Weight: 0.58% of selectable set | Window: Mar 30 – Jul 10, 2026 | Verdict: FILTER

Factor Profile

The anti-momentum loading is the decisive signal. In a momentum-heavy index, β_MTUM = -0.56 is structural drag. MAR (-0.57) and CSX (-0.46) were removed for the same factor profile. Framework consistency demands the same treatment.

Market Consensus: Three Layers

1. Analyst consensus (what they SAY):

Consensus: NEUTRAL (30% bullish, 11% bearish). Median target $32.50 (+14%). Range $23–$44.

Most recent analyst action: Jan 30, 2026 — nearly two months stale. Post-Versant spin, post-Q4 print, but targets may not reflect the March 16 pro forma recast or recent price deterioration. The $23 Barclays low target is essentially current price — a polite Sell.

59% Hold on a $104B name at 7.3x forward is the tell. At that valuation, the Street should be pounding the table. They're not. The broadband trajectory is why.

2. Options consensus (what they BET):

• Put/call ratio (OI): 1.29 — bearish skew
• ATM IV: 27.1% (55th percentile of 52-week range 13%–38%)
• Max pain: $30.00 (4.9% above current) — market-makers positioned for modest drift up by April opex
• 1 unusual put vs 0 unusual calls — directional flow bearish

3. Price consensus (what money DOES):

• RSI 20.8 — deeply oversold. Buyers have stepped away.
• 1-month: -7.1%. 1-year: -15.0%.
• Below 50DMA ($30.22) and 200DMA ($29.66) — technically broken
• Volume 0.4x 3-month average — selling on low volume = buyers absent, not sellers exhausted
• 52-week range: $24.13–$35.35. Currently at 40% of range.

What's Mispriced

Not the stock. The market efficiently prices CMCSA's broadband erosion, Versant complexity, and cable-to-streaming transition. 7.3x forward, 4.6% yield, RSI 21 — this is a stock the market correctly marks as a low-growth yield play with structural headwinds.

The index. QQQ holds CMCSA at benchmark weight mechanically. The index doesn't adjust for factor profiles. But CMCSA's factor profile is anti-correlated with QQQ's dominant factors:

• QQQ rewards momentum. CMCSA loads -0.56 on momentum.
• QQQ has beta ≈1.1. CMCSA has beta 0.40.
• QQQ is growth-tilted. CMCSA is a 4.6% yield value stock at 7.3x forward.

The mispricing is structural, not informational. We don't know something the market doesn't. We're exercising discretion the index can't — removing a factor-mismatched name that QQQ is forced to hold. At 0.58% weight, the contribution is ≈3–5 bps of filtration alpha. Small, correct, boring.

Fundamental Case: Connectivity in Accelerating Decline

Residential C&P EBITDA trajectory (YoY, pro forma constant currency, 8-K March 16 2026):

Q4 as-reported (non-constant-currency): residential -5.9%, total -4.2%.

The acceleration is the point. Not the -6.0% level, but the slope from +0.9% to -6.0% in four quarters, with management confirming the trajectory continues: "expect incremental EBITDA pressure over next couple quarters" as investments in simplified pricing and free wireless haven't lapped (Q4 2025 transcript).

Q4 2025 Residential C&P detail (8-K Jan 29 2026):

• Revenue: $17.6B (-2.1% YoY). Domestic broadband $6,316M (-1.1%) — first negative broadband revenue quarter.
• EBITDA: $6,099M (-5.9%). Margin 34.6% vs 36.0% prior year (-140 bps).
• Programming costs fell 4.3%, but non-programming opex rose 2.5% (customer experience investments, wireless line costs).

Broadband subscriber losses (8-K Jan 29 2026, customer metrics table):

• Total domestic broadband: 31,255K vs 31,842K = -587K FY2025
  • Residential: 28,719K vs 29,373K = -654K
  • Business: 2,536K vs 2,469K = +67K
• Q4 net losses: -181K total (-178K residential, -3K business)
• Q4 2024 net losses were -139K — losses accelerating quarter over quarter

ARPU under pressure: Management guided "further ARPU pressure next couple quarters" from three sources: no rate increase in H1 2026, free wireless line costs, ongoing migration to simplified pricing. All three are deliberate investments in retention — the sacrifice is intentional, not accidental.

Competitive pressure is structural:

• FWA: Big 3 capacity of 32M subscribers. T-Mobile at 7.3M (targeting 12M by 2028), Verizon 5.1M, AT&T 1.8M (most room to grow). 70% of new FWA customers are urban — direct cable overlap. Pricing ≈50% of cable.
• Fiber overbuild: AT&T expanding to 30M+ passings. CFO Armstrong acknowledged fiber competition "picked up" in Q4 and "remains" into January 2026.
• New Street Research (Dec 2025): cable broadband net adds "not expected to turn positive before 2030."
• Upper C-band auction by July 2027 could add 4M more FWA capacity.

Peacock: Improving but Dilutive in Our Window

Q4 2025 (8-K Jan 29 2026):

• Revenue: $1.6B (+20% YoY), 44M paid subs (+8M YoY)
• EBITDA loss: $552M vs $372M prior year = losses widened $180M
• Full year losses improved $700M+ YoY, but Q4 worsened on NBA rights launch

NBA game volume peaks in Q1 (≈50% of regular season in Oct-Mar). Peak Peacock EBITDA dilution falls in Q1 2026 — directly in our window. Path to breakeven is 2027-2028, not 2026.

Bull Case: Theme Parks and Valuation Floor

Theme parks are genuinely strong (8-K Jan 29 2026):

• Q4 EBITDA: $1,035M (+23.5%) vs $838M — first $1B+ quarter
• Revenue: $2,893M (+21.9%)
• Epic Universe driving higher per-cap spending, longer stays, hotel ADR +20%, occupancy +3%
• Not yet at full run-rate capacity — throughput optimization ongoing
• Universal Kids Resort (Frisco TX) opening later in 2026

But parks are ≈15% of total EBITDA ($3.08B of ≈$37.4B consolidated adj EBITDA). Even +30% parks growth = +2.5% total EBITDA lift, which doesn't offset connectivity decline on the ≈60% that's residential C&P.

Valuation:

• GAAP trailing P/E: 5.31 — misleading. Includes one-time $9.4B pre-tax Hulu gain (sold June 2025). GAAP EPS ≈$5.39 vs adjusted EPS $4.31.
• Adjusted trailing P/E: 6.64 ($28.59 / $4.31)
• Forward P/E: 7.34
• Dividend yield: 4.60% ($1.32/share, 18th consecutive year of growth)
• FY2025 buybacks: $6.8B (≈5.8% of current market cap)
• Net leverage: 2.3x. FCF $19.2B (but ≈$2B was non-recurring cash tax benefit)

Oversold (RSI 20.8): Mean-reversion signal. But at β_SPY = 0.40, even a bounce captures only 40% of a QQQ rally. The asymmetry works against the basket: if QQQ rallies 5%, CMCSA rallies ≈2% — we're better off without it.

Insider Activity

• Feb 27 cluster (≈$8.6M across 6 insiders): CEO Roberts $1.3M, CEO Cavanagh $1.0M, CFO Armstrong $3.0M, plus 3 officers. All "Acquire" on the same date = compensation grants (RSU vesting), not open-market conviction. Date clustering is the tell.
• CFO Armstrong (Mar 5): Converted 73,497 shares ($2.1M), immediately sold 4,494 ($143K). Sold 6% of conversion, held 94%. Tax optimization, not bearish signal — but not bullish either.
• CEO Cavanagh (Feb 11): Sold $1.9M. This is a real discretionary sale from the incoming co-CEO. Not a panic sell at $32.71 (stock has fallen 12% since), but insiders aren't buying the dip either.

No open-market insider buying during the -15% 1-year slide. That silence is information. When the CEO sells and nobody buys, management isn't signaling that the stock is cheap.

Earnings Catalyst: April 23 (Week 4)

Q1 2026 reports April 23 BMO — Week 4 of our window.

Consensus: $0.78 adj EPS (yfinance, Mar 27). Q1 2025 as-reported was $1.09 but included Versant operations (spun off Jan 2 2026). The YoY comparison is not apples-to-apples. The March 16 8-K provides pro forma recast numbers but exact pro forma Q1 2025 EPS is truncated in the filing.

Management is pre-announcing the bad news. "Incremental EBITDA pressure" guidance means Q1 won't be a positive surprise. Comcast has beat EPS estimates in all four trailing quarters (+1.7% to +10.8%), so a beat on lowered expectations is likely. The question is whether guidance deteriorates further (bearish) or management signals H2 inflection (modest positive that doesn't change the 15-week math).

Evidence Table

Independence structure:

• Signals 1-3: one regression, three readings → one cluster (factor profile)
• Signals 4-6, 17: broadband secular decline → one cluster (fundamental bear)
• Signals 10-12: market positioning → one cluster (consensus confirms bear)
• Signal 8-9: timing (Q1 earnings + Peacock dilution) → semi-independent
• Signals 13-14, 16: bull case → one cluster (parks + valuation + simplification)
• Signal 15: mean-reversion → independent but mechanically weakened by low β

Two independent bear clusters (factor + fundamental) confirmed by consensus. One timing signal (Q1 earnings). One independent bull cluster (parks + valuation) insufficient in magnitude.

Memo-level LR: 0.55 (moderately bearish)

Verdict: FILTER

Where's the trade? QQQ is a forced holder of a factor-mismatched name. We have discretion the index doesn't.

What's mispriced? Not CMCSA's absolute price — the market sees the broadband erosion. The index's implicit assumption that all constituents earn their weight. CMCSA's factor profile (anti-momentum, low beta, value/yield) is wrong for a momentum/growth index over 15 weeks.

Who's on the other side? Passive QQQ holders. They own CMCSA because it's in the index, not because they analyzed the factor profile. Our counterparty is mechanical, not informed — that's the cleanest edge available.

When does it resolve? Continuously. Anti-momentum drag accrues every day QQQ moves on momentum. Q1 earnings (Apr 23) is the discrete catalyst — management-guided EBITDA deterioration in our window.

What could go wrong? Value rotation. If market shifts from momentum to defensive/yield (macro shock, tariff escalation, risk-off), CMCSA outperforms QQQ and the filter costs us. At 0.58% weight, the damage is ≈3-5 bps — survivable.

Precedent: MAR (β_MTUM = -0.57) and CSX (β_MTUM = -0.46) removed for the same factor profile. CMCSA at -0.56 is the strongest case of the three.

Expected filtration alpha: +3–5 bps. Small, correct, boring.

Corrections applied: Broadband losses -587K (not -711K). Adjusted trailing P/E 6.64 (GAAP 5.31 includes Hulu gain). Theme parks +23.5% (not +24%). 52-week low $24.13 (not $25.75). Analyst consensus NEUTRAL (not Buy). Q1 EPS YoY comparison non-comparable (pre/post Versant). All corrections verified against 8-K primary sources.