Skip to main content
Before your agent can trade real capital, it has to prove it can manage risk on paper. Halyrd runs this evaluation automatically in the background from the moment your agent starts — no setup required on your part. When all three criteria are met, the agent earns the ELIGIBLE status, and the decision to promote it to live stays with you.

The three criteria

Your agent becomes ELIGIBLE when, over its evaluation window, it satisfies all of the following at the same time:
1

Positive expectancy after fees

The agent’s average trade outcome — accounting for wins, losses, swap fees, and gas costs — must be positive. This is calculated as:
Expectancy = (win rate × average win) − (loss rate × average loss) − fees
Fees here means the real cost of each trade: pool swap fees and the gas paid on BSC. An agent that looks profitable before costs but breaks even or loses after them is not ELIGIBLE. The definition of “profitable” never changes across configurations — only the thresholds in your chosen preset move.
2

At least the trade floor of closed trades

Expectancy calculated from two or three trades is noise, not signal. The trade floor is the minimum number of closed (fully exited) trades required before the expectancy figure is statistically meaningful. Until your agent has reached this floor, the evaluation shows “Evaluating — not enough trades yet” regardless of how good the early results look.The trade floor is set by your evaluation preset — see the table below.
3

No max-drawdown breach during the window

Even if expectancy is positive and the trade floor is met, a max-drawdown breach during the evaluation window disqualifies this configuration run. The agent’s risk ceiling must hold for the entire window.
All three criteria must be true simultaneously. Meeting two out of three does not make your agent ELIGIBLE.

The evaluation window

The evaluation window starts when your agent starts running and has a minimum duration that depends on your preset. At the end of the window, the agent checks whether all three criteria are met. If the trade floor has not been reached by the end of the window, the window automatically extends. The agent keeps running and keeps accumulating trades until it has enough closed positions for the expectancy figure to be trustworthy. There is no hard timeout — the window grows until the floor is met or a max-drawdown breach ends the run. This means an agent that rarely trades will take longer to reach ELIGIBLE. That is correct behaviour, not a bug.

Evaluation presets

When you spawn an agent, you choose a preset (or go Custom). Each preset sets the evaluation window length, the trade floor, and the drawdown caps together — they are designed as coherent sets.
PresetEvaluation windowTrade floorMax drawdownDaily drawdownMax consecutive losses
Conservative14 days15 trades15%5%3
Balanced (default)7 days10 trades20%8%4
Aggressive5 days8 trades25%12%5
  • Conservative — prove it slowly, fail safe. A longer window and higher trade floor mean the expectancy figure is based on more evidence before the agent is considered ready.
  • Balanced — the honest middle. The default for most agents; a reasonable window and sample size without being overly conservative.
  • Aggressive — reach ELIGIBLE faster, with a shorter window and lower trade floor. The evidence is shakier, the risk tolerances are wider, and the agent accepts higher drawdowns.
The profitability definition — positive expectancy after fees — is identical across all three presets. Presets only move the thresholds and windows; what “profitable” means never changes.

What ELIGIBLE means

ELIGIBLE is a signal, not a trigger. When your agent reaches ELIGIBLE, the dashboard shows a green outline badge reading “Ready for live — your call.” Nothing else happens automatically. You still decide when — or whether — to promote the agent to live trading. You can wait, keep observing, or promote immediately. Per the default configuration, there is no auto-promotion; that decision stays with you.
You can also promote your agent to live before it reaches ELIGIBLE. The Settings screen offers a force-promote path with an extra warning. The evaluation criteria exist to give you evidence — they are not a lock on the door.

Dashboard evaluation progress

While your agent is evaluating, the dashboard shows three live progress indicators so you can see how close it is to ELIGIBLE at any moment:
  • Profitability bar — current expectancy after fees across all closed trades in the window. Tooltip reads: “positive expectancy after fees, over the window.”
  • Trade count vs floor — how many closed trades the agent has accumulated versus the floor required by your preset.
  • Drawdown headroom — how much room remains before the max-drawdown limit would be breached.
These bars update in real time from WebSocket push events as new trades close and new equity snapshots arrive.

A note on capital and trade frequency

An agent running on a small account with the same fee structure as a larger one will naturally trade less often. Each trade must clear the fee filter — the agent only enters a position when the expected gain exceeds the swap fee and gas cost by a meaningful margin. On a small account, this bar is harder to clear, so trades are less frequent, and reaching the trade floor takes longer. This is the honest signal that the account’s capital is too low for its own cost structure at the current configuration. It is not a reason to weaken the evaluation criteria. The remedy is your capital and configuration choice — tighter presets, a watchlist with more frequent opportunities, or additional capital.
Halyrd makes no recommendations about account size. The evaluation criteria do not change based on how much capital you have. A configuration that stalls before ELIGIBLE is accurately reporting that it cannot cover its own costs at your current size — not asking for a lower bar.