https://nullberg.com/ The replication-first research institute for systematic investing. Every claim of alpha, dissected. en-us https://nullberg.com/primers/four-verdicts-four-shapes/ https://nullberg.com/primers/four-verdicts-four-shapes/ Sat, 11 Apr 2026 00:00:00 GMT Primer The first primer argued that most published alpha does not replicate. The first four verdicts test that claim empirically on four different factor classes and produce four qualitatively different outcomes. This primer reads the patterns in them, separates what the data shows from what the mechanisms might be, and sets out what the archive will track next. https://nullberg.com/primers/replication-crisis-in-systematic-investing/ https://nullberg.com/primers/replication-crisis-in-systematic-investing/ Sat, 11 Apr 2026 00:00:00 GMT Primer Three published meta-studies that anchor the entire reason Nullberg exists. Exact numbers, no paraphrasing, with links to every source. https://nullberg.com/archive/accruals-out-of-sample-2016-2026/ https://nullberg.com/archive/accruals-out-of-sample-2016-2026/ Sat, 11 Apr 2026 00:00:00 GMT Verdict: failed Sloan's 1996 accruals anomaly claims that low-accruals firms (earnings backed by cash) outperform high-accruals firms by ~10.4% per year. On 123 months of US equity data with cash-flow-statement accruals, the primary VW NYSE-breakpoint spec produces +0.883% per month — essentially exactly the canonical 0.87% monthly magnitude — but the i.i.d. t is +1.85, just below the rubric threshold, and the Newey-West 12-lag t is +1.74. Verdict: FAILED by pre-registered rubric. Headline caveat: two of the three EW sensitivities REPLICATE robustly under both i.i.d. and Newey-West (Sens A: +3.81 NW t, Sens B: +2.39 NW t), implying the accruals effect has migrated to smaller stocks where the VW primary underweights it. This is a new kind of near-miss in the archive: FAILED at the canonical magnitude, with sensitivities that clearly replicate. https://nullberg.com/archive/gross-profitability-out-of-sample-2016-2026/ https://nullberg.com/archive/gross-profitability-out-of-sample-2016-2026/ Sat, 11 Apr 2026 00:00:00 GMT Verdict: failed The original claim is that gross profits-to-assets predicts the cross-section of returns about as strongly as book-to-market does. In 123 months of value-weighted NYSE-breakpoint US equity data, the long-short spread goes in the right direction at 0.503% per month (above Novy-Marx's own canonical 0.31%) but with t = +0.93, below the conventional significance bar. Equal-weighted variants flip sign on smaller stocks. This is the closest of Nullberg's first three verdicts to REPLICATED, but the rubric requires t > 2 so it lands as FAILED with an underpowered-survivor interpretation. Backfill update 2026-04-11: Newey-West 12-lag t is +1.05 (slightly above the i.i.d. t of +0.93, implying mildly negative monthly autocorrelation), and both sub-sample halves are positive and consistent (+0.42% and +0.58%), strengthening the underpowered-but-consistent interpretation. https://nullberg.com/archive/max-factor-out-of-sample-2016-2026/ https://nullberg.com/archive/max-factor-out-of-sample-2016-2026/ Sat, 11 Apr 2026 00:00:00 GMT Verdict: failed The original claim is that the lowest-MAX decile beats the highest-MAX decile by over 1% per month. On 122 months of US equity data, value-weighted with NYSE breakpoints and stocks-only filtering, the spread is not merely absent but inverted: high-MAX beats low-MAX by 1.80% per month with t = -2.57. The finding is robust across three equal-weighted sensitivities with t-statistics from -1.81 to -4.12. Backfill update 2026-04-11 confirms the inversion survives the Newey-West 12-lag HAC adjustment at t = -2.18, and is concentrated in the 2016-2021 first half (t = -2.29) versus the 2021-2026 second half (t = -1.38). https://nullberg.com/archive/momentum-out-of-sample-2017-2026/ https://nullberg.com/archive/momentum-out-of-sample-2017-2026/ Sat, 11 Apr 2026 00:00:00 GMT Verdict: failed The original abstract says past winners beat past losers by a significant margin over three- to twelve-month holding periods. On 111 months of US equity data, value-weighted with NYSE breakpoints and stocks-only filtering, the winners-minus-losers spread is -0.13% per month with t = -0.16. The effect has decayed to zero in the standard spec and only approaches significance in one shorter-lookback equal-weighted sensitivity at t = +1.45. All four specifications fail to reach significance. Backfill update 2026-04-11: Newey-West 12-lag t is -0.26 on the primary, and both sub-sample halves are near-null (+0.02% and -0.25%), confirming the decay is consistent across time rather than regime-driven. https://nullberg.com/archive/value-out-of-sample-2016-2026/ https://nullberg.com/archive/value-out-of-sample-2016-2026/ Sat, 11 Apr 2026 00:00:00 GMT Verdict: replicated First REPLICATED verdict in the Nullberg archive, and also the most heavily caveated. The Fama-French 1993 HML 2x3 construction on 122 months of US equity data delivers +1.698% per month with i.i.d. t = +2.82, clearing the pre-registered rubric. But the effect is regime-driven: the first half of the sample is statistically null (+0.585%, t = +0.73) while the second half is huge (+2.810%, t = +3.20), with 2022 alone averaging +9.24% on an annualized-monthly basis. The Newey-West 12-lag t is +1.94, just below the rubric threshold. The simple D10-D1 sensitivity replicates under both i.i.d. and Newey-West, but the equal-weighted sensitivity is flat at t = +0.35, consistent with value being a value-weighted large-cap phenomenon in this regime.