Opening

Hi everyone, I'm Mu-Yan, and this is "AI Stock Market Chat." This episode was recorded on 2026-11-02 (Monday), and we'll be looking at the TAIEX outlook for the next trading day, 2026-11-03 (Tuesday). The model is bearish on both the next-day and short-term horizons, and the biggest wildcard today is the SOX index, which plunged more than 7% in a single session. Let me say this up front: everything in this program is the output of an AI quantitative model. We do not provide individual stock buy/sell recommendations, we do not manage accounts, and we do not promote stocks. Please make your own judgments.

Today's Index Forecast

This service forecasts only the TAIEX, across two horizons: "next day" and "short term." Here, "next day" refers to the target trading day of 2026-11-03 (Tuesday), and "short term" means roughly one week starting from 11-03.

Fine print: the percentage is the change of the forecast range midpoint relative to the current price; the confidence value is an internal model score, not a hit probability.

There's something I need to be upfront about here. The model's raw output listed -2.49% in the percentage column for both the next-day and short-term rows, but if you check against the actual target ranges, you'll find that the next-day range midpoint of 44714.78 is only about -1.10% from the current price, while the short-term midpoint of 44100.39 works out to about -2.46%, which is close to -2.49%. In other words, that -2.49% in the next-day row was very likely carried over from the short-term figure. To avoid giving listeners the wrong idea about the expected decline, this entire episode uses a single consistent calculation: the range midpoint relative to the current price. So the table above shows -1.10% for next day and -2.46% for short term, one standard throughout.

Also, the direction and confidence are identical across both rows (both bearish, both 75.0%). That's consistent in the model's own output: the directional call and confidence score come from the same source, and the only difference is the time window. Next day looks at just 1 day; short term looks at 5 days. Stretch the timeframe and uncertainty grows, so the range naturally widens and the midpoint drops lower. That's why the two ranges differ, not because the table contradicts itself.

On market tone, the sentiment report leans bearish: the TAIEX closed down -2.49% on the day, and market breadth data showed 92 advancers against 952 decliners, totaling 1,044 issues, meaning roughly 90% declined. One caveat though: the raw data doesn't specify the scope of those 1,044 issues (whether it covers the listed market only or listed plus OTC, and whether unchanged or zero-volume stocks were excluded). So "90% declining" should only be read as a proportion within that 1,044-stock sample, with the denominator defined by the data source. Add the heavy selloff in U.S. semiconductors, and the bearish pressure is mainly external.

Global Market Snapshot

The following are current closing figures only, with no forecasts attached.

One apples-to-apples comparison: the TAIEX futures night session at 44433 is already close to the lower bound of the next-day forecast range at 44383. In other words, the night session has already priced in part of the external bad news. That's a statement of current conditions, not a claim that the next day's open will necessarily follow. SOX at -7.2% is the heaviest hit of the four indices, and the Nasdaq's -2.86% is clearly weaker than the Dow's -1.09%. The selling is concentrated in tech and semiconductors, and I don't need to spell out what that means for the TAIEX's weighting structure.

AI Forecasts for Key Names

Let me set the ground rules for this section first: this service's AI forecast model only outputs direction, target range, and confidence for the TAIEX, and today's data contains no forecasts for any individual stock. So this episode will not, and cannot, assign a direction, target price, or confidence score to any individual stock.

The "Strategy Screening Radar" in the next section is the output of a separate strategy screener and is a completely different thing. It only presents scores, valuation tags, and technical signal states. It makes no bullish or bearish calls and is not a buy/sell forecast. Please don't conflate the two.

Strategy Screening Radar

A reminder: the following are screening results, not individual stock buy/sell forecasts. No direction, no target price, no confidence score.

A patch on the valuation tags first: the data source only provides three labels, expensive / cheap / fair, and gives no criteria for how they're assigned (whether relative to the industry median or relative to the stock's own historical P/E range is never stated). That's how you end up with counterintuitive combinations like Yang Ming at a P/E of 14.3 tagged expensive while Formosa Petrochemical at 17.07 is tagged cheap. I'm not making this up, the data just looks like that. With the baseline unclear, treat this column as reference only, not a valuation conclusion. One more bit of common sense: for cyclicals like Yang Ming, a low P/E often coincides with peak earnings, so valuation tags on cyclical stocks are prone to distortion by nature.

Novatek's P/E and valuation fields are blank in the data source, with no reason given (possibly stale data or not computable for the period). I'm uniformly marking those as "data missing / N/A" and making no estimates.

The screener commentary for each name is presented strictly as raw system scores and indicator values:

• Yang Ming (2609): Composite score of 1.0 today (system rule: ≥2 leans buy, ≤ −2 leans sell, in between is neutral, so this currently sits in neutral territory). The system flagged a bullishEngulfing candlestick pattern. That pattern name is standard technical analysis terminology and does not represent a bullish or bearish view from this program.
• Formosa Petrochemical (6505): Composite score of 1.2, also in neutral territory. The system flagged Williams %R at -95.94, a value falling within the system-defined oversold range.
• Hon Hai (2317): Composite score of 0.0; the system found no clear signal today. Technical description: moving averages tangled, no clear direction, price below the monthly MA, RSI 48 (>70 is hot, <30 is cold), KD turning down in the short term.
• Novatek (3034): Composite score of 0.0, no clear signal. Technical description: short-, medium-, and long-term moving averages in bullish alignment, price above the monthly MA, RSI 54, KD turning down in the short term.
• Chunghwa Telecom (2412): Composite score of -1.7 (negative territory), which under the system rules has not yet crossed the ≤ -2 signal threshold, so it's classified as no signal triggered. Indicator states: CCI 122.18 (system flags strengthening momentum), RSI 83 and Williams %R 0, both falling within the system-defined overbought range.

On chip scores, the first four names all scored 100 and Chunghwa Telecom scored 88. This is the screener's institutional-flow dimension score, again just a number, with no buy or sell implication attached.

Recent Forecast Accuracy Review

I have to report this honestly. The recent hit rate for the TAIEX's next-day direction is 47.4% (9 out of 19), with the full range hit only 5 out of 19 times and an average error of 1.17%. The short-term direction hit rate is 53.3% (8 out of 15), with the full range hit 3 out of 15 times and an average error of 2.51%.

Bluntly put: a next-day directional hit rate of 47.4% is below a coin flip, which means this kind of short-horizon forecast has limited reference value. Please do not base any trading decision solely on this program's content. The short-term 53.3% is slightly above half, but a 15-sample set isn't enough to prove anything, and the average error widens to 2.51%. So about today's 75% confidence value, let me stress it one more time: that's an internal model score, not a win rate. Don't equate the two.

One final reminder: this program does not provide individual stock buy/sell recommendations, does not manage accounts, and does not promote stocks. All content is for research reference only.

Disclaimer

This content is automatically generated by an AI model based on quantitative data and is for research reference only. It does not constitute investment advice or any form of recommendation. This site and the program's producers are not a securities investment consulting enterprise, do not provide analysis to specific individuals, and receive no compensation for this content. Investing carries risk; please exercise your own judgment and bear your own responsibility.