A price prediction answers "where will this go?" A scenario answers "if this condition appears, what does it mean and what will I check?" The two look similar on a screen, but they lead to very different monitoring setups. This article explains the difference, shows how to write scenarios as if-then cards, and is explicit about what scenario monitoring cannot do.
Two different questions
Prediction commits to an outcome. "SYN-12 will reach 104 by Friday" can be right or wrong, and almost all of the work goes into estimating that one number.
Scenario monitoring commits to a response. "If SYN-12 closes above 104 with volume above its 20-session average, I will review whether the range thesis still holds" makes no claim about whether that will happen. The work goes into defining conditions precisely and deciding in advance what each one means.
Regulators are direct about the limits of the first approach. The CFTC's customer advisory on AI trading bots (external site) states plainly that "AI technology can't predict the future or sudden market changes", and warns that promises of very high or certain returns are red flags for fraud. Any ai trading bot pitched on forecasting accuracy deserves that scepticism.
What serious scenario work looks like elsewhere
Scenario analysis is a mainstream tool in risk management precisely because it does not pretend to forecast. When the Federal Reserve publishes its annual stress test scenarios (external site), the release says: "The scenarios are not forecasts and should not be interpreted as predictions of future economic conditions." The 2026 scenario has U.S. unemployment rising to a peak of 10 percent — a condition chosen to test resilience, not a projection of what will happen.
That is the mindset to borrow: a scenario is a condition worth being ready for, chosen for its relevance, not its probability.
Prediction and scenario monitoring compared
| Aspect | Price prediction | Scenario monitoring |
|---|---|---|
| Core output | A target value or direction | A condition is present / absent |
| What is tested | Forecast accuracy | Whether the condition was defined clearly and observed reliably |
| When it is "wrong" | Price misses the target | Condition was ambiguous, or the planned response did not fit |
| Typical alert | "Price approaching target" | "Condition B met: volatility above threshold for 3 bars" |
| Failure mode | Overconfidence in a single path | Too many scenarios, none reviewed |
| Review question | Was I right? | Did I see it, and did my response make sense? |
The last row matters most. Prediction reviews collapse into a scoreboard. Scenario reviews ask whether your observation and your process worked, which you can actually improve.
Anatomy of a scenario card
A scenario card is a short, structured note. Each card has six fields:
- Name — short and specific ("Range break with participation").
- Condition (if) — observable, measurable, with thresholds.
- Meaning — what the condition would suggest, in one sentence, stated as a hypothesis.
- Response (then) — what you will check or review. In a research workspace, this is an analytical step, not an order.
- Invalidation — a condition under which the card no longer applies.
- Expiry — a time limit after which the card is retired unreviewed.
Each condition maps to one or more alert rules, so the monitor does the watching and you do the interpreting.
Worked example: three cards on one instrument
The values below are illustrative, for the synthetic instrument SYN-12 in a demo replay. They are not market data.
Context: SYN-12 has traded between 98.00 and 102.00 for 12 sessions. Its 20-bar average true range is 0.40.
Card A — Range break with participation
- If: close above 102.00 and volume above 1.5× the 20-bar average, held for 2 bars.
- Meaning: the range may be resolving upward.
- Then: review regime label and whether the volatility rule has also fired.
- Invalidation: close back below 101.50 within 5 bars.
- Expiry: 5 sessions.
- Rules: level cross at 102.00 (re-arm below 101.70), volume spike at 1.5×.
Card B — Break without participation
- If: close above 102.00 with volume below the 20-bar average.
- Meaning: the move may lack follow-through.
- Then: mark for review; no escalation.
- Invalidation: Card A conditions met.
- Expiry: 3 sessions.
Card C — Volatility expansion inside the range
- If: average true range above 0.70 (1.75× its current value) while price stays between 98.00 and 102.00.
- Meaning: the quiet regime is ending without a direction.
- Then: review whether range-based cards A and B are still appropriate.
- Expiry: 5 sessions.
In the replay, Card C triggered first, at bar 140: range widened to 0.72 while price stayed inside. Card A triggered at bar 251. Card B never triggered. Note what was produced: a sequence of observed conditions with timestamps, each linked to a planned review. Nothing in it claims SYN-12 "was going to" break out.
Keeping scenario monitoring disciplined
Scenarios can sprawl just as forecasts can. Some working limits:
- Three to five live cards per instrument, at most. More than that and you are describing every possible path, which is prediction by enumeration.
- Every card has an expiry. Unreviewed cards are noise.
- Thresholds come from the data, for example multiples of recent range, not round numbers.
- Review cards, not outcomes. After a session, ask which cards triggered, whether the conditions were defined precisely enough, and whether the planned response was the right check.
This mirrors what supervisors expect of automated trading generally. ESMA's guidelines on systems and controls in an automated trading environment (external site) ask investment firms to monitor their electronic trading systems and algorithms in real time and to deal with identified problems "in order of priority". Monitoring defined conditions with defined priorities is the same idea at research scale.
The limits of scenario monitoring
Scenario monitoring is more honest than forecasting, but it is not a safety net.
Further limits:
- Conditions are only as good as the data. In Noctivex, all instruments and replays are synthetic and labelled as such. Behaviour on synthetic data does not tell you how real markets will behave.
- Triggering is not validation. Card A firing does not prove the "meaning" field was correct. Treat it as one observation.
- No execution. The response field describes a review step. Noctivex does not connect to brokerages, hold funds or place trades, and nothing here is a trading recommendation. See the risk disclosure.
From cards to a working monitor
To move from idea to setup: write the cards first, then translate each "if" into rules with thresholds, cooldowns and severity. Group the instruments on a watchlist with a stated purpose, run a demo replay, and record which cards triggered in your session journal. The result is a monitor that tells you what is present, in the terms you defined, and leaves the interpretation where it belongs — with you.