Seven Things We Learned from Analyzing 515 Million Words –

Seven Things We Learned from Analyzing 515 Million Words – The New York Times

Millions of people play Wordle every day and share, discuss and debate how they tried to win.

For the first time, we analyzed how players performed in half a billion of these Wordle games last year and compared their results with the strategies recommended by our WordleBot.

Here are seven things we learned:

Many, many words have been written about the best opening word for Wordle. Answering this question was actually one of the motivations for developing WordleBot. In his robot brain a handful of words – SLATE, CRANE, TRACK – receive the bot's seal of approval because, on average, they lead to the solution with the fewest guesses.

But for human Wordle players, the most popular opening word is by far ADIEUwith AUDIOanother four vowel word not far behind.

The strategy seems to make sense: figure out the vowels and the other letters will match. The bot never preferred to start that way, but it's possible that the best strategy for a computer is not the best strategy for a human.

However, our new analysis shows that this is initially the case ADIEU or AUDIO actually puts human players at a disadvantage. How much of one?

On average, players who started with ADIEU took about a third of a round longer to solve their word problems compared to the players who started with SLATEwhich gives 132 additional revolutions over the course of a year.

Even worse: While 1.7 percent of SLATE Starters who fail to solve a wordle on average increase the likelihood of failure ADIEU more than doubled to 3.6 percent.

We can't say that with certainty at first ADIEU makes the players worse off. It could be that players start with ADIEU are more casual gamers or place less value on solving as few puzzles as possible. After all, not everyone values ​​pure efficiency when playing. But we can say that of the 30 most popular Wordle openers ADIEU performed worst.

In general, the average WordleBot user solves a puzzle in about a quarter turn less than the average Wordle player.

The bot starts with SLATE every day, which can be boring. Human players, freed from the burden of mathematical precision, can and do change their Wordle openings throughout the year. (Players often explain their daily picks in the Wordle Review.)

Here are the main opening words that were particularly popular on certain days:

  • Christmas Eve (December 24, 2022): CHEERFUL

  • Christmas Day: CHEERFUL, GIFTS, PEACE

  • New Year's Eve: PARTY, YEARS

  • New Year (January 1, 2023): YEARS, PARTY, HAPPY, FRESH

  • The Super Bowl (February 12): EXCELLENT

  • Valentine's Day: HEART, LOVER, CUPID

  • St. Patrick's Day: HAPPY

  • Easter: RISEN, HARE

  • Coronation of Charles III and Camilla (May 6): CROWN, ROYAL

  • Labor Day: WORK

Another fun pattern: PARTY has a small increase every weekend.

In the list above, we have excluded initial guesses, which were today's Wordle solution.

That's because in about 1 in 250 games, a reader gets the answer right on the first try. This happens much more often than you would expect if you just left things to chance, although not every case is necessarily true fraud.

Some may use a new window to make additional guesses after failing to solve the puzzle in six attempts. Some may re-enter a solution they found on another device to maintain a streak or test a technical issue. Others may have garbled the answer or looked it up.

Here you can see which opening words have become more popular over time. ADIEU remains dominant and its popularity as an opening tip has actually increased.

The increase in popularity through external events – when “ADIEU“was featured in the mini crossword puzzle in August – or from the time a word appeared as a solution – AUDIO on August 30th, CRANE on June 21st – are obvious.

The hardest words to solve began with J, ended with Y, or had a double letter somewhere. The most difficult of all words last year, JAZZY, has all three. Last year's five most difficult words to solve, measured by rounds solved:

  • JAZZY (June 1): 5.82 rounds to solve

  • JOKER (April 25): 5.69 rounds to solve

  • NANNY (June 3): 5.68 rounds to solve

  • JUDGE (Dec 26): 5.57 rounds to solve

  • MATURE (February 22): 5.52 rounds to solve

In terms of solution rate, the Wordle puzzle was the most difficult JOKERwith only 71 percent of players finding the solution.

There has been no solution list for WordleBot to consult for over a year. Instead, the bot assigns a probability of being a solution to nearly every five-letter word in the English language – all 14,855 accepted Wordle guesses. (The bot plays no role in selecting the words that appear as Wordle solutions; if it did, this task would be much easier.)

Most of these words – from AAHED To ZYMIC – are considered too unclear to be a likely solution and are given a probability of zero percent. For the remaining words, the more frequently a word occurs (measured by how often it appears in the New York Times since 2000), the higher the probability that the bot will assign it.

Most of the words that appeared as solutions – all of REMAIN To YOUTH – were given the greatest possible chance. But every now and then a solution comes along that surprises the bot. His solution model is not perfect! This is an area where a smart human player can outperform the bot.

Here are the solutions from last year that surprised the bot the most, along with the bot's estimated likelihood of them appearing compared to the most common words.

  • ANODE (August 5): 67.9%

  • COMPANY (January 22): 73.2%

  • BORAX (April 12): 81.2%

  • GUPPY (May 4): 86.0%

  • DEBUG (February 11): 89.4%

Human players can freely choose one of the approximately 15,000 accepted Wordle words when playing. The bot isn't so lucky.

Deciding which words were common enough for the bot to recommend was a matter of judgment. We want the bot to be as skilled as possible. On the other hand, it wouldn't make sense to have a bot that reprimands players for not guessing words like AWDLS, LAARF or YOKEG.

By incorporating data from players' guesses, we can make more targeted decisions about which words the bot recommends. In the coming days, we'll remove a handful of words from the bot's list and make room for about 50 additions. A new set of scientific terms – MALIC, ZILIA, BORON And NODE – along with many words related to food – PANKO, MOREL, CHILE, UMAMI And BENTO – will come up (again, this has nothing to do with actual Wordle solutions). We assume the bot spent more time reading the Science and NYT Cooking sections.

Note: In the cases where we calculated average guesses, failed puzzles counted towards 7.5 guesses.