When: Price below $215.97 (10 EMA)

Then: Do NOT add; respect technical breakdown and Q3 binary risk

HOLD - maintain current positioning, respect $209.87 stop on existing positions, do not add into current technical structure, wait for Q3 earnings to resolve credit overhang

• 50 SMA has risen from $192.85 to $209.87 but is being tested from above in a confirmed lower-low sequence since August 14, not a completed base pattern
• MACD histogram at most negative reading since July base; RSI rolled from 70 to 45 in 13 sessions without printing oversold
• ATR of $4.95 means $211 to $205.92 is roughly two average daily ranges, making the tight stop vulnerable to gap through on headline events
• $11 billion in annual buybacks against $129.65 billion market cap represents 8.5% of float annually but is funded by same $26 billion FCF that requires credit costs to stay controlled; buybacks get pulled if charge-offs accelerate
• 30-day delinquency bucket is leading indicator with 1-2 quarter lead time; 90-day delinquencies have nearly doubled since 2022 while 30-day is flat, but Q3 print will be first where 2022 vintages roll into peak charge-off territory at scale
• Berkshire Hathaway Q2 13F trim represents changed marginal buyer calculus; Buffett was loudest external validator of post-Discover thesis and his partial trim signals narrowed asymmetric upside
• Probability-weighted risk-reward: ~25-30% probability COF closes above $240 in 6 months versus ~35-40% probability it trades below $205.92, yielding expected value closer to coin flip than 5.3-to-1 arithmetic
• Fed 'higher for longer' creates consumer credit stress with lag; same policy tailwind for NIM is simultaneously creating headwind most likely to break bull case
• Stablecoin consortium: 21 banks, no disclosed economics, earliest revenue impact early 2027—16+ month optionality not consistent with 6-month trade thesis
• Trump AML litigation headline risk is unpriced; discovery process can produce weeks of negative news flow that moves stock 5-10% independent of credit fundamentals
• Asymmetry: being wrong on patience costs 10-15% missed rally that gets faded anyway; being wrong on adding aggressively ahead of earnings costs 15-25% drawdown that is hard to recover

→ vs aggressive: Framing HOLD as 'paralysis' is rhetorical trick; discipline is the point, and patience is appropriate three weeks before Q3 print where 90-day delinquencies are headline number

→ vs aggressive: Rising 50 SMA interpretation ignores confirmed lower-low sequence, most negative MACD histogram since July, and RSI rolling from 70 to 45 in 13 sessions without printing oversold—setup where first retest often fails

→ vs aggressive: $11 billion buyback math ignores that buyback rate is funded by same $26 billion FCF that depends on credit costs staying controlled; if Q3 shows charge-off acceleration, buybacks get pulled and EPS growth goes from 15-20% to single digits

→ vs aggressive: 'New delinquencies have not moved' doesn't foreclose bear case; 30-day bucket is 1-2 quarter leading indicator and Q3 is first print where 2022 vintages hit peak charge-off territory at scale

→ vs aggressive: Berkshire trim framing as 'harvesting gains' ignores that Q2 13F is market information that changed marginal buyer calculus; Buffett was primary external validator and his trim signals narrowed asymmetric upside

→ vs aggressive: Linear 5.3-to-1 risk-reward presentation ignores probability weighting; real distribution is fat-tailed and negatively skewed by October binary event with expected value closer to coin flip

→ vs aggressive: Fed zero cuts bullish for NIM simultaneously creates consumer credit stress with lag; higher for longer historically produces exactly the credit deterioration most likely to break bull case

→ vs aggressive: Stablecoin 2027 revenue optionality is internally inconsistent with 6-month buy thesis at $211; strategic partnership among 21 peers with no disclosed differentiated role for COF won't re-rate franchise on announcement alone

→ vs aggressive: Trump AML litigation headline risk is entirely ignored; political dynamics around Trump-related lawsuit against consumer bank can pressure stock 5-10% on single news cycle independent of credit fundamentals

HOLD with staged scale-in framework; maintain current position sizing without adding into weakness, wait for technical confirmation at $215.97 or flush to $205.92 with reversal candle for higher-conviction entry, stop zone $204-207

• Buybacks are credit-cycle dependent accelerator, not guaranteed return - $11B annual pace depends on NIM staying elevated and credit costs controlled
• Berkshire trim is a confirming indicator, not a thesis-breaker; should be weighted at 5-25% of decision, not zero
• Probability estimates: 30-35% probability of $235+ over six months, 30% probability of breaking $205
• Third realistic bucket: stock goes sideways $205-$220 for three months, not binary bull/bear
• Rising 50 SMA at $209.87 acting as dynamic support in base-building structure; hold probability 65-70% if tested
• No data yet showing 30-day delinquency buckets rolling despite 90-day doubling; projects narrative onto missing data
• Even if buybacks moderate to $5-6B annually, still 4-5% of float absorbed annually - mechanical EPS support doesn't disappear
• Stablecoin 21-bank consortium is strategic asset for multi-year valuation, not near-term catalyst
• Q3 earnings is tiered decision point: credit deterioration signals Underweight, credit stability with synergy beat signals Overweight, mixed signals mean reduce sizing by a third and wait
• Both the aggressive analyst (buy thesis) and conservative analyst (avoid thesis) are overconfident; asymmetry is more subtle than either acknowledges

→ vs aggressive: Buyback math is logically flawed - treating buybacks as exogenous to credit cycle when they're endogenous; both legs of the 'machine' require benign credit cycle

→ vs aggressive: Risk-reward of 5.3-to-1 is not probability-weighted; implies 70-80% bull case probability when realistic is 30-35%

→ vs aggressive: Technical 'textbook mean-reversion setup' ignores that mean reversion works only 60-70% of time; MACD already bearish-crossovered with most negative histogram since July base

→ vs conservative: 50 SMA at $209.87 is rising for six weeks and acting as dynamic support; 'hold vs break' are not equal probability - hold is 65-70%

→ vs conservative: 35-40% probability of breaking $205 is too high given macro backdrop with Fed funds zero cuts through 2026 at 89% probability supporting NIM

→ vs conservative: 25-30% probability of $240 seems too low given Q3 earnings upside has more legs; closer to 35%

→ vs conservative: Buyback point treats them as binary when they're continuous; even halved to $5-6B annually is still 4-5% float absorption

→ vs conservative: Stablecoin dismissal ignores strategic positioning in 21-bank consortium affecting multi-year franchise valuation