Order Flow & Tape Reading

Question 1 (12 marks)

(a) (3 marks)

• Bar delta $= \sum(\text{buy-aggressor volume}) - \sum(\text{sell-aggressor volume})$ within the bar. (2)
• Cumulative delta = running total of bar/trade deltas across the session; it never resets intrabar. (1)

(b) (5 marks)

def compute_cumulative_delta(trades):
    cum = 0
    out = []
    for t in trades:
        signed = t['size'] if t['aggressor'] == 'buy' else -t['size']
        cum += signed          # accumulate
        out.append(cum)
    return out

Marks: signed volume via aggressor (2), running accumulator (2), returns per-trade list (1).

(c) (4 marks) Signed volumes: $+40, -25, +60, -30, -15$. Bar delta $= 40 - 25 + 60 - 30 - 15 = 30$. (2) Cumulative path: $40, 15, 75, 45, 30$. Final cumulative delta $= 30$. (2)

Question 2 (14 marks)

(a) (2 marks) POC = price level with the greatest traded volume. Here max volume = 700 at 101.1 → POC = 101.1. (1 def + 1 identify)

(b) (8 marks) Total volume $= 120+300+500+700+400+250+130 = 2400$. Target $= 0.70 \times 2400 = 1680$. (1)

Algorithm: start at POC; repeatedly compare the sum of the two rows above the current top vs the two rows below the current bottom; add whichever pair is larger; continue until accumulated volume $\ge 1680$. (2)

Apply (start band = {101.1} = 700):

• Above pair (101.2, 101.3) = $500+300 = 800$; Below pair (101.0, 100.9) = $400+250 = 650$. Above larger → add. Total $=700+800=1500$. Band now 101.0? no — band = [101.1..101.3]. (1)
• Next above pair (101.4, —) = $120$ (only one row left, +0) = 120; Below pair (101.0, 100.9) = 650. Below larger → add. Total $=1500+650=2150 \ge 1680$. (2)

Value area now spans 100.9 → 101.3.

• VAL = 100.9, VAH = 101.3. (1.5)
• Volume inside VA $= 300+500+700+400+250 = 2150$. (0.5)

(Accept stopping exactly when threshold crossed; the crossing row 100.9 is included by convention.)

(c) (4 marks)

• Value area = where ~70% of business was done → "fair value" / balance region. (1)
• Price accepted outside VA (holds beyond VAH/VAL) signals a genuine breakout → follow. (1.5)
• Price rejected back inside VA signals a failed auction/false break → fade back toward POC. (1.5)

Question 3 (10 marks)

(a) (6 marks — 3 each)

• Absorption: (i) heavy aggressive volume hitting a level but price barely moves — a large passive limit order soaks it up. (ii) Delta is large (same direction as aggression) yet price flat/diverges from delta. (iii) Reversal against the aggressor as they get exhausted / trapped.
• Exhaustion: (i) climactic volume/large prints at the extreme, then aggression dries up. (ii) Delta spikes then collapses / momentum fades. (iii) Trend stalls and reverses as the last participants are used up.

(b) (4 marks)

• Strong positive delta but price fails to advance = classic delta/price divergence → absorption. (2)
• Justification: aggressive buyers are trading heavily but a passive seller is absorbing all of it at the level; price can't rise, signalling sellers control the level and a reversal down is likely. (2)

Question 4 (12 marks)

(a) (4 marks)

• Spoofing: placing a large order with intent to cancel before execution, to create false impression of supply/demand and move price. (1.5)
• Layering: placing multiple orders at several price levels on one side to build a fake wall / stacked deception. (1.5)
• Illegal because it's manipulative intent to deceive the market (fraud), no genuine intent to trade. (1)

(b) (8 marks)

def flag_spoof(order_events, book_stats):
    flags = []
    for o in order_events:
        big     = o.size > 5 * book_stats.avg_order_size          # unusually large
        far     = abs(o.price - book_stats.touch) <= NEAR_TICKS    # near enough to spook
        life    = o.cancel_time - o.place_time
        canceled_fast = (o.status == 'canceled') and (life < SHORT_MS)
        never_filled  = (o.filled == 0)
        if big and far and canceled_fast and never_filled:
            flags.append(o.id)
    return flags

Marks: size-vs-average (2), distance-from-touch (2), short lifetime + cancel-before-fill (3), returns flagged IDs (1). Accept a repeat-count / layering extension for full credit.

Question 5 (12 marks)

(a) (4 marks)

• Market depth: the resting bid/ask volume at each price level away from the best (top of book). (1.5)
• Imbalance: ratio/difference of aggregated bid vs ask size indicating buy/sell pressure. (1.5)
• Iceberg: displayed size is small but keeps replenishing after fills — repeated refills at one price reveal hidden size. (1)

(b) (4 marks)

• Bid total (top 3) $= 500+800+650 = 1950$. (1)
• Ask total (top 3) $= 120+150+200 = 470$. (1)
• Ratio $= 1950/470 \approx 4.15$. (1)
• Interpretation: bids heavily outweigh asks (~4:1) → bullish depth imbalance, upward pressure — but note thin size could itself be spoofed. (1)

(c) (4 marks)

• Genuine institution: passive order gets filled/refilled (iceberg), tape shows real absorption, order stays through pressure. (2)
• Spoofer: large size but cancels before fill, pulls when price approaches, no prints match the displayed size. Cross-check book display against actual executed tape. (2)

[
  {"claim":"Q1 bar delta = 30 and final cumulative delta = 30","code":"vols=[40,-25,60,-30,-15]\ndelta=sum(vols)\ncum=0\npath=[]\nfor v in vols:\n    cum+=v\n    path.append(cum)\nresult = (delta==30 and path[-1]==30)"},
  {"claim":"Q2 total volume 2400 and 70% target 1680","code":"vols=[120,300,500,700,400,250,130]\ntotal=sum(vols)\nresult = (total==2400 and abs(0.7*total-1680)<1e-9)"},
  {"claim":"Q2 value area volume = 2150 (100.9 to 101.3 inclusive)","code":"va=[300,500,700,400,250]\nresult = (sum(va)==2150 and sum(va)>=1680)"},
  {"claim":"Q5 top-3 imbalance ratio approx 4.15","code":"bid=500+800+650\nask=120+150+200\nratio=bid/ask\nresult = (bid==1950 and ask==470 and abs(ratio-4.148936)<1e-4)"}
]