ψ Psi Quant Research

Signal Report · US

AMD BUY

Composite decision report grounded in 1-XH strategy signals and a completed multi-agent debate (agents-3node-v1), generated at 2026-09-02 09:53.

Final Agent Decision

Action
BUY
Weight
30%
Confidence
55%
Variant
xiaohan_gogo_v2

Reasons

Bull Case (Analyst)

- Technical signal is a clean buy (score 0.60) with a 1h/3h/4h bottom formation and daily blue fully above yellow — the gogo_detail flags a multi-timeframe bottoming structure that often precedes mean-reversion rallies.
- The cross-sectional factor view is constructive: composite z = +0.38, rank 31/252 (top 12.3% of the market) in a BULL regime catalog with zero direction-flips — the regime itself is supportive of longs.
- The dominant lifters are powerful: xh_gogo_score (+2.48) is the single strongest factor in the entire catalog, corroborating the technical buy signal; beta_60d (+1.36) and momentum_6m (+1.25) add further long conviction.
- News flow shows operational strength: record Q2 2026 sales with data center growth and Q3 guidance above Wall Street expectations (item 4), plus a live AI infrastructure partnership with Cisco (item 6) — fundamentals are not the problem.
- At $456.30, the stock has already de-risked: down 4.08% over 7 days and ~12.7% from its 52-week high, while the technical setup suggests the selling pressure is exhausting at these lower levels.

Bear Case (Analyst)

- The cross-sectional rank (31/252) is good but NOT top decile — AMD was excluded from the production basket, meaning the model's highest-conviction long book does not include this name despite the buy signal.
- The draggers are telling: close_to_high_60d (-0.96) and bollinger_b_20d (-0.81) indicate the stock is far from highs and sitting in weak price territory; volume_momentum_60d (-0.67) shows the recent decline was on heavy participation — distribution, not noise.
- Insider selling is a red flag: CEO Lisa Su sold 15,000 shares at $480.94 and SVP Ava Hahn sold 2,993 shares at $488.69 on August 18 — both sold near recent highs, before the subsequent slide, signaling insiders did not see value at those levels.
- The stock fell >8% after a beat-and-raise quarter (item 7), with Musk's NVIDIA pledge cited as the catalyst — this suggests AMD is losing the AI narrative war to NVDA, a sentiment overhang that technicals alone cannot reverse.
- Price action is broken: opened down 3.14% on Sep 1 (item 5), 7-day return is -4.08%, and the close ($456.30) sits below both ema_low_fast ($459.20) and ema_high_slow ($473.80) — the buy signal is a counter-trend bounce against a persistent downtrend, not a reversal.

Devil's Advocate

COUNTER TO BULL:
- The "clean buy" technical score is contradicted by your own price-action evidence: close below both fast and slow EMAs, and a 7-day decline on rising volume. A 1h/3h/4h bottom is noise unless confirmed on the daily close — you have no daily reversal candle, only a daily "blue above yellow" that may be lagging.
- Composite z = +0.38 is weak sauce. Rank 31/252 sounds good, but the model excluded AMD from the production basket — that is the model telling you this signal lacks conviction where it matters. You are cherry-picking a supportive regime catalog while ignoring the dragger factors that are equally part of the same composite.
- xh_gogo_score (+2.48) being the "strongest factor in the catalog" is a red flag, not a blessing: extreme single-factor dominance usually means the signal is crowded and prone to mean-reversion against you. Beta and momentum are generic risk premia, not catalysts.
- Fundamentals "not the problem" is irrelevant — the stock fell >8% on a beat-and-raise. That means the market is pricing narrative and positioning, not EPS. Cisco partnership and record sales were already known; they are not new marginal buyers.
- "De-risked" is an assumption, not a fact. Down 12.7% from highs in a name that ran parabolic is still expensive on a P/E basis. Selling pressure exhausting is a hope — volume_momentum_60d (-0.67) says the opposite: heavy distribution is ongoing, and a 3-4% bounce off lows is exactly what distribution looks like before the next leg down.

COUNTER TO BEAR:
- Rank 31/252 is top 12.3% — excluding from the production basket may reflect risk limits (e.g., sector concentration, liquidity) rather than low conviction. Using basket exclusion as a bear argument is a logical leap without knowing the exclusion rule.
- Insider selling of ~18k shares total is trivial against daily volume and CEO's remaining holdings. Insiders sell for diversification, taxes, or planned 10b5-1 programs — you have no context on whether these were scheduled sales. Calling it a "red flag" without that context is weak.
- The post-earnings drop on Musk/NVIDIA news is a sentiment snapshot, not a structural verdict. AMD has repeatedly been "losing the AI narrative" for years and still rallied — narrative overhangs are exactly the kind of sentiment that bottoms form against. Your own data shows a multi-timeframe bottoming structure forming.
- "Broken price action" is backward-looking. Close below EMAs is a description of the past, not a prediction. Counter-trend bounces are how every reversal starts — you need to show why this bounce fails, not just that the trend was down. The 7-day -4.08% decline could be the exhaustion move, not the start of a new leg.
- You ignore the strongest factor in the entire catalog (xh_gogo_score +2.48) and the constructive regime. Dismissing the bull's best evidence while leaning on weak insider-sale data and narrative arguments makes the bear case feel like a collection of grievances, not a quantitative thesis.

Technical Signal (1-XH)

action
buy
score
0.60
variant
xiaohan_gogo_v2
timestamp
2026-09-02T15:30:00-04:00
close
456.3
ema_low_fast
459.2
ema_high_slow
473.8
gogo_score
0.6
gogo_detail
1h_bottom,3h_bottom,4h_bottom,daily_blue_fully_above_yellow

📈 Backtest Track Record xiaohan_gogo_v2 · signals simulated over available history (~3-month 30-min window, accumulating daily)

Trades
1
Win rate
0%
Avg / trade
-4.7%
Total (compounded)
-5%
best / worst
-4.7% / -4.7%
avg holding
8.0 days
open position
yes — marked-to-market -4.7%
Entry Exit Entry Exit Return Exit reason
2026-08-25 open 478.87 456.30 -4.7% open_position_mark_to_market

复盘 = each past xiaohan_gogo_v2 buy signal simulated: enter next bar, manage adds / adaptive sells / −25% hard stop. Demonstration only, not investment advice.

News & social (8 items)