ψ Psi Quant Research

Signal Report · US

SO HOLD

Composite decision report grounded in 1-XH strategy signals and a completed multi-agent debate (agents-3node-v1), generated at 2026-09-29 10:06.

Final Agent Decision

Action
HOLD
Weight
0%
Confidence
35%
Variant
xiaohan_gogo_v2

Reasons

Bull Case (Analyst)

- Technical signal is a clean `buy` (score 0.60, xiaohan_gogo_v2) with a stacked multi-timeframe bottom: `1h_bottom, 2h_bottom, 3h_bottom, 4h_bottom` at close 83.46, just above ema_low_fast (82.73) — a mean-reversion entry setup.
- Cross-sectional lifters support the long case: `xh_gogo_score` z = +1.76 (the dominant positive contributor, corroborating the technical buy), plus `downside_vol_60d` +1.17 and `volume_momentum_60d` +0.67 — favorable volatility and volume dynamics.
- News flow is constructive: Seeking Alpha upgrade to buy at $82.89 citing 17 GW contracted large load and 22.83% YoY growth; a new power agreement with a leading Big Tech name; and Zacks 2026/2027 EPS estimates revised up 6.74% / 7.12% to $4.59 / $4.91.
- Stock is at/near 52-week lows ($81.69 touched intraday) and ~15% below all-time highs, with a 3.67% dividend yield — a valuation/income floor that supports risk-reward on a bounce.

Bear Case (Analyst)

- Composite z is only +0.08 (rank 97/252, best 38.5%) — the cross-section does NOT place SO in the top decile, and it is **not in the production basket**. The technical buy is not corroborated by basket membership.
- Draggers are severe and directly contradict the long case: `sharpe_60d` z = -2.17 (worst-in-class risk-adjusted returns), `rsi_14` z = -1.40 (weak momentum), and `close_to_high_60d` z = -0.97 (price pinned near lows, no trend strength).
- Price action is weak: three consecutive down days into late September, and the stock only bounced after hitting a 52-week low of $81.69 — a falling-knife profile, not a confirmed reversal.
- The bull thesis leans on news/valuation (upgrades, contracted load, yield) while the quant evidence is thin: a 0.60 gogo score with a sub-median composite rank and no basket inclusion means the signal lacks cross-sectional confirmation.
- Note: the technical timestamp (2026-09-29 15:30) and news items are not fully aligned in date (some reference Sep 22), so the freshness of the catalyst set is uncertain — evidence for a durable turn is incomplete.

Devil's Advocate

COUNTER TO BULL:
- "Clean buy" is a single-model score (0.60) with no basket confirmation; a 0.60 is middling, not high-conviction. Stacked multi-timeframe bottoms are a lagging, mean-reversion pattern that fires repeatedly on the way down — it is not evidence of a floor.
- Close 83.46 vs ema_low_fast 82.73 is a 0.9% cushion. That is noise, not support. One down day invalidates the "just above" claim.
- The "cross-sectional lifters" are cherry-picked. `xh_gogo_score` z=+1.76 is the same signal family as the technical buy — it is not independent corroboration, it is double-counting the same factor. `downside_vol_60d` +1.17 could just as easily mean the stock is a high-volatility falling knife, not "favorable."
- News is stale and non-quantitative. An upgrade at $82.89 is already below current price action context; "17 GW contracted" and "Big Tech agreement" are unverifiable from the signal set and typically priced in before retail sees them. EPS revisions are analyst lag, not leading.
- "Near 52-week lows + 15% off highs + 3.67% yield" is a value trap framing. Yield rises because price falls. Nothing here says the decline has stopped — it says it has been persistent.
- Hidden assumption: that mean-reversion works here. With `sharpe_60d` z=-2.17, the regime is trending-down, exactly where mean-reversion entries fail.

COUNTER TO BEAR:
- Composite z=+0.08 / rank 97/252 is *not* a bearish signal — it is neutral. Absence of top-decile rank is not evidence of downside; it is evidence of no edge. The bear case over-reads a non-signal.
- "Not in production basket" is a process fact, not a price prediction. Baskets exclude names for capacity, liquidity, or turnover reasons unrelated to expected return.
- The draggers are backward-looking: `sharpe_60d`, `rsi_14`, `close_to_high_60d` all describe the *past* 60 days. In a mean-reversion setup, weak trailing stats are the precondition, not the refutation. The bear is using the same data the bull is, just inverted.
- "Three down days" and "bounced off 52-week low" is pattern-matching, not analysis. A bounce off a low with a technical buy trigger is literally the bull's setup. Calling it a "falling knife" is an assertion, not a measurement.
- The bear's strongest point (no basket inclusion) is also its weakest: it concedes the quant evidence is *thin*, not *negative*. Thin evidence cuts both ways.
- Date-alignment complaint is symmetric — if the catalyst set is stale, the bear's "three down days into late September" is equally stale. You cannot disqualify the bull's news on freshness while relying on equally old price action.
- Hidden assumption: that sub-median composite rank predicts underperformance. Rank 97/252 is above median-ish territory and within noise of the middle; treating it as a short signal is unjustified.

Technical Signal (1-XH)

action
buy
score
0.60
variant
xiaohan_gogo_v2
timestamp
2026-09-29T15:30:00-04:00
close
83.46
ema_low_fast
82.73
ema_high_slow
84.94
gogo_score
0.6
gogo_detail
1h_bottom,2h_bottom,3h_bottom,4h_bottom

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

Trades
2
Win rate
0%
Avg / trade
-3.2%
Total (compounded)
-6%
best / worst
-0.3% / -6.2%
avg holding
22.6 days
open position
yes — marked-to-market -6.2%
Entry Exit Entry Exit Return Exit reason
2026-08-28 open 88.95 83.46 -6.2% open_position_mark_to_market
2026-05-15 2026-05-28 93.24 92.97 -0.3% adaptive_breakdown_30m

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 (7 items)