BidWatch

백테스트

official validation은 strict complete canonical record만 사용합니다.

케이스

none · -

적중률

neutral range

평균 오차

P90 -

Official Validation

backtest result pending

best-

score-

strongest baseline-

BidWatch delta-

target-

unit-

confidence-

conflicts excluded0

near miss- / - / -

v1 tuning deferred until 1,000+ evaluated cases.

Hit Win, Error Loss

BidWatch v1 is ahead on hit rate, but MAE/P90 still favors the strongest baseline.

hit delta-

MAE delta-

P90 delta-

v2 strategy appliedno

Router report is diagnostic only; production strategy is unchanged.

Win/Loss	Count	Share	Avg BW Error	Avg BL Error

Router Candidate	Value
baseline fallback segments	-
BidWatch candidate segments	-
insufficient sample	-
P90 risk segments	-
production strategy changed	no

v2 Router Candidate

v2 router is diagnostic only and is not applied to production predictions.

report-onlyproduction changed: false

v2 hit

vs v1 -

v2 MAE

vs baseline -

v2 P90

big miss -

Compare	Hit Δ	MAE Δ	P90 Δ	Big Miss Δ
v2 vs v1	-	-	-	-
v2 vs baseline	-	-	-	-

Router Distribution	Count	Share

Fallback Reason	Count

Big Miss Analysis

P90 threshold - · 0 cases

Case	Segment	Actual	BW Error	BL Error	Delta	Reason

Segment Candidate	Cases	Hit Δ	MAE Δ	P90 Δ	Use

Model Score View	Top Strategy	Score

실행	상태	케이스	적중률	생성

공고	예측 구간	실제	결과

전략 비교

동일한 prior-only 케이스 기준

전략	적중률	MAE	Median Error	P90 Error	평균 폭	Coverage	Score	Cases

Baseline Family	Best Strategy	Hit	Strategies

Validation Notes

source split pending

rate unit-

metric unit-

source gap-

mixed unitsno

Source diagnostics	Cases	Median rate	Miss	Avg participants

Baseline diagnostics	Value
dominance reason	-
global median	-
within 0.05pp	-
global median hit/miss	0 / 0

Source	Cases	Hit	MAE	P90	폭

Conflict	Official	Participant	Diff
conflict warning 없음

Ablation	적중률	MAE	P90	폭	Cases

Segment	Best Strategy	적중률	Cases	Status

Miss Analysis

0 misses

Category	Count