Damodaran 상대가치 검산

공개 호출 방식

import dartlab
import importlib.resources as resources
import json
from pathlib import Path

import polars as pl
from dartlab.synth.damodaranL15 import buildDamodaranMemo

target = "005930"
c = dartlab.Company(target)
market = getattr(c, "market", "US" if not target.isdigit() else "KR")
currency = getattr(c, "currency", "USD" if market == "US" else "KRW")
company_name = getattr(c, "corpName", getattr(c, "companyName", target))


def _loadReference(name):
    return json.loads(resources.files("dartlab.reference.data").joinpath(name).read_text(encoding="utf-8"))


def _safeShow(topic):
    try:
        table = c.show(topic, freq="Y")
    except TypeError:
        table = c.show(topic)
    except Exception:
        return pl.DataFrame()
    return table if isinstance(table, pl.DataFrame) else pl.DataFrame()


def _latestPrice(frame):
    if not isinstance(frame, pl.DataFrame) or frame.height == 0:
        return {}
    date_col = "date" if "date" in frame.columns else "Date" if "Date" in frame.columns else None
    close_col = "close" if "close" in frame.columns else "Close" if "Close" in frame.columns else None
    latest = frame.sort(date_col).tail(1).to_dicts()[0] if date_col else frame.tail(1).to_dicts()[0]
    out = {}
    if close_col and latest.get(close_col) is not None:
        out["price"] = latest.get(close_col)
    if date_col and latest.get(date_col) is not None:
        out["priceDate"] = str(latest.get(date_col))
    return out


def _marketData():
    out = {}
    try:
        price_frame = dartlab.gather("price", target, market="US") if market == "US" else dartlab.gather("price", target)
        out.update(_latestPrice(price_frame))
    except Exception as exc:
        out["priceError"] = type(exc).__name__

    if market == "KR":
        krx_path = Path("data/krx/prices/raw-2026.parquet")
        if krx_path.exists():
            try:
                krx = (
                    pl.scan_parquet(str(krx_path))
                    .filter(pl.col("ISU_CD") == target)
                    .select(["BAS_DD", "TDD_CLSPRC", "MKTCAP", "LIST_SHRS"])
                    .sort("BAS_DD")
                    .tail(1)
                    .collect()
                )
                if krx.height:
                    row = krx.to_dicts()[0]
                    out.update(
                        {
                            "price": row.get("TDD_CLSPRC") or out.get("price"),
                            "priceDate": str(row.get("BAS_DD") or out.get("priceDate")),
                            "marketCap": row.get("MKTCAP"),
                            "shares": row.get("LIST_SHRS"),
                        }
                    )
            except Exception as exc:
                out["marketCapError"] = type(exc).__name__

    if market == "US" and out.get("price") is not None:
        cik = str(getattr(c, "cik", "") or "")
        for path in (Path(f"data/edgar/finance/{cik}.parquet"), Path(f"data/edgar/finance/{target}.parquet")):
            if not path.exists():
                continue
            try:
                shares = (
                    pl.scan_parquet(str(path))
                    .filter((pl.col("unit") == "shares") & pl.col("tag").str.contains("SharesOutstanding"))
                    .select(["val", "filed"])
                    .sort("filed")
                    .tail(1)
                    .collect()
                )
                if shares.height:
                    out["shares"] = shares["val"][0]
                    out["marketCap"] = float(out["price"]) * float(out["shares"])
                    break
            except Exception as exc:
                out["marketCapError"] = type(exc).__name__
    return out


country_defaults = _loadReference("damodaranDefaults.json")
industry_defaults = _loadReference("damodaranIndustryDefaults.json")
statements = {topic: _safeShow(topic) for topic in ("IS", "BS", "CF")}
memo = buildDamodaranMemo(
    target=target,
    market=market,
    currency=currency,
    companyName=company_name,
    statements=statements,
    countryDefaults=country_defaults,
    industryDefaults=industry_defaults,
    marketData=_marketData(),
)

emit_result(
    table=memo["tables"]["relativeCheck"],
    values=memo["headline"],
    date=memo.get("asOf"),
    units=memo["units"],
    sources=memo["sources"],
)

호출 동작

1. 결론 도출

상대가치는 DCF를 대체하지 않고 sanity check로만 쓴다. DCF가 peer multiple 분포와 크게 어긋나면 가정 재검토를 요구한다.

2. 핵심 근거 수집

KR은 scan("valuation") snapshot과 가격 path를 쓴다. US는 v1에서 가격 path만 확인하고 peer valuation primitive 부재를 gap으로 남긴다.

3. 메커니즘 분석

EV/Sales는 마진과 sales-to-capital 가정의 sanity check, EV/EBIT은 정상 마진의 sanity check, PB는 금융업·자본집약 업종의 보조 체크로 쓴다.

4. 반례·한계

적자 기업 PE, 현금 과다 기업 EV multiple, 회계 기준이 다른 peer group은 결론 강도를 낮춘다.

5. 후속 모니터링

US valuation scan 구현, peer group mapping, market cap/share count normalization을 gap ledger로 넘긴다.

대표 반환 형태

relativeCheck : dict — multiples, peerCoverage, sanityFlags, missingPrimitives, status를 담는다.

연계 절차

1. recipes.fundamental.valuation.damodaran.fcffDcf - DCF 결과 입력.
2. recipes.fundamental.valuation.damodaran.scenarioFalsifier - multiple이 깨는 가정 반증.
3. recipes.fundamental.valuation.damodaran.deepDive - 최종 memo에 gap 반영.

기본 검증

• US 기업은 partial 또는 blocked 표시 없이 relative valuation 완료 선언 금지.
• multiple 결과는 DCF 가정 검산으로만 사용한다.