Daily Forecast 2026-09-17

📋 Morning Board 📏 Range Forecast Γ Gamma Map 🕰 Analog Days 🎯 9:00 Forecast ✅ Scorecard 🌡 Sentiment

📋Morning Board

Symbol	Prev close	Change	VIX / VXN (pct)	GEX ($B/1%)	Expected range
QQQ	$704.72	+0.03%	22.4 (38th pctile)	+1.27B flip 709.5	1.21% (0.74–2.05)
SPY	$754.05	-0.44%	17.7 (57th pctile)	+0.68B flip 759.5	0.88% (0.52–1.56)

QQQ RSI(2)19 neutral

QQQ vs 200-day+6.5%

QQQ GammaDealers long gamma (+$1.27B per 1% move), flip 709.5 sits 0.4% below spot

How to read this · VIX

σ Volatility Indices (VIX / VXN)

VIX = S&P 500 30-day implied volatility (maps to SPY); VXN = the same measure for the Nasdaq 100 (maps to QQQ). Percentile = share of the past year (252 trading days) that closed below the current value.

Index	Current	Percentile	1-year range	As of
VIX · SPY	17.71	57% pctile · Near median	13.5 ─ median 17.2 ─ 31.1	2026-09-16
VXN · QQQ	22.44	38% pctile · Near median	17.2 ─ median 23.3 ─ 33.5	2026-09-16

CBOE official daily closes (VIX from 1990, VXN from 2009). These are closing values, not intraday — the morning brief uses the prior session's close and the evening review the current close, so the series stays consistent with history with no point-in-time bias.

📏Range Forecast

Today's Range Forecast (pre-open)

⚠ 数据陈旧：本预测基于 2026-09-16 的日线（距今约 22 天）。落盘日线可能未更新（检查 PostClose_Rebuild 任务），下面的数字不可用于今日决策。

This section answers one question only: how far price is likely to swing between the open and the close (high minus low). It does not forecast direction — the directional version has been backtested repeatedly and produced no reliable signal.

Symbol	Tier	Forecast day range	80% interval	vs last 20d
SPY	Normal	0.88%	0.52% ~ 1.56%	Well above normal
A normal day with no particular signal. Run your usual parameters; no volatility-driven adjustment needed.
QQQ	Normal	1.21%	0.74% ~ 2.05%	Slightly above normal
A normal day with no particular signal. Run your usual parameters; no volatility-driven adjustment needed.

How to read this table
· Forecast day range: how far today's high is from today's low, as a percentage of the current price.For example 0.88% above means the day's high and low are expected to differ by about 0.88%.
· 80% interval:on 8 days out of 10 the actual value lands between 0.52% and 1.56%; on the other 2 it falls outside. That is the honest error band — the real value may be half the forecast, or half again as large.
· Tier: days are bucketed into large / medium / small. When it says “High”, about 6 times out of 10 the day really is large, versus barely 3 in 10 for a blind guess. Good enough to size positions and choose strike width; not good enough to compute precise P&L.
· “Tier” and “vs last 20d” can look contradictory, and that is normal: the tier compares against the past several years, the right-hand column against the last 20 days. “High tier but about normal” means the recent period has itself been high-volatility — today is not exceptional, but the overall environment is jumpy and position sizing should stay cautious.
· A 10:00 intraday revision email follows, folding in the actual range and volume from the first half hour; the forecast gets sharper then.

Technical: out-of-sample R² SPY 0.51 / QQQ 0.48, model refits on every run.

ΓGamma Map

Symbol	Spot	Net GEX ($B per 1%)	Regime	Zero-gamma flip	Call wall	Put wall	Pin
QQQ	$715.79	+1.46B	Dealers long gamma	$712.33 (spot above)	$720 +0.59% · OI 32,431	$710 -0.81% · OI 37,988	$710
SPY	$762.08	-1.30B	Dealers short gamma	$763.67 (spot below)	$765 +0.38% · OI 26,988	$760 -0.27% · OI 70,918	—

Positive gamma: dealers sell rallies and buy dips, damping moves. Negative gamma: dealers chase, amplifying moves.

Walls are gamma-weighted, not raw OI. 503-day test: a NEAR call wall is real resistance and breakouts through it run shorter; a FAR call wall breaks more often (a volatility tell). Put walls show no measurable support.

DTE≤7 · Walls use the front expiry · post-close snapshot

🕰Analog Days

Prior close profile (2026-09-16, QQQ $704.72): session +0.03% · 5-day -1.62% · vs 200-DMA +6.5% · below 52-week high -5.5% · RSI2=19 · VIX 17.7 · policy rate 3.63% (1-year -0.70)

Closest analog 2024-11-01 (distance 0.32): session +0.74% · 5-day -1.59% · vs 200-DMA +6.3% · below 52-week high -3.1% · RSI2=37 · VIX 21.9 · policy rate 4.83% (1-year -0.50)
What followed: next day -0.29% · 5-day +5.48% · 20-day +5.72%

Side by side | 2026-09-16 (prior session) +0.03% ↔ 2024-11-01 +0.74%; today maps to the day after 2024-11-01, which was -0.29% (single days are mostly coincidence — the Top-20 statistics below are more reliable)

Analog (top 5)	Distance	That day	Next day	5 days on	RSI2 then	20 days on
2024-11-01	0.32	+0.74%	-0.29%	+5.48%	37	+5.72%
2019-10-01	0.39	-0.82%	-1.72%	-0.99%	31	+4.68%
2025-02-26	0.42	+0.24%	-2.78%	-2.44%	17	-5.87%
2015-03-27	0.42	+0.40%	+1.15%	+0.44%	34	+4.49%
2024-04-16	0.44	+0.01%	-1.22%	-1.40%	15	+3.44%

Top-20 analog statistics (deduplicated — far more reliable than any single day):
Next day mean +0.03% (up 50% of the time, range -2.8% to +2.9%) · 5-day +0.93% (65%) · 20-day +2.75% (75%)
Note: QQQ drifts up over time, so the unconditional probability of any given next day being positive is already about 55%. Compare the “up %” above against that baseline, not against 50%.

25 features in three equally weighted groups: price position (return / distance to moving averages / position in the 52-week range / trend slope / consecutive up-down days / gaps) + technicals (RSI2 / RSI14 / volatility / relative volume / volume trend / range / KDJ-J / Bollinger %B / MACD histogram) + environment (VIX / rate cycle / credit spreads / dollar index / oil / S&P P/E / CBOE equity P/C / AAII bull-bear spread). Weighted z-score Euclidean distance; matching pool 1999-03-22–2026-09-16 (6915 days, excluding ±60 days around the target). Narrative context, not a trading signal: after four rounds of feature engineering the Top-1 next-day directional hit rate tops out at 54.9%, still below the 57% you get by always guessing up. The data is given in full, but do not use it for direction — it answers “which historical setup does today resemble”, not “will tomorrow be up or down”.

🎯9:00 Forecast

9:00 premarket price$716.30 (vs prior close +1.64%, Prev close $704.75)

LeanBase — no strong signal, lean is small

P(close above 9:00 price)56%

Expected close$716.4

Expected high$720.5

Expected low$711.8

Expected range1.21% (0.74–2.05%)

Gatebase

Features used5 (Missing inputs: oi)

Model versionv0.3-gates · 2026-09-17 09:02:52 ET

Posted at 09:05 ET, scored at 16:45 ET. The ledger keeps every call — misses included.

✅Scorecard

Called	Actual
P(up) 56% · close $716.4 · range $711.8–$720.5	Open $715.95 · High $718.04 · Low $713.32 · Close $716.92 (Close vs 9:00 price +0.09%)

Direction✓ hit

Range inside band✗ miss 0.67% vs 0.74–2.05%

Expected high touchedno

Expected low touchedno

Brier0.196 Brier (naive) 0.196

Close error0.07% Naive error 0.09%

Running tally · 40 calls scored · direction hit rate 18/40 · avg Brier vs naive 0.259 / 0.259

Naive baseline = "close equals 9:00 price, always 55% up".

🌡Sentiment

AAII Retail Sentiment Survey · week of 2026-09-02

Bull−bear spread +2.1 bulls 39.7% / bears 37.6%

Automatic retrieval found no newer release; this reading is 15 days old
39-year percentile 40.1% · 5-year percentile 58.1% (the distribution has shifted down; use the 5-year figure)

Historically a contrarian indicator, and it takes 3–4 weeks to work, not 1–2. Since 2020 the effect has broken down and at times inverted — low-weight atmosphere only.

CNN Fear & Greed Index · live

28.7 fear

prior close 26.5 · one week ago 32.2 · one month ago 59.1
Percentile since 2022 18.5%

Backtested: no predictive power for same-day or next-day direction (its components are that day's price action, which makes it circular). Forward absolute returns are positive from every bucket; only the magnitude differs — pure atmosphere.

CBOE Equity Put/Call · 2026-09-16

0.69 index P/C 1.02 · total P/C 0.98

Percentile since 2023 77.4% (median over the same period 0.59)
Low = retail crowding into calls (speculative heat); high = heavy put demand (defensive)

Backtested: an extremely low P/C (call mania) does not mark tops — forward returns are in fact slightly positive, the opposite of the popular claim. Atmosphere only.

All four have been backtested: none carries tradable predictive power for same-day or next-day direction. They describe the sentiment backdrop and are not a basis for buying or selling.

Full report: Morning report · Evening report

Research output, not investment advice.