Model accuracy

Model accuracy

Measured walk-forward on real history — we don’t cherry-pick winners. The weak numbers are shown as they are.

Walk-forward results · US · every model disclosed

Market US

ModelPrice toward value in 12 mo.nBelow floor: excess return, 12 mo.
Lynch Fair Value44.6%280-8.3% (n=61)
Sector P/E42.2%1,411-7.4% (n=463)
Buffett DCF40.7%359-1.1% (n=44)
Shiller CAPE39.7%479-8.4% (n=78)
Greenblatt Magic36.0%1,314-4.3% (n=161)
Pabrai 10x35.4%1,451+0.5% (n=109)
Gordon DDM35.1%1,054-16.4% (n=8)
Graham Number34.4%1,315+3.0% (n=68)
Justified P/E34.4%1,399+2.7% (n=66)
Nash NCAV34.2%849+31.5% (n=23)
Graham NCAV33.7%205
Schloss Net-Net33.7%205

Market Thailand (SET)

ModelPrice toward value in 12 mo.nBelow floor: excess return, 12 mo.
Gordon DDM62.6%3,073+14.4% (n=149)
Schloss Net-Net62.4%2,571+22.1% (n=70)
Graham NCAV59.7%2,571+11.6% (n=196)
Justified P/E57.3%4,772+8.6% (n=629)
Greenblatt Magic54.7%3,709+2.5% (n=748)
Sector P/E53.9%3,711+2.9% (n=807)
Pabrai 10x53.5%3,915-2.0% (n=832)
Nash NCAV52.8%4,233+7.6% (n=1,105)
Buffett DCF51.4%3,410-2.5% (n=1,004)
Graham Number50.7%3,709+3.7% (n=1,023)
Lynch Fair Value49.1%1,581-1.6% (n=640)
Shiller CAPE45.5%3,975+2.0% (n=2,220)

DAVI Score composite · USv1 (equal vote across models) vs v2 (weighted by measured accuracy)

Measured walk-forward over the same window (2022→now): hit = the 12-month peak after entry ≥ +20% · the numbers show whether a higher score really means a higher hit rate — not a forecast.

v1 — raw votes

0 votes55.5%n=1,073
1-2 votes56.7%n=446
3-5 votes62.4%n=173
6-8 votes79.0%n=19

v2 — accuracy-weighted (live)

0% weighted votes55.2%n=1,236
1-25% weighted votes60.1%n=338
26-50% weighted votes60.6%n=99
51-75% weighted votes66.7%n=18
76-100% weighted votes (top)70.6%n=17

DAVI Score composite · THv1 (equal vote across models) vs v2 (weighted by measured accuracy)

Measured walk-forward over the same window (2022→now): hit = the 12-month peak after entry ≥ +20% · the numbers show whether a higher score really means a higher hit rate — not a forecast.

v1 — raw votes

0 votes27.6%n=780
1-2 votes27.1%n=1,023
3-5 votes35.1%n=709
6-8 votes37.6%n=447
9+ votes (DAVI Score top)47.2%n=199

v2 — accuracy-weighted (live)

0% weighted votes25.9%n=1,529
1-25% weighted votes34.1%n=552
26-50% weighted votes34.9%n=427
51-75% weighted votes37.8%n=355
76-100% weighted votes (top)47.2%n=265

ML Ranking Model (XGBoost v1)

Went live May 2026 — the live history is still too short to judge 12-month forecast accuracy. First 2–5 week rank-IC (Spearman): US -0.019 (a value near 0 = no forecast signal yet in the short window). We’ll keep reporting this every month as it is, good or not.

Computed 10 Jun 2026

Backtested results are not investment advice and do not guarantee future outcomes · a hit-rate near 50% = no different from random.

Past accuracy does not guarantee future results · DAVI is an educational calculation tool, not investment advice.