Market Context

market-context

Purpose

Compute session VWAP, POC / Value Area, ATR, realized-vol, price-vs-VWAP z-score, and cross-symbol relative strength from market_history. Produce opinionated narrative over the numbers — not raw JSON dumps.

Descriptive, not predictive. Every output describes the fetched bar window: session high, session low, and where volume clustered. It also reports the session's volatility level. Outputs do not forecast future prices or recommend trades. A user may ask about the future. An example is "will it break out?" For that kind of question, state the descriptive limit clearly in the narration. The skill can describe the conditions before past breakouts. It cannot predict the next one.

Environment routing

Symbol/market data only — no account binding. A sibling skill's account resolution might already pin the session to demo (simulation) or live. If so, stay on that same MCP server.

MCP tools used

Tool names below are bare. The NinjaTrader MCP server provides them. Your client adds its own prefix. See AGENTS.md at the repo root.

• market_history — the bar data for the symbol. Set symbol, barType, barSize, and volumeProfile=true for histogram-based VWAP or market profile.
• market_snapshot — fallback source for tickSize when market_history doesn't carry it.
• search_contracts — alternate source for tickSize.

Workflow

1. Pull bars

Call market_history with:

• symbol — the contract symbol (e.g., ESU6)
• barType — Minute for intraday, Daily for daily
• barSize — e.g., 5 for 5-minute bars
• volumeProfile: true for histogram-based VWAP or market profile

For a typical session-context read, 5-minute bars over the RTH session are a reasonable default. For volatility-regime reads, use Daily bars across a 20–60 day lookback.

2. Get tick size (only for histogram mode)

histogram[].price is a tick offset from bar.open, not an absolute price. Actual price = bar.open + histogram[i].price × tickSize.

tickSize is not in the market_history response. Fetch it from one of these:

• market_snapshot with the same symbol
• search_contracts

For full detail and the common-products cheat sheet, see references/histogram-offset.md.

3. Pipe bars to the right script

Every script reads the market_history response as JSON and prints JSON to stdout. Save the tool result to a file and pass its path with --file. Never re-type or inline a large JSON payload into the command. You can still pipe input on stdin as a fallback.

--tick-size is contract-specific — source it from Step 2 (market_snapshot.tickSize or search_contracts.tickSize). Common values: ES/MES/NQ/MNQ 0.25, YM/MYM 1.00, ZB 0.03125, ZN 0.015625, CL/MCL 0.01, GC 0.10, SI 0.005. A hardcoded 0.25 works for ES/NQ. It silently mis-prices YM (4× too small). It also mis-prices ZB (8× too large) and CL (25× too large). Always look it up. Never paste the example value.

VWAP (chooses histogram mode if volumeProfile=true, else typical-price):

python3 scripts/vwap.py --file market_history.json --tick-size <TICK_SIZE>

Output: {vwap, mode, bars_used, total_volume, price_z_score}.

Market profile (requires volumeProfile=true):

python3 scripts/profile.py --file market_history.json --tick-size <TICK_SIZE>

Output: {poc, vah, val, value_area_pct, total_volume, levels_in_value_area, total_levels}.

The optional --value-area 0.70 is the default and matches CME convention. Lower values (e.g., 0.50) give the tighter "developing value area."

Bar delta + cumulative delta + order-flow imbalance (no histogram or tick size required — reads upVolume / downVolume from the bar):

python3 scripts/delta.py --file market_history.json

Output: {bars_used, total_up_volume, total_down_volume, net_delta, cumulative_delta_path, final_imbalance, bar_imbalance_path}. The two *_path arrays are per-bar views. Feed them to a chart-render or a pattern detector.

ATR + realized volatility + close z-score:

python3 scripts/atr.py --file market_history.json --n 14

For intraday bars, pass --periods-per-year to match the bar count per trading year, so the annualized-vol output stays meaningful. Examples:

• Daily bars: --periods-per-year 252 (default)
• 5-minute bars during RTH: 78 bars/day × 252 = --periods-per-year 19656
• 1-minute bars during RTH: 390 × 252 = --periods-per-year 98280

Cross-symbol relative strength (needs at least two symbol files):

python3 scripts/cross_symbol.py --symbol ES=es.json --symbol NQ=nq.json --symbol RTY=rty.json

Each NAME=PATH points to a saved market_history response. Output includes per-symbol change-today and pairwise spreads. This is intra-day only. Historical rolling correlation lives in correlation-hedge.

4. Narrate

Use references/regime-narrative.md to translate the numeric output into short, opinionated sentences. Rules of thumb:

• Lead with dollar / point values. Put R-multiples second, when there's a position context.
• Snap VWAP / POC to tick precision. Never report "7151.4328".
• State regime as fact, not forecast. Say "trading above VWAP"; not "headed higher".
• Never invent a baseline comparison. If no 20-day median exists, report the raw value and say so.

Typical queries and their responses

• "What's ES doing?" → 5-minute bars with volumeProfile; run vwap.py + profile.py; narrate ("ES at 7151, inside value — VAL 7148, POC 7150, VAH 7154 — volume focused around open").
• "Is ES volatile today?" → Daily bars, 20-day lookback; atr.py with --n 14; narrate realized_vol_annualized_pct against the atr_n comparison.
• "What's the delta on ES this morning?" → Use 5-minute bars for the morning window. Run delta.py. Narrate net_delta and final_imbalance. Flag any bar in bar_imbalance_path with |imbalance| > 0.4 as aggressive one-sided activity.
• "How is ES doing vs NQ and RTY?" → three market_history calls, saved to files, then run cross_symbol.py; narrate the largest spread_pct.
• "Where's VWAP?" → just vwap.py; report the number and the z-score narrative.

Disambiguation

• vs correlation-hedge: this skill covers intra-day relative strength (who leads today). Historical rolling correlation across a lookback window lives in correlation-hedge.
• vs chart-render: this skill produces numbers and narrative. Chart-render draws them as a PNG.
• vs trade-journal / trade-replay: those skills look at the user's own trades. market-context is market-wide — no portfolio or fill data involved.

Resource layout

• scripts/vwap.py — histogram-based VWAP with typical-price fallback
• scripts/profile.py — POC / VAH / VAL (default 70% volume area)
• scripts/delta.py — bar delta + cumulative delta + order-flow imbalance
• scripts/atr.py — ATR(N), realized-vol (annualized), close z-score
• scripts/cross_symbol.py — intra-day relative strength across N symbols
• scripts/fixtures/ — a small synthetic market_history response for regression testing
• references/histogram-offset.md — the tick-offset convention. Load it when the user asks about histogram data or VWAP math.
• references/regime-narrative.md — output templates. Load it when you write narrative sentences.