Previous Month | RSS/XML | Current
WEBLOG
August 20th, 2026 (Updated: 8/21/2026) (Permalink)
Straw Polls
Something calling itself "Median Strategies" (MS) put out press releases claiming to have done three different polls: one for the Democratic primary for governor of Wisconsin, another for the contest for the governor of Nevada, and finally one for the race for mayor of Los Angeles1. The last one received the most media attention by claiming that Democrat Karen Bass, incumbent mayor of Los Angeles and candidate for re-election, had a twelve percentage point lead over challenger Nithya Raman. Bass' campaign trumpeted the results before it was revealed that all three polls were hoaxes2.
Political partisans frequently claim that polls whose results they dislike are fake, or at least rigged; for instance, President Trump loves the word "fake", calling news he doesn't like "fake news" and poll results he hates "fake polls"3. While there have been "push polls" in the past, which could be rightly called "fake"4, outright hoaxes have been rare.
Since this news may well lead to a decline of trust in polling, I wondered whether any polling had been done to measure the level of trust. Surprisingly, there was such a poll just last week released by a real pollster as opposed to a fraudster: half of respondents said they don't trust polls, with only 28% saying that they did5. According to the pollster, this result has not significantly changed in the last decade, so the pollster has been polling on this topic for at least that long. Assuming that it keeps asking the same question―and I hope it does―it'll be interesting to see whether the current news leads to a loss of trust.
What was the motive for the hoax? MS claimed in its announcement that it was "a short-term social experiment examining how purported polling information could enter and spread through the political information ecosystem without independent verification"6. Every report on this hoax that I've seen seems to take this claim at face value, but that's the same mistake that those who reported on the poll as if it were genuine made. Given that MS has lied about everything else, why should we believe this claim?
Assuming that it's claim is true, this "social experiment" had mixed results. The alleged poll of Los Angeles voters received the most media attention probably because the Bass campaign publicized it. So, there was at least some "spread through the political information ecosystem without independent verification" of that one. This led some news media to try to investigate MS, and that seems to have triggered its public confession and the whole hoax only lasted a few weeks. I'm about as cynical about our "political information ecosystem" as anyone, but this is a reassuring result. It's not as easy as you might have thought to get away with hoaxing poll results.
If this was a "social experiment" by a scholar or scholars, one would expect that those responsible would immediately reveal their identities, and a publication of some sort should ensue. There is a recent history of scholarly hoaxes, starting with the Sokal hoax and followed by the so-called "Sokal squared" hoaxes7, but in both cases those involved revealed their identities and explained the points of the hoaxes.
If it wasn't a "social experiment", what was it? Perhaps it was an attempt to influence the elections by injecting hoaxed polling data that might contribute to a "bandwagon effect"8. This might explain why the fake polling numbers for Bass were so lopsidedly in her favor, while contemporaneous polls had the race between her and her opponent closer. As a result, the "poll" was an outlier, thus drawing attention to the formerly unknown "pollster".
I haven't been able to find any information on the hoaxed Nevada poll, but the poll on the Democratic primary for governor in Wisconsin was similarly skewed, showing Francesca Hong with a 23-point lead in a race that she subsequently lost9.
In its last statement, MS claimed: "No individual involved with Median Strategies received compensation or financial benefit from the project, and the project was not funded, commissioned, or directed by any candidate, campaign, political party, PAC, media organization, or other outside entity." Again, why should we believe this disclaimer? Even if true, it doesn't rule out that an individual or individuals involved in it―and, thus, inside entities―wanted to help one or more of the candidates and tried to do so by means of the hoax.
Unfortunately, we'll probably never know the correct explanation of this hoax unless the party or parties behind it are identified.
Update, 8/21/2026: That was fast! When I wrote the above, whoever was behind "Median Strategies" was unknown, but now the perpetrator, a young man, has confessed to making up the polling results with the help of "artificial intelligence". As in the final statement put out by MS, he claims to have simply wanted to see whether hoaxed polling results would be picked up by the news media. As I pointed out above, the stunt was only partially successful. He also claims, as he did in that statement, that he didn't make any money from it, but why should we believe him? The Grauniad tracked him down based on his domain name registration for MS' website, so he didn't do a very good job covering his tracks10.
Notes:
- ↑ Steve Peoples, Amelia Thomson-Deveaux & Ken Sweet, "What to know about the fake election poll numbers shared and retracted by a mysterious company", Associated Press, 8/18/2026. The following news article is the only one I've found that treated any of the hoaxes as real polls: Justin Choi, "Poll: Spencer Pratt Voters Propel Karen Bass to Double-Digit Lead Over Nithya Raman", California Post, 8/14/2026.
- ↑ Noah Goldberg, "Poll showing Bass leading Raman by double digits was bogus, company says", Los Angeles Times, 8/18/2026.
- ↑ Jesus Mesa, "Donald Trump Slams 'Fake Polls' After Worst-Ever Approval Rating", Newsweak, 1/22/2026.
- ↑ Herbert Asher, Polling and the Public: What Every Citizen Should Know (7th edition, 2007), pp. 156-158.
- ↑ "Only 28% Trust Political Polls", Rasmussen Reports, 8/13/2026. You have to subscribe to read this entire article, but the first couple of paragraphs give the results cited.
- ↑ "Median Strategies Polling Discontinued", Median Strategies, accessed: 8/18/2026.
- ↑ Mike McRae, "A Massive Hoax Involving 20 Fake Culture Studies Papers Just Exploded in Academia", Science Alert, 10/4/2018.
- ↑ Walter Veit, "What Is the Bandwagon Effect?", Psychology Today, 3/14/2020.
- ↑ Sean James, "Polling Company Collapses After Giving Francesca Hong a 23-Point Lead in a Race She Lost, Admits to 'Bogus' Data on LA Mayor Karen Bass", Mediaite, 8/18/2026.
- ↑ Sophie Sullivan, Anna Betts & Sam Levine, "Website behind widely shared fake election polls run by 21-year-old", The Grauniad, 8/20/2026.
August 9th, 2026 (Permalink)
Newsweak Nostradamus
Check out the dueling headlines:
Gas Prices Predicted To Hit Five-Year Low in 20261
Gas Prices on Course To Beat 2022 Record: Which States Are Charging Most?2
Both of these headlines were on articles by the same writer in the same source, Newsweak, about a half-year apart. The first is from early January of this year, prior to the beginning of the current conflict with Iran, and the second is from late last month, months after that conflict had begun. When Newsweak doesn't have any actual news to report, it predicts what the news is going to be.
The actual prognostications reported in these articles are not due to Newsweak itself, but come from the company GasBuddy, which provides information to consumers on gasoline prices3, and when not predicting current prices tries to predict future ones. Such information would be useful if it were correct; for instance, if you knew that the price of gas was going to go up, you could fill up your car prior to an increase, and if you knew it would go down, you could wait until it did to do so.
What's the "five-year low" that the first headline predicted for gas prices this year? According to the article: "Average nationwide gas prices are expected to fall below $3 per gallon this year, according to GasBuddy's predictions―hitting the lowest level since 2020…". According to GasBuddy itself4, the average price of gas for the U.S. was already below $3 per gallon at the time this article was published: it's easy to predict what's already happened.
What's the "2022 Record" that the second headline predicts for the price of gas for the rest of this year? According to the article: "Gas prices in the U.S. are up by more than 30 percent since the start of the Iran war and are likely to reach the highest August level ever next month, according to analysts, surpassing even the record hit in August 2022 when inflation was rampant." So, apparently the record is only for the month of August, which is not nearly as alarming as it would be if it were for the entire year or an all-time record. The nominal price record for the entire year of 2022 was just over five dollars, and the real price record―adjusted for inflation―was set in 20086.
The Newsweak article goes on to say:
Patrick De Haan, head of petroleum analysis at GasBuddy and one of the most prominent experts in the field in the U.S., said on Tuesday that in approximately a week, Americans are "going to see new all-time records set for highest August gas prices ever―even compared to 2022." Considering the national average gas price of $4.08 per gallon then, according to GasBuddy data, De Haan estimated that prices nationwide will reach their "highest level ever for the date" by August 5. In early August 2022, the national average gas price was just above $4; by mid-August, it had fallen to just below $4―the lowest level since early March of that same year.6
According to GasBuddy's own data, the price peaked at about $4.10 a couple of days after this was said, and has been trending downward ever since. On the fifth of this month, the price was $4.05, which might be a penny or two higher than it was on the same date in 2022. Big deal. The current price has already fallen to "just below $4" and we're not even at the middle of the month yet.
The real problem with these prognostications is not that they're wrong, or make a big deal out of nothing much―though they do―but that they simply extrapolate current trends. If the current trend is upward, then they predict that it will keep going up; whereas, if it's currently trending downward, they predict that the trend will continue down. In January, the price of gas was trending downward, so it was predicted to reach a five-year low this year; whereas, since the Iran conflict began, the price has risen, and now it's predicted to beat the 2022 record high. However, the only certainty about current trends is that they won't continue, though when they will change is unpredictable.
I don't blame GasBuddy for not realizing in January of this year that in February we would begin a bombing campaign against Iran that would cause the trend of gas prices to reverse. Perhaps the only person who could have seen that coming was the President, and he may have expected the fighting to end very soon, in which case the impact on gas prices might have been a short spike.
I chose to examine this example not just to pick on Newsweak―that's a bonus―but because it illustrates a larger phenomenon. The future would be predictable if only unpredictable things like wars didn't happen, but things have a way of happening unpredictably. As a result, attempts to predict the future based on current trends are bound to fail eventually. Still, people keep trying: you can count on that.
Notes:
- ↑ Giulia Carbonaro, "Gas Prices Predicted To Hit Five-Year Low in 2026", Newsweek, 1/7/2026.
- ↑ Giulia Carbonaro, "Gas Prices on Course To Beat 2022 Record: Which States Are Charging Most?", Newsweek, 7/30/2026.
- ↑ "About", GasBuddy, accessed: 8/9/2026.
- ↑ "12 Month Average Retail Price Chart", GasBuddy, accessed: 8/9/2026.
- ↑ Inflation and "Record-High" Gas Prices, 6/26/2022.
- ↑ See note 2, above. Paragraphing suppressed.
August 2nd, 2026 (Permalink)
Crack the Binary Combination
In previous "Crack the Combination" puzzles*, the combination has always been decimal, that is, the digits in the combination were the ten from 0 to 9. Now, the combination lock you need to open has a binary combination of five of the two binary digits―"bits"―0 and 1. The following are some incorrect combinations:
| Clue | Combination | # of bits correct |
|---|---|---|
| 1 | 1 0 0 0 0 | 1 |
| 2 | 1 0 0 1 0 | 10 |
| 3 | 0 0 1 1 1 | 11 |
| 4 | 1 0 0 0 1 | 10 |
| 5 | 0 0 0 1 0 | 11 |
Your mission is to determine the correct combination from the above clues.
Compare the first two clues.
01011
Explanation: Here's one way to solve it: Clues 1 and 2 differ by only the fourth bit, and 1 has one correct bit whereas 2 has two. Thus, the fourth bit of 2, which is a one, must be correct. So, the combination is: _ _ _ 1 _. Similarly, clues 1 and 4 differ only by the last bit, while 1 has only one correct bit and 4 has two. Hence, the last bit of 4, which is a one, must be correct. Therefore, the correct combination is: _ _ _ 1 1. This leaves only the first three digits to be determined, and there are only eight possible permutations of three bits. However, clue 5 tells us that there's only one 1 in the first three digits, since two of its first three bits must be correct for it to have three correct bits. Thus, there are only three possibilities for the first three digits of the combination: 1 0 0, 0 1 0, and 0 0 1. The first trio is ruled out by clues 2 and 4, both of which would have four digits correct instead of only two. The last trio is eliminated by clue 3, which would be entirely correct. This leaves the middle trio. Therefore, the full combination is: 0 1 0 1 1.
* ↑ I, II, III, IV, V, VI, VII, VIII, IX, X, XI.