Select expiry and date in Control Desk to launch the anomaly engine.
Operating Signals
Signal cards hidden.
The sections below include research proposals such as news matching, pattern memory, and outcome analysis. These are not inputs to the current activity score.
Detection Framework
Reliable detection starts by asking whether participation, positioning, price behavior, and peer context are all abnormal for the same time of day. A single volume spike is a clue; a multi-signal cluster is a tradable investigation.
Peer z-score > 1.5 or absolute Vol/OI above 1.0 for stock futures and 1.5 for index futures.
Detects fresh participation that is large relative to outstanding positions. This is often the earliest sign that a contract is being actively repriced.
Price change and OI change classify long buildup, short buildup, short covering, and long unwinding.
Separates initiative positioning from liquidation or covering. The same price move means different things when open interest is expanding versus contracting.
Price z-score > 1.5, stock future range > 2% of last price, or index future range > 1%.
Flags contracts where the intraday range is no longer normal for the time of day. This helps catch breakouts and failed auctions early.
Triggered symbols divided by sector universe, then ranked by average confidence and leader count.
Turns isolated symbol alerts into a market-structure read. A single high score is useful; five related high scores can reveal institutional rotation.
Notional value, value per trade, rapid volume intensity, and rapid OI shifts.
Futures snapshots do not expose the full order book, so the system uses turnover concentration and delta speed as a proxy for aggressive flow.
Match alert time to news, results, policy events, bulk deals, or sector headlines, then measure follow-through.
A headline-driven spike needs a different workflow from a quiet accumulation setup. Event tags help distinguish tradable repricing from one-bar noise.
Historical Data
Historical data should do more than backfill charts. It should define normal behavior, label old anomalies, identify false positives, and compare the live session with similar prior intraday patterns.
Compare every 5-minute, 15-minute, and hourly bucket with the same bucket across prior sessions.
Opening volume should not be compared with lunch volume. Build baselines by symbol, day type, expiry week, and time bucket so the anomaly is judged against the right context.
Require either self-history abnormality, peer abnormality, or both for stronger confidence.
A high-volume stock can look unusual only because it is always active. The model should compare each symbol to itself and to the current futures universe.
Store every fired alert with next 30-minute, close, and next-session outcomes.
Historical anomalies become a labeled library. The system can show whether a current setup resembles a continuation, squeeze, failed breakout, or liquidation pattern.
Find older sessions with similar RVOL, OI delta, range expansion, sector breadth, and timing.
When a live name fires, compare it with historical vectors instead of only looking at current rank. Similarity helps a trader judge if the signal is early, mature, or exhausted.
Penalize stale, illiquid, single-bucket, and expiry-roll distortions before pushing alerts.
A reliable system must say no often. Confidence should fall when a move has weak liquidity, poor sector confirmation, or no second signal after the first burst.
Indicator Stack
The strongest setup is not the highest value in one column. It is a pattern where volume, OI, volatility, structure, and peer confirmation all move out of their normal bands.
| Indicator | Trigger Logic | Why It Matters | UI Cue |
|---|---|---|---|
| Relative volume by bucket | Current bucket >= 2x median or z-score >= 2 | Captures participation before price has fully moved. | Heat bar with bucket percentile. |
| Volume/OI ratio | Peer z-score > 1.5 or absolute threshold by instrument type | Finds turnover that is large versus open positions. | Vol/OI column with z-score sublabel. |
| OI change percent | Peer z-score > 1.5 or absolute OI change > 20% | Separates new positioning from pure price momentum. | Price/OI matrix tag. |
| Price expansion | Price z-score > 1.5, stock move > 2%, index move > 1% | Highlights contracts breaking normal intraday behavior. | Momentum chip and range badge. |
| Range and ATR expansion | High-low range exceeds normal time-of-day range | Detects breakouts, failed auctions, and volatility ignition. | Range stretch alert type. |
| Volume share | Stock futures > 5% share, index futures > 15% share | Shows where market attention is concentrating. | Share grab badge. |
| Sector breadth | Multiple sector names fire in adjacent buckets | Distinguishes isolated news from broad rotation. | Sector pressure leaderboard. |
| VWAP and opening range state | Alert fires near ORB, VWAP reclaim, or VWAP rejection | Turns anomaly rank into an execution map. | Chart overlay and setup label. |
| Catalyst proximity | Alert within 5 to 15 minutes of verified news or event | Separates headline repricing from quiet accumulation. | News marker on timeline. |
Confidence Model
Use score to prioritize human attention, not to automate entries. A strong model blends anomaly strength, signal agreement, participation concentration, and decay control.
Largest weight because participation shock is the cleanest early tell.
Confirms whether the move is new positioning, covering, or unwinding.
Raises rank when participation is already changing price behavior.
Protects against peer noise by asking if this symbol is abnormal for itself.
Start with weighted z-scores, add bonuses for multiple trigger flags, volume share, and rapid intensity, then cap the score at 100. Penalize stale alerts, poor liquidity, and patterns that historically fade.
Alert Logic
Alerts should explain the evidence behind the score and the next decision. A trader needs to know whether to watch, validate, prioritize, or escalate immediately.
One strong anomaly, or early activity without enough price/OI confirmation.
Add to monitor list, wait for a second trigger, and check market regime.
At least two trigger flags or one strong anomaly plus rising participation.
Open the chart, validate VWAP/opening-range structure, and define invalidation.
Volume/OI, OI shift, price expansion, and sector or notional confirmation align.
Prioritize for execution review; avoid chasing if the move is extended from VWAP.
Multiple anomaly families fire together, usually with broad sector or squeeze behavior.
Escalate immediately, but apply liquidity, spread, and risk-size checks first.
UI and UX
The interface should reduce scanning time, expose why a signal fired, and make historical context reachable without forcing traders into separate research tools.
Recommended display
Show broad long build, broad short pressure, short-covering squeeze, two-way rotation, or selective rotation before any symbol table.
Recommended display
Use hourly or 15-minute tiles showing unusual percentage, leader count, dominant setup, and first leader. Traders should see when the move started.
Recommended display
Separate First Detected from Conviction Leaders. Early names are not always strongest; strongest names are not always early enough to trade.
Recommended display
Keep the table dense: symbol, signal, confidence, trend, Vol/OI, price, OI change, participation, and short commentary.
Recommended display
Let analysts replay older sessions bucket by bucket, compare similar historical anomalies, and label outcomes after market close.
Recommended display
Alert cards should state why the score fired, what changed since the prior bucket, and whether breadth is isolated or spreading.
Market-Hours Workflow
A robust workflow changes by time of day. Opening anomalies, midday rotations, and closing squeezes require different confirmation rules and different risk controls.
Best for catching initiative long buildup or short buildup before the broader market notices.
Helps distinguish real accumulation/distribution from an opening trap.
Useful when the index is quiet but institutions are rotating inside a sector.
Protects traders from chasing end-of-day extensions while still surfacing forced-flow setups.
Every alert should be treated as a focus signal, not an entry command. The trader still validates spread, liquidity, VWAP/structure, catalyst quality, and a clear invalidation level before acting.
FAQ
Use these rules as a research and prioritization framework. They are designed to surface where attention belongs, not to replace execution judgment.
It means volume, open interest, price behavior, volatility, or participation concentration is materially different from normal behavior for that symbol, peer universe, and time of day.
Measure volume versus open interest, open-interest change, price expansion, self-history z-scores, peer z-scores, sector breadth, and notional concentration. Higher confidence requires multiple signal families to agree.
Historical data creates time-of-day baselines, symbol-specific norms, peer distributions, and labeled anomaly outcomes. Current intraday behavior can then be compared with similar older patterns to estimate follow-through quality.
A practical design uses watch alerts around 45 to 59, actionable alerts above 60, high-conviction alerts above 75, and critical escalation above 85 after liquidity and spread checks.
Not always. It can support bullish continuation, but the outcome depends on broader regime, liquidity, catalyst quality, volatility, and follow-through after the first pullback.
Yes. Live mode is intended for market-hours monitoring, while snapshot mode helps analysts replay prior sessions and build pattern memory before applying the model in real time.
Unusual Futures Activity highlights contracts whose price, volume or OI differs from their recent reference behaviour. Component readings help distinguish an unusually large price move from an unusual increase in trading or outstanding positions.
Read the baseline and absolute values behind the score, then compare expiry, sector and recent sessions. A normally quiet contract can produce a large relative deviation from a small change. Unusual activity does not identify participant intent or guarantee follow-through; rollovers and event-driven trading also change the readings.
It means volume, open interest, price behavior, volatility, or participation concentration is materially different from normal behavior for that symbol, peer universe, and time of day.
Measure volume versus open interest, open-interest change, price expansion, self-history z-scores, peer z-scores, sector breadth, and notional concentration. Higher confidence requires multiple signal families to agree.
Historical data creates time-of-day baselines, symbol-specific norms, peer distributions, and labeled anomaly outcomes. Current intraday behavior can then be compared with similar older patterns to estimate follow-through quality.
A practical design uses watch alerts around 45 to 59, actionable alerts above 60, high-conviction alerts above 75, and critical escalation above 85 after liquidity and spread checks.
Not always. It can support bullish continuation, but the outcome depends on broader regime, liquidity, catalyst quality, volatility, and follow-through after the first pullback.
Yes. Live mode is intended for market-hours monitoring, while snapshot mode helps analysts replay prior sessions and build pattern memory before applying the model in real time.
