# extras

Files made for the website and for the video from the same study results as the main folder. Each
was checked against the main files before it was written; the check is given with each file.
Dollar figures use 1 R = $100 on a $10,000 account.

## coin_runs.csv

Every individual coin run of the coin test: 2,000 runs x 4 versions, with the number of survivors
each run produced in each market and in all seven together. The main folder
(`lucktest_sensitivity_by_symbol.csv`, `lucktest_sensitivity_all_markets.csv`) gives each
version's distribution as a median, percentiles and a p-value; this is the distribution itself.
`version` is as `variant` in `lucktest_sensitivity_settings.csv`; `run = 0` is the real strategy
(the unflipped data), included as a reference row; then `ALL` and one column per market.
**Check:** every observed count, median, 5th and 95th percentile and p-value recomputed from these
runs matches the two files above, for all 4 versions and 8 scopes.

## one_trade_ticks.csv

Every tick (bid and ask, in points of 0.001) of the S&P 500 (US500) from 13:30 to 14:30 UTC on 20
October 2025, around the survivor trade in `tick_vs_ohlc_example.csv`, bought at 13:36.
**Check:** the tick-exact stop-out (13:41:32, bid 6705948, ask 6706455), the 13:41 one-minute mid
and bid bars, and the 14:24 target bar all reproduce `tick_vs_ohlc_example.csv` exactly.

## coin_twin_days.csv

The daily net result and cost, in R, of seven real survivors, for every trading day of rounds two
and three: the strategy whose trade is replayed on the page (US500, ORB-1.17, cell 33,
`r_multiple@1.5`, `ORB-1.17_gap_and_go_q75`) and the best survivor in each of six markets, picked
with hindsight. The page's "flip it yourself" coin test runs on these days.
**Check:** each strategy's days sum to its `val_total_R` and `oos_total_R` in `survivors_all.csv`.
On 1,000,000 seeded coin worlds (seed 20260925) the replayed strategy's coin twin cleared the cost floor in both rounds 1.182% of the time; the six others between 0.250% and 7.687%.

## video_figures/

The figures the video uses that are computed from the results rather than read from one file.
**Check:** every configuration's daily result, summed, equals its round total in
`survivors_all.csv` or `is_clearers_all.csv` (to 1e-6 R). Daily figures are per trading day, so
drawdowns are daily-close figures; within a day a trade-level drawdown can be larger.

- `dream_portfolio.csv`: the strategy at its very best. The best survivor per market, chosen on
  rounds one and two together from the survivors (so it already knows who survived). Six markets:
  gold had no survivors. `dream_portfolio_daily_oos.csv` is its combined daily result over the last
  two years, with the running total.
- `honest_portfolio.csv`: the same idea using only what you could know at the time. The best
  round-one winner per market, chosen on round one alone, in the same six markets.
  `honest_portfolio_daily_val.csv` follows a $10,000 account through round two; `account_usd`
  reaches zero on the day given in the summary.
- `honest_incl_gold_portfolio*.csv`: the same with gold included (seven markets), for comparison.
- `portfolios_summary.csv`: the totals per round, the dream portfolio's worst daily-close dip over
  the last two years, and the honest portfolio's zero date.
- `nasdaq_open_ticks.csv`: price updates in the first minute after the 09:30 New York open, one row
  per trading day (daylight saving handled), counted from the Nasdaq 100 tick data.
  **Check:** the tick count equals NAS100's `n_ticks` in `tick_data_physical.csv`.
- `nasdaq_open_summary.csv`: over the last two years (20 Aug 2024 to 21 Aug 2026) the median first
  minute has 464 updates, the median second 8, and the busiest second 19. Over all 3,247 days since
  2014 the median first minute has 419; activity has grown over the years.
- `eurusd_quiz_math.csv`: the video's "about one in eleven trillion". The naive maths that treats
  every EURUSD version as passing by luck on its own: 35,681 versions, luck's average of 529.842
  survivors (the original coin), a binomial count, and the chance of 707 or more. Beside it, the
  coin's own range, which keeps the shared trades and is more than twenty times wider. Worked from
  `lucktest_flip_by_variation.csv` and `lucktest_population_by_symbol.csv`.