Korea Reserve Prices (복수예비가격) — All 15, and Which Were Drawn

Korean public tenders do not name a price and take the lowest bid. They do this:

base price published
   → 15 candidate reserve prices generated within a band around it
   → 4 drawn at random when bids are opened, averaged
   → that average is the predetermined price
   → the lower-limit rate applied to it sets the award threshold

Every one of those 15 candidates is public after opening — along with which four were drawn. This Actor returns them, one record per candidate, in English.

About 10,000 candidate prices are generated a day, and the draw rate comes out at ~24-27% exactly as 4-of-15 predicts.

#1     52,758,900 KRW    98.892% of base
 #1  3,126,493,300 KRW   101.494% of base   ★ drawn
 #2     52,333,200 KRW    98.094% of base
 #2  3,036,327,900 KRW    98.567% of base   ★ drawn
 #3     52,945,100 KRW    99.241% of base

This is the only public route to the actual distribution of reserve prices and the draw outcomes. If you are modelling what the predetermined price is likely to be on a comparable tender, this is the raw material.

What you get

Group	Fields
The candidate	reserve_price_seq (1-15), reserve_price_total, base_predetermined_price_krw, reserve_vs_base_pct (computed)
The draw	was_drawn (boolean), draw_no
The outcome	predetermined_price_krw, predetermined_net_construction_cost_krw
The anchor	base_price_krw, base_price_step_no
Timing	reserve_prices_created_at, opened_at, registered_at
Link back	notice_no, notice_seq, rebid_no, title_ko, category, url

reserve_vs_base_pct is computed here — the upstream API gives amounts only, and the percentage is what you actually want when comparing across tenders of different sizes.

Input

Every field is optional. The default is the last 7 days across all four categories.

Field	Type	Notes
startDate / endDate	date	YYYY-MM-DD, Korea time
categories	array	goods, construction, service, foreign
keywords	array	Match on notice title or agency. Korean or English
drawnOnly	boolean	Return only the 4 drawn candidates instead of all 15
minBasePriceKrw	integer	Skip tenders below this base price
maxItems	integer	Stop after N records. Each tender is ~15 records

drawnOnly is the fast path if you are studying where in the band the draw lands rather than the full distribution.

Who this is for

• Bidders modelling the predetermined price before submitting
• Quantitative analysts studying the draw distribution across thousands of tenders
• Bid-support tools that need the full band, not just the summary
• AI agents reasoning about Korean procurement pricing

Join on notice_no with Korea Public Tenders (나라장터) for the notice, Korea Tender Base Prices (기초금액) for the published band, and Korea Tender Award Results (낙찰정보) for how it actually ended.

Notes

Source. 조달청 나라장터 낙찰정보서비스, published on Korea's national open-data portal under an unrestricted licence (이용허락범위 제한 없음). Every record carries a source field and a url back to the original notice.

This is an API client, not a scraper. Data comes from the government's own REST API with a registered key, so nothing breaks when g2b.go.kr is redesigned.

No personal data. Contact details are stripped from free-text fields and are never emitted.

---

Development

export G2B_SERVICE_KEY='...'   # data.go.kr 인증키 (다른 Actor와 같은 키)
python3 probe.py               # 오퍼레이션·필드명 대조
apify push                     # 배포