After any ambient agent session, whether 15 minutes or a full trading day, you want to know: what did it find, what did it deploy, what did it pass on, and what did it miss? The opportunity board and missed-alpha ledger answer those questions.
The opportunity board
The opportunity board is a live ranked list of all candidates the agent has evaluated. Every instrument the agent scanned, researched, or traded gets an entry. Entries have one of three states:
| State | Meaning |
|---|---|
watching | Researched and promising. Monitoring for a better entry or waiting on regime confirmation. |
deployed | Currently in a live paper position |
rejected | Evaluated and rejected (with a reason) |
How entries are ranked
The board ranks entries by a composite score that blends:
- Quality (backtest tier A/B/C/D): the primary signal
- State bonus: deployed positions rank above watching, which rank above rejected
- Freshness: older entries decay in rank (stale thesis)
- Regime alignment: entries whose thesis matches the current market regime rank higher
This ranking is computed without hindsight. The agent only uses information available at the time of evaluation.
Reading the board after a session
Open the Ambient tab, click into your agent, and select Opportunity Board. You'll see:
Rank Instrument Setup Type Quality State Score
──────────────────────────────────────────────────────────────────
1 NIFTY23150PE PE momentum A deployed 9.2
2 RELIANCE EMA crossover B watching 7.1
3 HDFCBANK Mean-reversion B watching 6.8
4 NIFTY22800PE PE momentum C rejected 4.1
Rejection: noise_stop_mismatch -- spread too wide for stop distance
The rejection reason is critical. It is not "the agent didn't like it"; it is a specific, auditable gate that failed.
Common rejection reasons
| Rejection reason | What it means |
|---|---|
quality_below_threshold | Backtest quality tier below mission minimum |
noise_stop_mismatch | Intraday price noise exceeds stop-loss distance |
spread_too_wide | Bid-ask spread consumes too much of the expected P&L |
depth_insufficient | Order book too thin for the required quantity |
daily_cap_exhausted | Mission contract's daily deployment limit hit |
regime_misaligned | Instrument's thesis contradicts the current market regime |
already_deployed | Same underlying already has an open position |
raptorbt_failed | Backtest couldn't run (insufficient historical data) |
If you consistently see daily_cap_exhausted on good candidates, consider raising the
mission contract's daily deployment limit.
The missed-alpha ledger
The missed-alpha ledger records good opportunities that the agent couldn't take. These are situations where the agent identified a real edge but was blocked by a contract cap, a regime gate, a market impact failure, or a stop that fired too quickly.
The ledger is distinct from the opportunity board's rejected entries. Rejections are correct decisions (no edge). Missed-alpha entries are trades that would have worked but were blocked by a constraint.
What gets recorded in the ledger
- Strategies that passed quality gate but hit
daily_cap_exhausted - Positions that were stopped out by noise before the thesis had time to play out
- Instruments that the agent flagged as high-conviction but couldn't deploy (budget capped)
Reading the ledger
In the Ambient tab under Missed Alpha:
Date Instrument Setup Quality Block reason Est. missed P&L
────────────────────────────────────────────────────────────────────────────────────────
Jun 5 12:44 NIFTY23150PE PE momentum A daily_cap_exhausted +₹4,200 (est.)
Jun 5 10:15 RELIANCE EMA cross B noise_stop_mismatch +₹1,100 (est.)
The estimated missed P&L is computed from subsequent price action. It is approximate and uses perfect-exit assumptions. Use it as a signal of magnitude, not as a precise number.
Using both for continuous improvement
The board and ledger together tell you how to tune your agent's mission:
Seeing many quality_below_threshold rejections?
The market isn't offering setups that meet your quality bar. Either lower the minimum tier
(accept C-quality trades) or change the watchlist to instruments with more reliable signals.
Seeing many noise_stop_mismatch rejections?
The instruments you're watching are too volatile for the stop distances the agent is using.
Consider wider stops, smaller sizing, or different instruments.
Seeing many daily_cap_exhausted entries in the ledger?
You're finding more good trades than the contract allows. Raise the cap or widen the capital
allocation.
Seeing many stops fired before the thesis played out? The stop distance is too tight for the instrument's intraday noise. The market impact gate should catch this. If it didn't, the noise calculation may need tuning.
The agent's self-review cycle
Each day-start wake, the agent reviews the prior day's ledger automatically. If the same rejection reason appears repeatedly, the agent updates its beliefs:
- Repeated
noise_stop_mismatchon a specific instrument: the agent may drop that instrument from its watchlist or switch to a wider-stop strategy family. - Repeated
regime_misaligned: the agent may pivot its entire thesis to match the prevailing regime.
This is how ambient agents improve over time without manual intervention, though you can always intervene by editing the mission or the watchlist directly.
Next steps
- Ambient Agents intro: what ambient agents are and how to launch one.
- Arming alerts and interest watches: control when your agent wakes.
- Research, deploy, and monitor: the full autonomous loop from scan to exit.