What does not workSAVNG Research · September 10, 2026

Net-nets: the ratio does not sort, and the strict rule is a lottery

The claim. If you had bought the fifth of companies that scored best on ncav / price (net-nets) at each annual report, 2010 to 2025, and held for a year, you would have beaten the S&P 500 by +8.2 pts on average (very unlikely to be luck) and the worst-scoring fifth by +1.6 pts; against small caps the gap was +1.3 pts. Bought during a sell-off the best fifth returned +33.5 pts (135 cases). Measured on 5,065 company-years across 406 companies.

Ranking companies by net current asset value over price barely separates winners from losers. The strict Graham rule (price below net current assets) did return a high average, but on fewer than a hundred cases with one year carrying most of it, and a median far below the average.

How we tested it
  • Universe: 657 U.S. companies with ten-plus years of prices (8,703 company-years). Non-financials. Net current asset value per share ÷ price; highest = top fifth. Classic net-nets (NCAV/price ≥ 1) are also reported as their own bucket.
  • Point in time: factor values from the latest fiscal year known on each 10-K filing date; price = first close on or after the filing.
  • Outcome: total return over the next 12 and 36 months minus the S&P 500, and minus small caps (IWM) for 12 months; companies ranked into fifths within each filing year; market state = S&P ≥10% below its one-year high on the entry day.
  • Limits: survivorship-tilted, small-cap-heavy universe; no costs; the spread between fifths is the finding, the levels are not.
Reproduce: node savng-insider-history/factor-study.js → reports/factor-study.json (generated 2026-09-11); served at /screener-evidence; any filter combination live at /backtest.
The picture
Q1 (best by this factor)+8.2 pts n 1,021 · median −4.8 ptsQ2−6.1 pts n 1,013 · median −10.6 ptsQ3−2.2 pts n 1,011 · median −5.5 ptsQ4−1.4 pts n 1,013 · median −4.3 ptsQ5 (worst)+6.6 pts n 1,007 · median −1.5 pts

Average twelve-month return minus the S&P 500 by fifth, best-scoring first. Medians in grey.

The strict rule
BucketCompany-yearsAverage vs S&PMedian% beatt
NCAV/price ≥ 1 (true net-nets)86+32.3 pts+7.2 pts54%2.0
Sell-offs versus calm markets
Market state at entryBest fifthWorst fifthSpread
Sell-off (S&P ≥10% off high)+33.5 pts n 135+24.1 pts+9.4 pts
Calm+4.3 pts n 886+4.3 pts+0.0 pts
Does it apply to your sector?
Sector (SEC group)Company-yearsBest fifth, 12 moWorst fifthSpreadRead
Industrials & manufacturing1,556+4.2 pts n 377+8.4 pts−4.2 ptsDoes not sort
Services575+9.7 pts n 72+5.6 pts+4.1 ptsSorts weakly
Tech hardware & semis486+6.3 pts n 181−4.7 pts+11.0 ptsSorts well
Biotech & pharma373+30.7 pts n 69+88.0 pts−57.3 ptsDoes not sort
Other295+0.7 pts n 82+21.0 pts−20.2 ptsDoes not sort
Retail & consumer293−0.7 pts n 29−1.0 pts+0.3 ptsDoes not sort
Medical devices260+18.6 pts n 69+20.4 pts−1.8 ptsDoes not sort
Energy240−14.9 pts n 14+15.6 pts−30.4 ptsToo few to say
Software & IT services224+15.6 pts n 55−39.8 pts+55.4 ptsSorts well
Utilities151+84.1 pts n 7−3.0 pts+87.1 ptsToo few to say
Finance & real estate132−16.0 pts n 13+6.8 pts−22.8 ptsToo few to say
Where it applies today
Market state today: Calm — the S&P 500 closed 3.1% below its 1-year high (as of 2026-09-16). Our rule, the same one used to split every backtest on this site: sell-off = 10% or more below the 1-year high, borderline from 5%, calm otherwise. A further 7.2% fall from here would put us in a sell-off. Since 2006 the market has been in a sell-off 16.6% of the time, across 13 episodes (red bands).
2007200920112013201520172019202120232025
S&P 500 (SPY, adjusted) distance from its trailing 1-year high, weekly. Amber band: borderline (−5% to −10%). Red band: sell-off (below −10%). The dot is today. History to 2026-08-19 from the research database; today from the live feed.
Backtest · point in time · 8,702 entries in the study

The same screen, live — every filter re-tested as you change it

Screen tested: Net current asset value ≥ 1.00× price (Graham net-nets) · bought the day of the annual report, held 12 months
Too few cases Buy every company that met this screen on the day it filed its annual report, hold 12 months: on average it beat the S&P 500 by +30.5 pts (93 entries, 31 companies, unlikely to be luck). 53% of entries beat the S&P; 27% beat it by 30 points or more; 27% trailed it by 20 points or more. Owning everything instead averaged +1.7 pts, so the screen's edge is +28.8 pts.
+30.5 ptsvs S&P, average
53%beat the S&P
27%up 30+ pts vs S&P
27%down 20+ pts vs S&P
+31.8 ptsvs small caps, 12 mo
−43.6 pts 2019worst year to buy ?
2.0luck check ?
Does the holding period matter?3 mo: +8.4 pts 50% beat6 mo: +13.0 pts 55% beat12 mo: +30.5 pts 53% beat24 mo: +48.9 pts 53% beat3 yr, per yr: +8.0 pts 48% beat
Market state at entryCompany-yearsvs S&P 12 movs small caps
Calm ◀ today82+10.8 pts t 1.8+12.5 pts
Sell-off (S&P ≥10% off high)11+176.8 pts t 1.5+176.0 pts
Sector (SEC group)nvs S&P 12 mo3 yrs / yr
Industrials & manufacturing27+35.0 pts t 1.4+11.1 pts

Holds up out of sample? Before 2018: +13.2 pts (n 48, t 1.9). From 2018: +48.9 pts (n 45, t 1.6). Same sign in both halves.

Year201120122013201420152016201720182019202020212022202320242025
vs S&P−27 pts+37 pts+33 pts−7 pts+6 pts+15 pts+9 pts+13 pts−44 pts+292 pts+29 pts+26 pts+4 pts−16 pts+28 pts
n5118866447765556

Read the research behind this screen: Net-nets: the ratio does not sort, and the strict rule is a lottery → · Test your own combination in the Screen Lab →

NCAV = current assets − total liabilities, per share, ÷ price at the filing date. Non-financials only. Point-in-time: each entry is one company bought at the first close after its 10-K filing date, using only filings available that day, held 12 months; result = total return minus SPY over the same months (vs IWM for 12 months; 3-yr annualised). Survivorship-tilted universe; no costs; read spreads, not levels. Educational, hypothetical, before costs.

Open the screen with today's names →Test your own combination
Where it does not apply
Modern net-nets are mostly failing micro-caps, warrants and shells. The screener now hides non-common securities and data mismatches, and shows the median next to the average.

SAVNG is an educational research publication, not a Registered Investment Adviser. Everything here is general-circulation information produced by an automated pipeline — the same for every reader — and is not personalized investment advice, an offer, or a recommendation to buy or sell any security. Any performance figure is hypothetical; past performance does not predict future results. Details.

Published by SAVNG Research. Every figure on this page is read from the study data at render time; if the study changes, the page changes with it.

← All research