A stock forecaster's low-end misses keep growing
A Korean outlet scores its own weekly KOSPI price forecasts against what actually happened. Three straight weeks: high forecasts stayed accurate, low forecasts missed by more each week — 2.89%, 6.61%, 8.10%.
AI-assisted, human-reviewed sourcing. Figures are pulled programmatically from the official sources listed at the end of this article and checked by an editor before publication.
kospiforecastinganalystskoreaequities
Every week, the Korean outlet BigTaNews publishes an expected high and an expected low for the 15 largest KOSPI stocks by market cap, then — the following week — publishes how far off it was. Three of those scorecards are now public, back to back, for the weeks of 18 August, 24 August and 31 August. Line them up and one half of the forecast holds steady while the other keeps getting worse.
The split
| Week | High-forecast error | Low-forecast error |
|---|---|---|
| 18 Aug | 3.74% | 2.89% |
| 24 Aug | 3.77% | 6.61% |
| 31 Aug | 2.56% | 8.10% |
The high-side error sits in a narrow band across all three weeks — never below 2.56%, never above 3.77%. The low-side error very nearly triples over the same stretch, from 2.89% up to 8.10%.
Both numbers are means of the absolute per-stock error across the same 15 names each week (Samsung Electronics, SK Hynix, SK Square, Samsung Electro-Mechanics, Hyundai Motor, LG Energy Solution, Samsung Life, Samsung C&T, Samsung Biologics, HD Hyundai Heavy Industries, SK, Doosan Enerbility, Kia, KB Financial and Hanwha Aerospace). Recomputing the average from BigTaNews’ own published per-stock figures matches what they report, in all three weeks.
Where it missed worst
The single biggest miss anywhere in the dataset is on the high side: Doosan Enerbility’s forecast for the week of 24 August was off by 13.79%. But the low side produced the two next-biggest misses, both on the same stock — Samsung Electro-Mechanics missed its low forecast by 18.87% the week of 24 August, then by 16.31% the week after. SK Hynix and SK Square each missed their low forecasts by more than 17% in the week of 31 August.
Those are exceptions in degree, not in direction. In the week of 31 August, 12 of the 15 stocks’ actual lows came in above the forecast low — meaning the stock didn’t fall as far as BigTaNews expected. Only one came in below.
What was happening in the market
This stretch overlapped with a KOSPI run higher — the index closed at a record 6,995 on 7 September, up 4.61% on the day. A market moving mostly one direction will tend to make “how low will it go” forecasts look too pessimistic in hindsight, and “how high will it go” forecasts easier to hit if the trend was already priced into them. That’s a plausible backdrop, not a proof: this piece measures what BigTaNews published, not why its low-end forecasts undershot.
Not investment advice.
Data & Verification Notes
- Data as of
- Sources
- Cross-checks
- Three weekly scorecards, each covering the same 15 KOSPI stocks by market cap, read directly from BigTaNews' own published tables: week of 18 Aug (high error 3.74%, low error 2.89%), week of 24 Aug (high error 3.77%, low error 6.61%), week of 31 Aug (high error 2.56%, low error 8.10%)
- Error is (forecast − actual) / forecast for each of the 15 stocks. BigTaNews' published weekly average is the mean of the absolute values across all 15; recomputing that mean from the individual figures matches BigTaNews' own published average in all three weeks
- The single largest miss anywhere in the three weeks is Doosan Enerbility's high forecast for the week of 24 Aug, off by 13.79%. The single largest low-side miss is Samsung Electro-Mechanics, also the week of 24 Aug, off by 18.87% — and the same stock's low forecast missed by 16.31% the following week
- High-forecast error moved narrowly across the three weeks — 3.74%, 3.77%, 2.56%. Low-forecast error nearly tripled — 2.89%, 6.61%, 8.10%
- Excluded figures
- Why the gap opened. This reads BigTaNews' own published error rates; it does not examine their forecasting method or model
- Any claim about future weeks. Three weeks is a short run, and whether the pattern holds beyond it is unmeasured here
- A ranking of which of the 15 stocks BigTaNews forecasts best — the aggregate weekly pattern is the finding, not a per-stock scorecard
- Not investment advice.
- Instruments
- 005930, 000660
The full dataset behind these figures — every listed company, every row — is available to license. SeoulMarkets data for licence →
Not investment advice. SMarkets publishes data journalism for general information only. Nothing here is investment advice, a recommendation, or an offer to buy or sell any security — any investment you make is at your own risk and responsibility. Figures are derived from official public data sources and may be revised by the issuing agency. Verify independently before acting.
← More Equities