# bitcoin-options-implied-interest-rate: findings (2026-10-24) The six checklist lines (37 to 42; en, ko, tr, ru, vi, de) are one claim: the three PICTURE alts 9, 13 and 25 joined together. The article and the manuscript (`originals/BLOG_implied-rate_1012.txt`) match word for word. Reproduce with `recheck.py`; raw re-queries are in `requery/`. ## Checklist verdicts: all six lines PARTIAL - **Alt 9 (parity fit): numbers verified.** The legend shows "rate 5.59%, forward $84,584" and "58 two-sided strike pairs". The fitted line DF·(F−K), with T = 73.9738/365, gives +63,856 at K = 20,000 and −153,665 at K = 240,000. The plotted endpoints are +63,849 and −153,705, so the gap is at most $41, which is residual-sized. 229.96 = 459.92/2 (block 32). The largest residual of 192.78 agrees with the roughly +$200 point at K = 220,000, but the chart only shows it approximately. Alt unchanged. - **Alt 13 (box spreads): numbers verified, one gap.** The chart annotation prints 18.2, 154.4, 337.2, 289.3, 77.9 and 44.3. The 16 October box (−12.6 to +24.1) reads off the axis, and every n label (16, 121, 1,570, 808) is on the image. 1,653 = C(58,2), and every n is no more than C(strikes, 2). The 24 September 2027 range of 5.16 to 5.58 is too thin to read, and the body does not state it, so it is kept as published. The alt left out the whiskers (10th to 90th percentile), which are now added. - **Alt 25 (perp funding): matches image, body and manuscript, but cannot be reproduced.** The values are Athenum feed readings. I re-queried each venue's own published funding history for 08:00 UTC, annualised: Binance 2.04 (published 2.55), Bybit 6.25 (4.47), OKX 2.73 (3.07), Bitget 2.41 (2.83), Hyperliquid 10.11 (10.44), and Deribit's trailing interest_8h 9.41 (7.99), which is not a settlement. The venue ranking is the same and the spread is 8.07 against the published 7.89. The alt is now scoped as "feed readings that differ from the venues' own published funding history for 08:00 UTC", and "preceding" becomes "past" because the 168-bar window (2026-10-05 09:00 to 2026-10-12 08:00) includes the 08:00 bar. - **Non-English lines 38 to 42:** all three alts are still in English. Localise them from `spec-en.json`. ## Sibling facts checked - **Inverse parity:** handled correctly. In coin terms, C−P = e^{−qT} − K·e^{−rT}/S. Multiplying by the index gives DF·(F−K), so the slope is the dollar discount factor and q sits only in the intercept (blocks 7, 8.6 and 21 are right). The lease rate of −0.06 at 74 days follows from the forward 84,584: q = 5.5851 − 5.6449. - **Day count and compounding:** the compounding is stated (continuous). The day count was not. The published 5.5851 and 5.6168 come back only with actual time from the 08:37:43 UTC reading to the 08:00 UTC expiry over 365 (74/365 would give 5.6148). Added to block 8 step 4. The Treasury conversions check out: 4.05 on a discount basis gives 4.127, 4.44 as a bond-equivalent yield gives 4.391, so the gaps of 1.46 and 1.03 stand. - **Mark vs mid:** kept separate throughout (mid fit; mark fit 5.5885; futures mid 5.6168 against mark 5.6249). Deribit trades at 08:37:35 to 08:37:41 (history.deribit.com) show an index of 83,626 to 83,629 and a 25DEC26 mark of 84,576.28. That agrees with the published index of 83,621.84 and future mid of 84,573.75. Deribit option tickers still return `interest_rate` 0.0. - **37 free tools:** confirmed. `apps/marketing/lib/tools.ts` lists 37 entries, 35 of them calculators by slug, and the file has been unchanged since 2026-09-27. ## Other substantiated defects (English indices) - **Block 12:** "the least squares slope is exactly the squared strike distance weighted mean of those same pairwise boxes". The identity holds for the fitted discount factor (minus the slope) and the pairwise box discount factors over all 1,653 pairs. It does not hold for the 1,570 filtered rates. "uses all of them at once" now reads "fits all 58 strikes at once". Corrected. - **Block 28:** "Hyperliquid printed within 0.22 points of 10.95 in 77 of those hours". Hyperliquid's public hourly history has 86 of the 168 hours at exactly 10.95. The 77, 54 and 4 can only be feed readings, because an 8h venue settles just 21 to 22 times a week. Reworded to keep the feed's counts and add the venue's 86. "Touches exactly zero" for Deribit is verified: interest_1h is 0.0 in 17 hours. - **Block 31:** "Deribit is the crypto native options venue whose entire board is readable without a key" overclaims. Bybit (684) and OKX (786) BTC option tickers are also key-less, so "the" becomes "a". "The box spread literature makes the same point … for exactly that reason" is a misattribution: van Binsbergen, Diamond and Grotteria (JFE 2022) find box rates about 40 basis points above Treasuries and put that gap down to the Treasury convenience yield, not to venue credit. Reworded. ## Not changed, noted - **Block 17:** quote fit and futures basis "agree to four decimal places" at 5.5851. The futures side recomputes exactly. The quote fit cannot be re-run, and an exact match is unlikely when the mark fit (which tracks the future by construction) is 0.0034 off. Flagged as unverifiable. - **Blocks 0, 3, 20, 21:** the headline intervals are the plain OLS ones, which block 27 itself calls less honest than the robust one (±0.21 at 74 days). The 74 day non-overlap survives with the robust interval. The other expiries cannot be checked. - **Block 27:** the feed's weekly means (Binance 0.48, Bitget 6.39) are not reproducible. Settled-rate means over the same window (2026-10-05 09:00 < t <= 2026-10-12 08:00, 21 settlements) are 0.43 and 5.78. ## Structure All six locales have 35 elements with identical tag and picture order, and one table each. English indices map 1:1, so no `indexmap.json` is needed. The cover image (term chart) has its own alt on the page and is outside the spec schema; its numbers match the image. No chart image shows a wrong number, so there is no `figures.txt`.