About the data
What the numbers are
All prices come from Average Price Data, which the U.S. Bureau of Labor Statistics (BLS) publishes each month next to the Consumer Price Index. BLS describes them as estimates of the average price paid by consumers, for narrowly defined items such as a dozen large Grade A eggs or a pound of white bread. Food prices exclude sales tax. Every question here uses the U.S. city average.
BLS publishes monthly figures only. A year in this game is the plain average of the twelve monthly figures for that year, and it is used only when all twelve months exist. “August 2026” is the latest month BLS had published when the question set was built.
The seventeen items
| Item and BLS series | Years used | Aug 2026 |
|---|---|---|
| Large Grade A eggs (dozen) APU0000708111 | 1990–2024 | yes |
| Whole milk (gallon) APU0000709112 | 1996–2024 | yes |
| White bread (1 lb) APU0000702111 | 1990–2024 | yes |
| Sliced bacon (1 lb) APU0000704111 | 1990–2024 | yes |
| Whole fresh chicken (1 lb) APU0000706111 | 1990–2024 (not 2020) | yes |
| Bone-in chicken legs (1 lb) APU0000706212 | 1990–2024 | yes |
| Boneless chicken breast (1 lb) APU0000FF1101 | 2006–2024 | yes |
| White sugar (1 lb) APU0000715211 | 1990–2024 | yes |
| Ground coffee (1 lb) APU0000717311 | 1990–2024 (not 2007, 2008, 2009, 2018, 2019) | yes |
| Long-grain white rice (1 lb) APU0000701312 | 1990–2024 (not 2000, 2001, 2002, 2020) | yes |
| Chocolate chip cookies (1 lb) APU0000702421 | 1990–2024 | yes |
| American cheese (1 lb) APU0000710211 | 1991–2024 (not 2018) | yes |
| Bananas (1 lb) APU0000711211 | 1990–2024 | yes |
| White potatoes (1 lb) APU0000712112 | 1990–2024 (not 2020) | yes |
| Dried beans (1 lb) APU0000714233 | 1996–2024 (not 2020) | yes |
| Potato chips (16 oz) APU0000718311 | 1990–2024 | yes |
| Regular gasoline (gallon) APU000074714 | 1990–2020 | no |
No 2025 year exists because BLS has no figures for October 2025 (data unavailable due to the 2025 lapse in appropriations). Coffee and rice have other gaps where BLS did not publish a month.
How the items were picked
BLS lists 160 national average price series, counting those it has stopped publishing. A series was kept only if all of these held:
- It runs under the same BLS series ID from 1990 to August 2026, and its definition is a fixed physical unit such as a pound, a dozen or a gallon. BLS documents its definitions in its average price factsheet.
- Nothing in BLS notices says its definition or data source changed in the years used. The notices read were the 2018 average price changes (33 series discontinued, 5 added, no redefinitions listed), the 2024 and 2025 publication changes, and the factsheet on the June 2021 switch to a secondary data source for gasoline.
- Its monthly changes show no step that the matching BLS price index does not also show. The check compares each month’s change in the average price with the change in the BLS index for the same food category. A series fails if any month differs by more than 15% (on a log scale), or by more than 8% in January 2015, April 2018 or June 2021, the months BLS announced changes. The largest difference among the kept series is about 14%. These thresholds are judgment calls, not BLS rules.
- It has twelve complete months in at least 15 years.
Gasoline needs a special rule. From June 2021 BLS calculates the gasoline average from a secondary data source instead of its own collection, and the BLS page on the change does not say whether the two periods are comparable. So regular gasoline appears only for 1990 through 2020, and never for August 2026.
What was left out
| Item | Why it is out |
|---|---|
| All-purpose flour (1 lb) | Its average price fell about 19% in June 2021 while the BLS index for the same category rose. The cause is not documented, so the series is out. |
| Spaghetti and macaroni (1 lb) | Its average price fell about 20% in June 2021 while the BLS index for the same category rose. The cause is not documented, so the series is out. |
| Cheddar cheese (1 lb) | Month-to-month swings of more than 15% beyond the BLS index in 1996 that reverse the next month, and many missing months before 1992. |
| Ice cream (half gallon) | Swings of more than 15% beyond the BLS index in October 2002 and May 2003 that reverse. |
| Ground chuck (1 lb) | All beef is out. Six beef series stepped up together in November 2003, and the ground beef series again in January 2013. BLS documents no cause. |
| Ground beef (1 lb) | All beef is out. See ground chuck. |
| Electricity (kWh) | Two BLS documents describe differently how electricity prices are collected (mail questionnaires to the Department of Energy in one, BLS staff in the other). When it changed is not documented. |
| Utility gas (therm) | Same collection question as electricity. |
| Fuel oil (gallon) | Same collection question as electricity, and very large monthly swings. |
| Gasoline, all types | Regular unleaded is the only gasoline grade used. |
| Table wine | BLS says average prices for items priced every other month, wine and beer among them, left out the off-cycle months before January 2015. |
| Malt beverages | Same as wine: priced every other month before January 2015. |
| Navel oranges | BLS does not publish the average in most off-season months (148 missing months since 1980). |
| Strawberries | Same as navel oranges (104 missing months). |
| Iceberg lettuce | 44 missing months and monthly swings of up to 79%. |
| Field tomatoes | Seasonal swings of up to 88% in a month. |
| Canned corn | 271 missing months. |
The rest of the 160 either stopped being published, begin too late to be useful, or were not examined, such as most cuts of beef and pork.
What the numbers cannot tell you
BLS says average prices are best used to measure the price level in a particular month, and not price change over time. Sample rotation, quality changes and shifting consumer preferences can all move an average, and the index is the right tool for inflation. The eggs series, for example, includes cage-free and organic eggs as well as conventional ones, so a change in what shoppers buy changes the average. The “how many times as much” questions simply divide one yearly average by another. They are not inflation figures.
The items are national averages, so they say nothing about what a particular store charged in a particular city.
How the answers are checked
The question set (995 questions) is built by a script from BLS files. A second script then reads the BLS files again with separate code, recomputes the yearly average and the correct answer for every question, and compares every monthly value used against a separate download from the BLS public API (6,604 values compared, no differences). It also checks answer spacing, rounding, years and series units. The check runs on every rebuild, and a rebuild is needed whenever BLS adds a month.
The scripts and the list of items are in the site’s source repository, which is private for now. If you find a wrong number, write to [email protected] and it will be fixed on this page and in the game.
Source and credit
Source: U.S. Bureau of Labor Statistics, Average Price Data, retrieved 12 October 2026 (series files from download.bls.gov, cross-checked against the BLS public data API). BLS material is in the public domain and BLS asks that it be cited. The rounding, yearly averages, ratios and question wording are this site’s own work, and BLS has not reviewed or endorsed them. The BLS emblem is not used.
Privacy
The game sets no cookies and has no accounts or ads. Your daily results, streak and a first-visit date are kept in your browser’s local storage under the key ep1, and they are never sent to the site. If storage is blocked, the game still works but forgets your streak.
The page sends small counting requests with no identifier: that a game was started or finished, which question was answered right or wrong, which share button was pressed, whether a visit looks like a returning one (a day or a week after the first), and, on a first visit, the kind of site that sent you (search, a social site, a news site or none). The host, Cloudflare, also keeps standard server logs and page-view statistics. The main site’s privacy policy is in Japanese and covers the same host.
Who made this
This is a one-person site run under the pen name Kanemochi no Hito. The game, the data scripts and the checks were built with an AI model (Claude), which wrote most of the code. The numbers were taken from BLS files by script, then checked by script and sampled by eye.