 minutes between consecutive DCA triggers
  minPositionAgeMs: 60000      // Must hold position for at least 1 min before DCA
}

A few important points about this logic:

• Trigger point: The -17% is relative to the average purchase price. This ensures clear thresholds are respected.

• Sizing: Purchases are currently pegged to the original order size, but in future we’ll allow scaling multipliers (e.g., 1.5x larger buys on deeper drops). If your original purchase was 0.1 SOL, your first DCA would be 0.1 SOL, then the second DCA 0.2 SOL.

• Risk controls: maxDcaPerPosition hard-limits exposure. No matter what, the agent won’t chase indefinitely.

• Temporal controls: minPositionAgeMs and cooldownMs prevent rapid-fire averaging in flash crashes, creating breathing room between actions.

Dashboard & UI Enhancements

Alongside the backend logic, we’ve also uplifted the dashboard to make DCA activity fully visible:

• You’ll see layered entries when an agent averages down.

• P&L calculations adjust dynamically to reflect the blended entry price.

• Trade history logs will show primary entries vs DCA entries, so you can audit and learn from the agent’s behaviour.

This transparency is critical - we don’t just want agents acting; we want you to understand exactly how and why they acted.

Immediate Roadmap: Configurable Logic

For the initial rollout (expected later this week), DCA will be hard-coded at the agent level. This helps us validate the mechanics safely and thoroughly before exposing the feature to every user.

But the near-term goal is to make this configurable:

• Adjust your own trigger thresholds (e.g., -10%, -25%).

• Set custom sizing strategies (fixed, percentage-based, or progressive multipliers).

• Define cooldowns and maximum DCA entries per agent.

Think of this as “parameterized resilience” - giving you the knobs and dials to shape how your agents behave in choppy markets.

Long-Term Vision: Context-Aware AI DCA

The really exciting part lies just beyond the configurable stage: enabling our AI agents to make their own DCA decisions based not just on static thresholds, but on live contextual data.

We want agents to consider:

• Market microstructure: liquidity depth, slippage impact, and order book imbalance.

• Volatility regimes: are we in a sharp panic sell-off, or a slow grind down?

• Correlation data: is this token’s drop isolated, or part of a wider market move?

• Agent memory: how has DCA worked historically for this asset or strategy?

Instead of blindly averaging down every 17%, an advanced agent could say:

• “Liquidity is thin — averaging down here risks massive slippage.”

• “Volatility is spiking — wait until the cooldown resets before acting.”

• “This drop is correlated to a broad market sell-off — hold position until stabilization.”

This is where AI meets trade execution strategy, and it’s the direction we’re actively working toward.

Stop Loss & Beyond

DCA is only one side of the coin. The other is risk protection. In parallel, we’re developing enhanced stop loss logic - dynamic, trailing, and multi-tiered.

The combination of adaptive DCA + intelligent stop losses will give Nexgent agents the ability to not only stay in trades longer, but also know when to cut losses systematically.

Closing Thoughts

The introduction of DCA marks an important milestone: agents are no longer just entering and exiting, but managing positions dynamically.

This first iteration is simple, guardrailed, and hard-coded - by design. But the roadmap is clear:

1. Hard-coded logic (current testing).

2. Configurable parameters (coming soon).

3. Context-aware, AI-driven decisioning (future state).

As always, this is early work. We’re sharing progress transparently because community feedback helps shape the system. Expect rapid iteration, testing, and evolution as we refine this capability.

Stay tuned for the rollout later this week - and keep an eye on your dashboards. You’ll soon see your agents stacking entries intelligently, and this is only the beginning.