ψ Psi Quant Research

Signal Report · US

DUK HOLD

Composite decision report grounded in 1-XH strategy signals and a completed multi-agent debate (agents-3node-v1), generated at 2026-09-25 10:07.

Final Agent Decision

Action
HOLD
Weight
0%
Confidence
35%
Variant
xiaohan_gogo_v2

Reasons

Bull Case (Analyst)

- Technical signal is a fresh `buy` (score 0.60, xiaohan_gogo_v2) with `gogo_detail` showing synchronized bottoms across 1h/2h/3h/4h timeframes — a multi-timeframe exhaustion-of-sellers setup, with close 113.3 sitting below both ema_low_fast (113.9) and ema_high_slow (117.1), i.e., a mean-reversion entry rather than a chase.
- The single strongest cross-sectional lifter is `xh_gogo_score` at +1.61 z — the same factor family that generated the technical buy — so the factor rank and the timing signal are internally consistent on the bullish side.
- `downside_vol_60d` (+1.02) and `volatility_60d` (+0.14) are positive-z lifters, meaning DUK's realized volatility profile is being rewarded in this BULL catalog; in a 34-factor regime with zero direction-flips, that's an unambiguous tailwind rather than a mixed signal.
- News flow is supportive-to-neutral: Morningstar names DUK among "best utilities stocks to buy," a TradingView comparison gives DUK the edge over NextEra (Zacks #3 Hold vs #4 Sell), and Fox News notes data-center expansion in North Carolina with DUK among holdings — a structural demand narrative for a regulated Carolina utility.

Bear Case (Analyst)

- Composite z is only -0.05, rank 139/252 (55.2nd percentile) — dead center of the cross-section, and explicitly **not in the production basket**. The model's own cross-sectional layer is not endorsing this name; the buy is a single-strategy timing call against a neutral factor backdrop.
- The draggers are momentum/trend factors, not noise: `rsi_14` at -1.50, `sharpe_60d` at -1.42, `bollinger_b_20d` at -1.41. Weak RSI, poor risk-adjusted 60-day returns, and a low Bollinger %B all say the tape is in the lower part of its range with deteriorating reward-per-unit-risk — the opposite of what you want underwriting a long.
- The bullish lifters are largely the same signal restated: `xh_gogo_score` (+1.61) is the technical score itself, and the volatility lifters (`downside_vol_60d` +1.02) reward *downside* variance — that is a compensation-for-risk factor, not evidence of upside. Strip those out and the residual factor picture is negative.
- Price at 113.3 is below both EMAs (113.9 fast, 117.1 slow), so the trend structure is still down; a "bottom" call across four hourly timeframes is a counter-trend entry with no confirmation from the daily/slow EMA, and the 117.1 slow EMA is ~3.4% of overhead resistance.
- News is thin and mixed rather than catalytic: Zacks Rank #3 (Hold) on DUK, a GE Vernova comparison that explicitly prefers GEV on earnings growth, ROE, and price performance, and the only substantive DUK-specific piece is a paywalled equity-units note — potential dilution/structural complexity rather than a fundamental upgrade. No evidence in the bundle supports an earnings or guidance catalyst.

Devil's Advocate

COUNTER TO BULL:
- "Synchronized bottoms across 1h/2h/3h/4h" are all sub-daily and highly autocorrelated — four timeframes of the same intraday tape is one observation, not four confirmations. The claim of "multi-timeframe" breadth is inflated.
- Mean-reversion entry below both EMAs is indistinguishable from "catching a falling knife." No daily/weekly confirmation, no volume, no reversal candle cited. "Exhaustion of sellers" is asserted, not evidenced.
- Internal consistency between `xh_gogo_score` and the buy signal is circular: the factor and the signal share a family. Agreement between a signal and its own parent factor is not independent corroboration.
- Positive-z on `downside_vol_60d` and `volatility_60d` being "rewarded in this catalog" is a description of the model's fitting, not a market fact. High vol being a lifter can equally mean the catalog is short-vol-biased or that the factor is compensating for risk you're about to take.
- "Zero direction-flips in a 34-factor regime" is presented as a tailwind but is equally consistent with a stale/overfit factor set. Stability is not the same as edge.
- News items are generic screeners and comparisons, not DUK-specific catalysts. "Best utilities to buy" lists and a Fox segment mentioning holdings are ambient, not actionable. The NextEra comparison is relative and says nothing about DUK's absolute setup.
- Score 0.60 is a modest threshold; no base rate, hit rate, or sample size given. A single-strategy buy at 0.60 with no cross-sectional backing is a low-conviction trigger dressed as a signal.

COUNTER TO BEAR:
- Composite z of -0.05 / rank 139 is statistically indistinguishable from zero. Calling it "not endorsing" overstates a number that is essentially noise; "not in the production basket" may reflect basket construction rules, not a negative view.
- The momentum draggers (rsi_14, sharpe_60d, bollinger_b_20d) are exactly the factors a mean-reversion thesis expects to be negative. Citing them as bearish assumes trend-following is the correct prior — which is the very thing the bull is betting against. This is question-begging.
- "Strip out the bullish lifters and the picture is negative" is true of any factor set. Removing the factors that disagree with you is not analysis.
- The claim that `downside_vol_60d` "rewards downside variance, not upside" is an interpretation. In a long-only catalog, a positive-z on downside vol can reflect a risk premium being paid to holders — which is a reason to own, not avoid.
- "Below both EMAs = trend still down" uses the same EMAs the bull already flagged as the mean-reversion target. Both sides are reading the same levels; the bear's "3.4% overhead resistance" is the bull's "mean-reversion upside."
- The bear dismisses the news as "thin and mixed" but then cites a GE Vernova comparison and a paywalled note as if they were evidence. Neither is DUK-specific negative information; a paywall is not dilution.
- "No earnings or guidance catalyst in the bundle" cuts both ways — absence of a negative catalyst is not a bearish argument, and the bear offers no downside trigger, level, or event either.

Technical Signal (1-XH)

action
buy
score
0.60
variant
xiaohan_gogo_v2
timestamp
2026-09-25T15:30:00-04:00
close
113.3
ema_low_fast
113.9
ema_high_slow
117.1
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
-3.8%
Total (compounded)
-7%
best / worst
-0.7% / -6.9%
avg holding
47.5 days
open position
yes — marked-to-market -6.9%
Entry Exit Entry Exit Return Exit reason
2026-08-25 open 121.64 113.30 -6.9% open_position_mark_to_market
2026-05-13 2026-07-16 125.54 124.70 -0.7% daily_sell_after_daily_blue_cross

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)