Eiko performance study — 2 October 2026

Questions
1. Can a depositor understand a historical result more easily as two ending amounts?
2. Can earlier vaults appear as one chronological record without changing the maths?
3. What can a new vault show before it has completed an epoch?

Inputs and scope
The owner supplied 18 same-window (25 September–2 October) strike comparisons.
Mockups use the TIBBIR subset; for example 1,000 USDC from a passive limit order
corresponds to 1,157.79 USDC for the vault at strike 0.26 in that window.
This is normalization to the benchmark outcome, NOT a 1,000-unit initial deposit.
The live screen uses the existing selected-strike completed-epoch calculation.
No table values, 18/18 headline, predicted yield, or wallet profit is hard-coded.
Reference: live Eiko TIBBIR 0.26 on /dev/eiko/app.html, disconnected, 1440x1000,
plus the approved eiko-mascot.png. Built-in imagegen, three separate concept calls.
Exact prompts are in prompts.txt. Generated words/bars are design suggestions only.

Adopt / adapt / reject
A — implement the paired ending amounts and an explicit benchmark scale.
    Remove repeated normalization text. Do not say a user deposited 1,000 or
    personally earned the example. Compute both bars with the same scale:
    benchmark = 1000, vault = 1000*(1+gainBps/10000); normalize widths by the greater value.
    The generated bars were not proportional; they are not copied as geometry.
B — retain as a later common-period strike comparison. Do not average USDC and
    base-token outcomes or assume every vault ran the same dates. New data view needed.
C — retain the distinction between fees building now and completed results.
    Reject repeated headings and nested frames for this minimal implementation.

Implementation
Gyro/Eiko/Meryl/Merry retain independent copies. Earlier-source transitions are
decoded independently, then flattened for presentation; no return bridges vaults.
Sort by recorded endpoint timestamp, not local epoch number. Retain equal-time
records from different sources/series. Current-vault rollup and historical scope
remain separate. Source notes and repeated-roll evidence remain under Sources.
Eiko alone adds performance.js/css. Latest current comparison is open; others
are optional. Negative, zero, tiny and unreadable results have explicit tests.
No new RPC, indexer job, transaction path or cache is introduced by the comparison.

Expectations discussion (current agent assessment, not an external consensus)
Fees so far and completed benchmark-relative results answer different questions.
Vault IV is a strategy input, not an annual yield. A Black–Scholes value can be a
reference price for a specified option and assumptions; it does not by itself
predict this vault's fees, path-dependent outcome or net result versus a limit.
Do not ship it labelled Target return until the mapping to this strategy is
specified and independently tested. A useful future optional label is Option
price reference. Distinguish its period/notional and intrinsic/time value.
For a new vault, show observed fees and time to roll; no history means no
historical performance figure. Do not turn missing history into a zero result.
Primary reference: https://www.optionseducation.org/advancedconcepts/black-scholes-formula
Fee context: https://support.uniswap.org/hc/en-us/articles/20901935681677-What-is-a-liquidity-provider-LP-fee

Opus and Z.ai brainstorming was requested by the owner. Scoped-brief approval
is pending; neither provider has run. This is not an external consensus.

Comprehension checks for a newcomer
Ask: Is 1,000 what you deposited, or the normalized limit-order outcome?
Ask: Does +157.79 describe completed outperformance, fees, or next epoch's yield?
Repeat with a negative result and a TIBBIR-denominated result. A wrong answer
means the labels need work; visual preference alone does not establish clarity.

Verification
- Fresh Astra source review: no remaining blockers. A proof/repeated-only source
  note bug was fixed and covered in all four copies.
- Targeted lineage, normalization, stale-context and return-help tests pass.
  Full four-site test sweep compared with dev aeacf5fe; no new failures remain.
  Existing Eiko failures: test-aave-fork, test-alpha, test-chain-focus,
  test-core-housekeeping, test-m-cutover, test-markets-r2-live (opt-in required),
  test-no-swap-withdraw, test-vault-generation. These remain unresolved.
- Served Eiko desktop 1440x1000 and phone 390x844 checked; also dark mode at
  320px. No horizontal overflow. Browser check found the old 240px inner list
  clipped the comparison; performance.css now allows the list to grow.
- Live TIBBIR 0.26: 1,000.00 USDC vs 1,157.79 USDC, 157.79 more USDC.
  Older comparisons start closed. Switching strike clears the old amounts.
  Empty history does not invent a result from the owner's example table.
- Served Gyro XSGD 0.785: one list spanning two earlier vaults, September 19,
  September 12, September 5, August 29, newest first; four source links kept.
- Served Meryl NVDAc 210: one earlier row, September 19, +0.51%, Sources closed.
  Merry SPY 750: no completed epochs and no earlier-vault panel, as before.
- Console inspected on all four: public RPC retry/throttle responses (429/500)
  and the existing WalletConnect setup warning occurred. Reads recovered;
  no new JavaScript exception was observed. No wallet connected or transaction sent.
- check-wiring.mjs aeacf5fe and git diff --check pass.