Simulate, calculate and forward-test the Roundhill XX ETF (S&P 500® Target 10,000 2030) and other defined-target call ETFs.
Open the browser simulator: no install, and nothing you type leaves your browser.
These tools price, forward-test and simulate defined-target call ETFs. The first fund is XX, which targets the S&P 500 at 10,000 by 2030. No issuer sponsors this project. It gives no investment advice.
This project provides general information only. Roundhill Investments and S&P Dow Jones Indices do not sponsor, endorse or affiliate with it. S&P Dow Jones Indices owns the "S&P 500" trademark. The fund names and tickers belong to their owners. Nothing here recommends a purchase or a sale. It ignores your objectives, finances and needs. Read the fund's prospectus before you invest.
A defined-target call ETF spends almost all its money on one batch of long-dated call options. The options share one strike and one expiry. The strike sets a "target" level of an index: for XX, the S&P 500 at 10,000. The expiry sets the "target date": January 10, 2030.
At the target date each share is worth about (options per share) × max(ETF price − strike, 0), plus a small cash balance. Below the target, the holder loses almost everything. Above it, gains grow point for point.
Before the target date the share acts as a leveraged bet. Its leverage changes every day. Volatility and interest rates move its price.
When the date arrives the fund rolls into a new target and date.
| Leverage | Can it go to ~zero? | Expiry | |
|---|---|---|---|
| Daily-reset leveraged ETF (e.g., a 3x S&P fund) | Fixed (3x), resets daily | Rarely; it decays when markets chop | None |
| Buffer / defined-outcome ETF | Below 1x, upside capped | No | Yearly outcome periods |
| DIY LEAPS call | You choose | Yes | Yes |
| Defined-target call ETF (XX) | Changes daily (~5.6x in Oct 2026) | Yes, below the target | Yes, then rolls |
Investors may treat XX as a lottery ticket: a small sum that could return five or ten times its cost. Option prices describe a different instrument. At the 2026-10-13 price, XX functions mainly as a leveraged bet that the S&P 500 grows about 10% a year until January 2030, and the bet carries about a 61% risk-neutral chance of a near-total loss.
| Outcome for a buyer at NAV on 2026-10-13 | Required S&P 500 price growth per year | Risk-neutral odds in listed option prices |
|---|---|---|
| Options expire worthless; about 1% cash remains | below 7.9% | about 61% |
| Recovers the purchase price | 9.8% or more | about 29% |
| Doubles | 11.6% or more | about 21% |
| Returns 5× | 16.6% or more | about 7% (3-11% across the bid/ask quotes) |
| Returns 10× | 24.1% or more | about 1% (extrapolated beyond the last listed strike) |
The 5× and 10× outcomes require the S&P 500 to compound at 17-24% a year for more than three years. Listed option prices assign them odds of about 7% and 1%, but wide quotes leave these two rows uncertain: they do not differ reliably from the simple model's 7.9% and 1.8%. Risk-neutral odds measure prices, not forecasts: if stocks outperform cash, the real-world odds run higher. The figures come from one snapshot of delayed quotes (F10). A later study measures how often the S&P 500 has delivered such growth.
The code here computes each fact from the issuer's public daily files and public market data. FINDINGS.md records each method. The figures cover the first closes after launch (2026-09-30 to 2026-10-13). The daily report updates them.
- Each XX share holds call options on 0.48 SPY shares (strike $995.01, expiration 2030-01-10) plus about 1% cash. New money buys options in proportion, so fund flows do not change a buyer's exposure per share. (F1, F3)
- The option's price implies a volatility of about 16.5-16.9%. The pricing vendor's interest rate drives most of the uncertainty in that number: +0.5 points of rate ≈ −0.8 points of volatility. (F6, F9)
- Leverage runs about 5.6×: a 1% SPY move moves NAV about 5.6%. Volatility counts as much: one volatility point moves NAV about 10%, and one percentage point of interest rates moves it about 15%. (F6)
- The finish line sits near S&P 9,995, not exactly 10,000, because SPY's price does not equal S&P/10. The range spans about 9,955-10,040, depending on how much history the estimate uses. (F7)
- A buyer at NAV on 2026-10-13 needs the S&P near 10,560 on the target date to recover the purchase price. That takes about 9.8% price growth a year; reaching the strike alone takes 7.9%. The report's simple odds model uses one volatility for every outcome and puts the risk-neutral chance of finishing above the strike near 29%. Its "10% mean total return" row still leaves the options worthless 52% of the time: a mean is not the typical path, and that row's median path grows only 7.4% a year in price. The options become more likely than not to pay off only near a 12% mean total return (median price growth 9.4%). (daily report, F6)
- The listed S&P 500 options price the odds differently. On 2026-10-13, call spreads on the Dec-2029 calls put the risk-neutral chance of finishing above 10,000 near 39% (simple model: 29%) and the chance of breaking even near 29% (simple model: 23%). The volatility skew causes the gap: it moves probability toward modest wins. For 5× and 10× outcomes the quotes leave too much uncertainty to separate the two models. One snapshot of delayed quotes. (F10)
- On 2026-10-13, XX fell 2.3% while SPY held flat. The day before was a creation day, when the fund's mark comes from its own purchase price; the mark fell back the next day (hypothesis H1 tests this pattern). The daily report books such a move as a fall in implied volatility (96% of it), because the attribution assigns any repricing of the option to volatility. (daily report)
- Speed: one CPU core of an Intel Core i7-13700K (16 cores, 24 threads) prices 1 million options in 0.033 s and solves 1 million implied volatilities in 0.34 s (median of 7 runs; single-threaded numba 0.68, Python 3.13). The Monte Carlo engine (v0.3) uses all 24 threads and, optionally, an NVIDIA GPU: 1 million S&P (SPY) paths over 814 real sessions take about 1.4 s on the CPU and 0.6 s on an RTX 3080 (lognormal), 0.65 s and 0.04 s (bootstrap) (FINDINGS F16). The XX layer, which reprices the fund on every session of every path, runs on the CPU only: about 10 s per million paths.
| Tool | Command (from the repository folder) |
|---|---|
| Browser simulator (v0.2): a date × S&P grid of one share's P/L, value or leverage, with the odds of finishing above each level, volatility scenarios, and presets for XX, a listed S&P 500 call or any fund you describe. No install; nothing you type leaves your browser. | open it online, or open web/index.html from a copy (how to read it); the self-test checks its maths against the Python engine |
| "What must the S&P do?" calculator: computes break-evens, a payoff table and odds across growth assumptions. Runs without downloaded data. | py -m definedtarget calc --price 25 --spot 760 --iv 0.17 --rate 0.048 --as-of 2026-10-01 |
| Daily tracker / forward test: downloads the issuer's public files, checks the accounting, prices the option and writes a one-page report. A new copy prices each close from the first day it runs; to price earlier closes, add SPY's past closes (data/README.md) | py -m definedtarget daily then py -m definedtarget report |
| Prospectus illustration: recomputes the prospectus table | py -m definedtarget prospectus |
| Pre-registered predictions: a hash-chained ledger of testable claims about XX's behavior | forward/HYPOTHESES.md |
| Verdicts: computes each prediction's verdict from its frozen rule (provisional numbers before its review date); records a verdict only when asked | py -I scripts/verdicts.py |
| Monte Carlo engine (v0.3): seeded paths on the real trading calendar (lognormal or blocks of real S&P 500 history), on the CPU or an optional GPU, with error bars, exact-answer tests, no look-ahead and a run manifest for every study. It imports nothing XX-specific. Its interface stays provisional until v0.3.3: names and arguments may change. | definedtarget.montecarlo (how it works and how to use it) |
| XX path studies (v0.3): what a holder lives through path by path (drawdowns, time below the price paid, the odds of reaching 2x / 5x / 10x), the volatility rule on real paths since 2007, and history's shape against the bell curve | py -I scripts/xx_paths_study.py, slide_rule_test.py, effective_leverage_study.py, bootstrap_study.py (FINDINGS F15-F19) |
| Monthly forecast check (v0.3, protocol FC1): the engine forecasts XX one month ahead under four models; a month later the log records where the real outcome landed | py -I scripts/forecast_check.py due, then forecast; a month later outcome --no-events after reviewing the issuer's notices (or --void "..."); report when due lists one |
| Plain-English explainers | docs/learn/ |
Install Python 3.10+ with numpy, pandas, numba and scipy (scipy runs the verdicts' Welch tests; openpyxl is needed only to read saved Cboe FLEX reports). The Monte Carlo engine's GPU path is optional: it needs CuPy and an NVIDIA card, and everything runs without it.
pip install numpy pandas numba scipy
py -m definedtarget calc --price 25 --spot 760 --iv 0.17 --rate 0.048 --as-of 2026-10-01
py tests/run_tests.py
For the daily tracker, read data/README.md. The scripts download the issuer's public files and
public market data into your own data/ folder. Run every script that reads downloaded files with py -I
(isolated mode), e.g., py -I scripts/daily.py.
To install the package itself, run pip install . from the repository folder (pip install .[flex] adds
openpyxl). The definedtarget calc and
prospectus commands then work from any folder; daily and report read and write the repository's data/ and
forward/ folders, so run them from a copy of the repository. CHANGELOG.md lists each release;
LICENSE gives the MIT terms.
The predictions mean something only if they existed before the data that tests them. The first 21 lines of
forward/predictions.jsonl hold protocol P1 and the 20 hypotheses, frozen on 2026-10-14 (each entry records its
UTC time).
OpenTimestamps anchors the SHA-256 of those exact bytes, bc2b73a352bb46ebebf184ee4fdbbb13a7576c22fbba259e897d9c96d0e8888b,
in the Bitcoin blockchain. To check it:
- Copy the first 21 lines, byte for byte, into a file named
predictions_2026-10-14.jsonl:py -c "d=open('forward/predictions.jsonl','rb').readlines()[:21]; open('predictions_2026-10-14.jsonl','wb').write(b''.join(d))" - Verify that file against
forward/predictions_2026-10-14.jsonl.otswith the OpenTimestamps client (ots verify) or on opentimestamps.org. The proof shows the date of the Bitcoin block.
Protocol P2 (clarifications that change no band) follows as line 22, appended on 2026-10-15; the ledger records the
append before the first capture of a test-period close. A second proof, forward/predictions_2026-10-15.jsonl.ots,
covers the first 22 lines (SHA-256 a213dfabc265a43a3a88e8b5e96529dc402df60675fb8316d22b6127d45f339e): repeat the
steps with [:22] and the name predictions_2026-10-15.jsonl. That proof was stamped later the same day, so it
shows that P2 existed by then, not that it came before the capture. Protocol FC1 (the monthly forecast check)
follows as line 23, appended on 2026-10-17 (UTC), before its first origin close, 2026-10-30; the repository's
public history dates it. Verdicts append below and never change earlier lines. The test
tests/test_ledger.py pins both sets of bytes, and .gitattributes stops git from changing their line endings.
- Option value: Black-76 (the Black-Scholes formula written with the forward price) prices the SPY option. It projects SPY's quarterly dividends as discrete payments and discounts with the US Treasury curve, bootstrapped from par yields. The report shows a financing spread over Treasuries as a sensitivity.
- Implied volatility: the solver backs it out of the issuer's daily option mark. The code scales the Greeks to one fund share.
- Odds: a lognormal model at the option's implied volatility computes them twice: under the market's own (risk-neutral) pricing and under several assumed future returns. The report shows them side by side. The model ignores the volatility skew; F10 compares it with the odds in listed S&P 500 option prices.
- Checks: the code rebuilds each day's NAV from the holdings. A broken identity, a stale input, a failed latest-day download or a changed fund document stops the run; a revised holdings file or a close without an SPY price adds a note. No failure passes silently.
ASSUMPTIONS.md and FUND_FACTS.md list every assumption and its status.
- Whether the S&P 500 closes above 10,000 on January 10, 2030.
- The prices the fund pays and receives in trades (the issuer publishes only its model marks).
- The roll rule for the next target period (unpublished).
- Your tax position.
- This repository contains no data. The scripts download each user's own copy.
- Fund files: Roundhill Investments' public website (informational use; republishing requires their permission).
- Volatility indices: Cboe. Interest rates: US Treasury (public domain). SPY dividends and past closes: files each user adds from a source they may use (data/README.md). F10's listed-option quotes: the author's TradingView account (display-only license; the repository holds only derived figures).
- Code: MIT license, provided "as is", without warranty.
Open an issue to report a bug, question a number or suggest a feature. Corrections to the findings matter most: each one comes with its method, so anyone can check it.
If you use these tools or findings, please cite:
ossilon (2026). defined-target-lab: open tools to price, forward-test and simulate defined-target call ETFs, version 0.1.0. https://github.com/ossilon/defined-target-lab
GitHub's "Cite this repository" button gives the same reference in other formats (CITATION.cff).