# FTDNA Discover ancient register — replication analysis (2026-08-18) Register: 362 ancient Z93 males with a usable date, classified into arms by subtree membership in the FTDNA haplotree (Y3 = below F2597; Z2124 = below Z2124; else basal/other). Pipeline disjoint from AADR v66. ## 1. Era x arm counts (register) basal/other Z93-Z94 Y3 arm Z2124 arm 1 pre-Sintashta 22 0 2 2 Sintashta-Andronovo 10 2 59 3 LBA-Iron 2 2 24 4 Scythian-Xiongnu 19 9 97 5 Migration-era 150-950CE 16 8 65 6 Medieval+ 4 1 20 Z2124-arm share by era: 1 pre-Sintashta 2 Sintashta-Andronovo 3 LBA-Iron 0.083 0.831 0.857 4 Scythian-Xiongnu 5 Migration-era 150-950CE 6 Medieval+ 0.776 0.730 0.800 AADR v66 audit shares for comparison (published): 0 / .51 / .54 / .67 / .70 / — ## 2. Temporal trend, register only logistic Z2124 ~ kyr-BP: coef = -0.281 p = 0.0116 (negative coef = Z2124 share RISES toward the present; same caveats as the audit: cemetery patrilines violate independence, era bins are choices.) ## 3. Y3-arm ancients in the register code year country I13265 -1795 b8-2 -1760 Russian Federation I20330 -1325 I23400 -900 I3774 -641 I26241 -500 DS21 -375 Russian Federation DA136 -350 Russian Federation I10517 -336 DS90 -300 Russian Federation DS30 -300 Russian Federation I8395 -275 C3316 -77 China I33986 250 ALN005 330 Kyrgyzstan Hegyko_20 525 Hungary I6942 832 India I19106 900 Russia (Republic of Tatarstan) I19107 900 Russia (Republic of Tatarstan) I19109 900 Russia (Republic of Tatarstan) I25489 900 Russia (Republic of Tatarstan) I19420 1600 Kenya study Akbari et al. 2026 Blöcher et al. 2023 Akbari et al. 2026 Akbari et al. 2026 Akbari et al. 2026 Akbari et al. 2026 Andreeva et al. 2025 Damgaard, Marchi et al. 2018 (Nature) Akbari et al. 2026 Andreeva et al. 2025 Andreeva et al. 2025 Akbari et al. 2026 Kumar et al. 2022 Akbari et al. 2026 Gnecchi-Ruscone et al. 2021 Y. Tian et al. 2026 Harney et al. 2019 Gyuris et al. 2025 Gyuris et al. 2025 Gyuris et al. 2025 Gyuris et al. 2025 Brielle et al. 2023 Pre-CE Y3-arm men: 13 of 22 ## 4. Modern inversion test (FTDNA haplotree census) Y3 arm (toward L657) Z2124 arm SouthAsia 199 65 EasternEurope 11 759 Fisher exact, South Asia vs Eastern Europe, Y3 vs Z2124: odds ratio = 208.7 p = 5.7e-139 (self-reported tester countries; ~63% of kits report no country) ## 5. Pattern-mining appendix (adversarial panel + verification by execution) Four derived analyses were reviewed by a two-model panel (archaeology, adversarial statistics); every claim below was then checked by direct computation. Register cleaned to 363 rows (denisova8, an archaic reference individual at haplogroup A0000 swept in from FTDNA's deep panels, removed). **Certified — Bronze Age radiation burst.** 65 of 106 dated branch formations fall in 2500-1500 BCE; the rate then collapses in stages — 16 in 1500-500 BCE (most before 1200 BCE) and 4 in the millennium 1000 BCE-1 CE. Triangulated three ways: the formation spike, the ancient-sample mass in the same bins, and near-zero attestation lags at the root (Z93 oldest sample 2718 BCE vs TMRCA 2660 BCE; Z2124 2708 vs 2519 — stem placements, expected, not dating error). Ascertainment check: median terminal kit count is identical (1.5) for pre- and post-1500-BCE formations, so the recent-bin depletion is not explained by a kit-count threshold; a formal Yule-null test with size-biased sampling remains open. **Descriptive only — the longitudinal wave.** Median longitude of located ancients runs 40E (3000-2500 BCE) -> 74-75E (2000-500 BCE) -> 53E (0 CE) -> 20E (500-1000 CE) at near-constant 46-56N latitude. Robust dispersion shows the bins are dominated by mega-cemetery clusters (MAD 0.8 deg when a single Xiongnu-era cemetery dominates; 1.9-3.2 deg when Hungary dominates), and the 201 located samples exclude the anonymized Akbari release. Without an opportunity-set null (all published aDNA per period), the wave describes where excavation happened as much as where lineages moved. Within-clade check: only two lineages have >=4 located samples spanning >=1000 years (R-YP1547, R-YP1548); both move west (-2.7 and -5.7 deg/kyr) — consistent with, but far from proving, real westward movement in the Iron-to-medieval window. **Not comparable as computed — entropy ranks.** Shannon entropy over tester countries is biased by kit count and country-count ceiling (Z2125: 875 kits, 83 countries vs YP413: 80 kits, 23 countries). Rarefaction to equal draws is required before ranking; raw ranks are not published. **Quantified — attestation lag.** Oldest-sample-minus-TMRCA for 35 clades: L657 +1,906 y (coalescence 1983 BCE, first physical attestation 77 BCE, a candidate call); the largest lags (Y2568 +3,094 y at the Swahili coast, YP1451 +3,101 y) mark lineages whose ancient record is entirely missing. Negative lags (max -189 y) are within combined dating uncertainty and mark stem placements.