LearnModel Accuracy
Measured walk-forward on real history · we do not cherry-pick; weak numbers show as they are
Walk-forward results · US · Thailand · Korea
How often the price moved toward value within 12 months, with the sample size
Lynch Fair Value
43.7% · n=334 · excess −1.3% (n=66)
Buffett DCF
40.2% · n=371 · excess +3.2% (n=49)
Shiller CAPE
38.9% · n=550 · excess +1.3% (n=78)
Nash NCAV
38.4% · n=2,356 · excess +16.3% (n=55)
Sector P/E
37.9% · n=1,521 · excess 0.0% (n=443)
Pabrai 10x
35.5% · n=1,508 · excess +2.4% (n=123)
Greenblatt Magic
35.4% · n=1,422 · excess
Every DAVI Score version · ML Ranking · Neural Score
DAVI calculates each stock's value range (floor–ceiling) from 12 published models and compares it with the market price · every value is an educational calculation
Public financials
The income statement, balance sheet and cash flows a company publishes
Run 12 models
Each model gives a floor, a ceiling or a value, by its published formula
Compare to the price
Shows where today’s price sits in the range; a description of position only
Each model follows a published investing framework, with its formula
Nash NCAV
Variant of Graham 1934 + Graham zero-growth earnings power (US)
Floor = max((CA − Non-Current Liab) / Shares, 8.5 × EPS_norm — US asset-light fallback)
Graham NCAV
Benjamin Graham, "Security Analysis" (1934)
Floor = (Current Assets − Total Liab) / Shares
Graham Number
Benjamin Graham, "The Intelligent Investor" (1949)
Floor = √(22.5 × EPS × BVPS)
Buffett DCF
Discounted Cash Flow (textbook)
10y FCF discounted + Gordon terminal value
Lynch Fair
Peter Lynch, "One Up on Wall Street" (1989)
Fair = EPS × Earnings Growth Rate (PEG = 1)
Gordon DDM
Myron Gordon (1959)
Value = D₁ / (r − g)
Justified P/E
CFA equity textbook
Value = BVPS × (ROE − g) / (r − g)
Sector P/E
Sector median multiplier
Value = EPS × sector median P/E
Greenblatt Magic
Joel Greenblatt, "The Little Book That Beats the Market" (2006)
Quality-tiered P/E (16× / 13× / 10×) × EPS — by ROIC
Pabrai 10×
Mohnish Pabrai, "The Dhandho Investor" (2007)
Fair = Owner Earnings × 10 (heads I win, tails I don't lose)
Schloss Net-Net
Walter Schloss, Graham disciple, WJS Partners (50y track)
Floor = NCAV × ⅔ (Schloss’s threshold: 67% of net-net)
Shiller CAPE
Robert Shiller, Yale (Nobel 2013)
Fair = avg EPS (10y cycle) × historical mean P/E (16)
Justified P/E
34.4% · n=1,456 · excess +7.6% (n=77)
Graham Number
34.2% · n=1,370 · excess +5.3% (n=70)
Gordon DDM
33.4% · n=1,132 · excess −13.6% (n=12)
Graham NCAV
33.3% · n=216 · excess +0.8% (n=1)
Schloss Net-Net
32.9% · n=216 · excess +0.8% (n=1)
About 50% = no better than chance · highest first, weak numbers shown as they are · Second line = price toward value in 12 mo. · sample size · 12-month excess return of the below-floor group
Schloss Net-Net
63.5% · n=3,060 · excess +55.8% (n=90)
Graham NCAV
61.6% · n=3,060 · excess +46.7% (n=238)
Gordon DDM
61.4% · n=3,314 · excess +39.0% (n=192)
Justified P/E
56.8% · n=5,695 · excess +23.4% (n=770)
Pabrai 10x
56.2% · n=4,615 · excess +5.2% (n=721)
Buffett DCF
54.8% · n=3,955 · excess +2.5% (n=903)
Greenblatt Magic
54.6% · n=4,402 · excess +9.7% (n=932)
Nash NCAV
53.0% · n=5,959 · excess +13.2% (n=1,708)
Graham Number
51.6% · n=4,402 · excess +12.1% (n=1,276)
Lynch Fair Value
48.5% · n=1,750 · excess +0.7% (n=730)
Sector P/E
48.1% · n=4,408 · excess +8.6% (n=1,600)
Shiller CAPE
46.4% · n=4,311 · excess +9.8% (n=2,462)
About 50% = no better than chance · highest first, weak numbers shown as they are · Second line = price toward value in 12 mo. · sample size · 12-month excess return of the below-floor group
Lynch Fair Value
62.5% · n=8 · excess +38.3% (n=6)
Graham Number
57.9% · n=57 · excess +11.6% (n=22)
Greenblatt Magic
54.4% · n=57 · excess +20.5% (n=19)
Pabrai 10x
51.7% · n=58 · excess −6.9% (n=16)
Shiller CAPE
47.4% · n=19 · excess +58.2% (n=9)
Justified P/E
43.9% · n=66 · excess +18.0% (n=15)
Sector P/E
35.1% · n=57 · excess +5.4% (n=19)
Nash NCAV
33.9% · n=59 · excess −20.8% (n=4)
Graham NCAV
33.3% · n=15 · excess —
Schloss Net-Net
33.3% · n=15 · excess —
Gordon DDM
28.1% · n=32 · excess —
Buffett DCF
9.1% · n=11 · excess +2.9% (n=2)
About 50% = no better than chance · highest first, weak numbers shown as they are · Second line = price toward value in 12 mo. · sample size · 12-month excess return of the below-floor group
Right column = 12-month excess return of the below-floor group vs the equal-weight market
The score in the app today is version 3 (value vs the floor + price momentum) · version 3 has no published measurement yet; below are the earlier versions
Walk-forward, same window (2022 → today): a hit is a 12-month peak at least 20% above the entry price · the numbers show whether higher buckets really hit more often. Not a forecast.
Version 3 · live now
No measurement recorded yet
Version 2 · weighted by measured accuracy
| Bucket | hit | n |
|---|---|---|
| 0% of accuracy weight | 52.6% | n=3,048 |
| 1–25% of accuracy weight | 58.6% | n=916 |
| 26–50% of accuracy weight | 65.0% | n=206 |
| 51–75% of accuracy weight | 54.5% | n=44 |
| 76–100% of accuracy weight · top bucket | 62.5% | n=40 |
Version 1 · every model weighted equally
| Bucket | hit | n |
|---|---|---|
| No model below its floor | 51.7% | n=2,634 |
| 1–2 models below their floor | 58.3% | n=1,188 |
| 3–5 models below their floor | 60.7% | n=364 |
| 6–8 models below their floor | 70.0% | n=50 |
| 9+ models below their floor · top bucket | 76.2% | n=21 |
The score in the app today is version 3 (value vs the floor + price momentum) · version 3 has no published measurement yet; below are the earlier versions
Walk-forward, same window (2022 → today): a hit is a 12-month peak at least 20% above the entry price · the numbers show whether higher buckets really hit more often. Not a forecast.
Version 3 · live now
No measurement recorded yet
Version 2 · weighted by measured accuracy
| Bucket | hit | n |
|---|---|---|
| 0% of accuracy weight | 28.0% | n=2,076 |
| 1–25% of accuracy weight | 35.1% | n=1,081 |
| 26–50% of accuracy weight | 40.1% | n=639 |
| 51–75% of accuracy weight | 40.0% | n=552 |
| 76–100% of accuracy weight · top bucket | 49.5% | n=311 |
Version 1 · every model weighted equally
The score in the app today is version 3 (value vs the floor + price momentum) · version 3 has no published measurement yet; below are the earlier versions
Walk-forward, same window (2022 → today): a hit is a 12-month peak at least 20% above the entry price · the numbers show whether higher buckets really hit more often. Not a forecast.
Version 3 · live now
No measurement recorded yet
Version 2 · weighted by measured accuracy
No measurement recorded yet
Version 1 · every model weighted equally
| Bucket | hit | n |
|---|---|---|
| No model below its floor | 57.5% | n=40 |
| 1–2 models below their floor | 58.6% | n=29 |
| 3–5 models below their floor | 91.7% | n=24 |
| 6–8 models below their floor | 100.0% | n=8 |
| 9+ models below their floor · top bucket | 100.0% | n=4 |
Live since May 2026; the live history is too short to judge 12-month accuracy · rank-IC (Spearman) over the first 2–5 weeks, US 0.029 (near 0 = no short-horizon predictive power seen yet) · reported monthly as it is
Measurement not published yet
Computed 20 Oct 2026
Backtest results are not investment advice, and past results do not indicate future ones · a hit-rate near 50% = no better than chance
| 54.8% |
| 3,955 |
| +2.5% (n=903) |
| Greenblatt Magic | 54.6% | 4,402 | +9.7% (n=932) |
|---|
| Nash NCAV | 53.0% | 5,959 | +13.2% (n=1,708) |
|---|
| Graham Number | 51.6% | 4,402 | +12.1% (n=1,276) |
|---|
| Lynch Fair Value | 48.5% | 1,750 | +0.7% (n=730) |
|---|
| Sector P/E | 48.1% | 4,408 | +8.6% (n=1,600) |
|---|
| Shiller CAPE | 46.4% | 4,311 | +9.8% (n=2,462) |
|---|
| 43.9% |
| 66 |
| +18.0% (n=15) |
| Sector P/E | 35.1% | 57 | +5.4% (n=19) |
|---|
| Nash NCAV | 33.9% | 59 | −20.8% (n=4) |
|---|
| Graham NCAV | 33.3% | 15 | — |
|---|
| Schloss Net-Net | 33.3% | 15 | — |
|---|
| Gordon DDM | 28.1% | 32 | — |
|---|
| Buffett DCF | 9.1% | 11 | +2.9% (n=2) |
|---|
| Bucket | hit | n |
|---|
| No model below its floor | 27.8% | n=1,229 |
|---|---|---|
| 1–2 models below their floor | 31.5% | n=1,463 |
| 3–5 models below their floor | 36.6% | n=1,070 |
| 6–8 models below their floor | 44.0% | n=669 |
| 9+ models below their floor · top bucket | 44.1% | n=299 |