A systematic income experiment: worst-of-two autocall notes on the dated S&P 500, priced from point-in-time market data and held to contractual runoff. What follows is the full decade, published as measured — every figure gross of fees.
Each position is a note on the worse performer of two S&P 500 companies. The note pays a fixed coupon while both stocks hold up, redeems early at par once the worse one stands above its starting level on a quarterly check, and returns full capital at six months unless the worse stock has fallen below half its starting value at that final observation. Losses column: how many notes crossed that line, out of the notes that ran their full term to an observation.
United States Dollar
3.14%
annualised, gross
The deepest book: admission requires several independent rules to agree, sizing fills what each week can actually absorb, and the dollar rate era gave the strategy something to buy.
| Year | Gain | Value |
|---|---|---|
| 2016 | 64k | 10.06m |
| 2017 | 93k | 10.16m |
| 2018 | 117k | 10.27m |
| 2019 | 388k | 10.66m |
| 2020 | 74k | 10.74m |
| 2021 | −20k | 10.72m |
| 2022 | 415k | 11.13m |
| 2023 | 687k | 11.82m |
| 2024 | 661k | 12.48m |
| 2025 | 1.13m | 13.61m |
Swedish Krona
1.48%
annualised, gross
Rate-blind gates in a negative-rate decade: the book waited out six lean years, then cleared par in 2023 once higher rates fed the coupon floor.
| Year | Gain | Value |
|---|---|---|
| 2016 | −64k | 9.94m |
| 2017 | −68k | 9.87m |
| 2018 | −70k | 9.80m |
| 2019 | −2k | 9.80m |
| 2020 | 15k | 9.81m |
| 2021 | −39k | 9.77m |
| 2022 | 199k | 9.97m |
| 2023 | 493k | 10.46m |
| 2024 | 638k | 11.10m |
| 2025 | 476k | 11.58m |
Euro
1.39%
annualised, gross
Euro funding was negative into 2022: six lean years priced the cost of waiting before the book cleared par in 2023.
| Year | Gain | Value |
|---|---|---|
| 2016 | −31k | 9.97m |
| 2017 | −35k | 9.93m |
| 2018 | −37k | 9.90m |
| 2019 | −1k | 9.90m |
| 2020 | −24k | 9.87m |
| 2021 | −16k | 9.86m |
| 2022 | 55k | 9.91m |
| 2023 | 383k | 10.29m |
| 2024 | 655k | 10.95m |
| 2025 | 523k | 11.47m |
Swiss Franc
0.55%
annualised, gross
Franc funding rarely cleared the coupon bar: the book held cash for most of the decade, then finally found coupons worth owning once rates turned.
| Year | Gain | Value |
|---|---|---|
| 2016 | −73k | 9.93m |
| 2017 | −72k | 9.85m |
| 2018 | −74k | 9.78m |
| 2019 | −33k | 9.75m |
| 2020 | −39k | 9.71m |
| 2021 | −34k | 9.68m |
| 2022 | 31k | 9.71m |
| 2023 | 360k | 10.07m |
| 2024 | 356k | 10.42m |
| 2025 | 139k | 10.56m |
Two figures deserve a word. A negative year in a book that held mostly cash is not a trading loss: for much of the decade the franc, euro and krona carried negative policy rates, so cash itself paid to wait, and those carrying costs are published like everything else. Losses counts notes that finished below their protection barrier; most notes are called back early at a gain, so the denominator is the few that ran their full term to a barrier observation. Max decline is a weekly model mark on open positions, not a dealer bid; realized losses were zero in every book.
Unlevered, one note at a time, admitted only where the deterministic stack agrees.
Measured risk2/10The engine's admission rules are audited one by one: each must earn its keep in measured risk avoided. The first release - a rally-pattern rule that blocked zero real losses across a decade of evidence - is measured live: more notes admitted, zero barrier losses, the gain currently capped by the book's own position limit.
Measured risk4/10Balance-sheet leverage on the Income+ engine with financing modelled explicitly. Tested at 1.5x and not evidenced: the single-name limit binds long before the leverage does.
Measured risk6/10The deterministic stack keeps admission authority; a language model chooses among what it allows.
Measured risk5/10A model runs the desk from a written market brief, sizing and structuring its own book.
Measured risk8/10The S&P 500 by dated membership from 2016, including companies that later left the index, so nothing is chosen with hindsight. Weekly coverage is bounded by which names carry a priced volatility surface that week.
Daily open, high, low, close and volume for every member, with corporate actions applied as they were traded.
Option-implied volatility surfaces by expiry and strike, read at the decision date. What the options market charged for risk that day, not what volatility turned out to be.
Policy and short-term funding rates in each currency — Federal Reserve, European Central Bank, Swiss National Bank and Riksbank. They set what idle cash earns, and what a note must pay to be worth owning.
The two-year to ten-year spread, as a read on whether the market expects tightening or relief.
The equity volatility index and high-yield credit spreads, used to classify the market regime a decision is made in.
Consumer prices and unemployment are tracked as context. They are revised after publication, so they inform research rather than trade decisions.
Requests for quote from bank counterparties, used to calibrate the pricing model where quote panels exist — calm markets first, with stressed-market panels still being collected.
Measured between the two underlyings of every candidate pair from trailing returns, because a worst-of note is a bet on how two companies move together.
One to ten, scored against the evidence: realised loss experience, proximity to the protection barrier, drawdown, leverage, and how much decision authority is delegated.