Over the past year, I've become quite invested in Wordle. Actually, Taylor and her family started playing first, and I quickly joined in. Soon enough, I even got my mom hooked on it!

If you're not familiar, Wordle is an engaging daily puzzle where you have six tries to guess a secret five-letter word, guided by helpful color-coded hints.

Throughout the year, I consistently chose the same starter: SLATE.

It felt safe and reliable—I always knew my next steps based on it. However, Taylor experimented with different starters and consistently crushed it.

This made me curious: what are actually the best starting words? After some digging, familiar contenders: CRANE, SLATE, TRACE, CRATE, CARET. I've compiled a comprehensive list below for (both of our) future reference.

Identifying the Optimal Starters

Rather than relying solely on intuition or AI, I wanted to see if a straightforward, "old-school" program could crack this puzzle. First, we needed a suitable dictionary of words. Unix and macOS provide the standard words library, but this includes obscure entries that Wordle may not recognize.

Next, we required an effective scoring method. A straightforward approach is to calculate letter frequency across all words—the more frequently a letter appears, the higher its score.

Quickly, some letters emerged as particularly common:

• Most frequent: S, E, A
• Moderately frequent: O, R, I

Interestingly, four of these letters are vowels: E, A, O, I.

Positional frequency matters too—letters are scored higher if they're common in specific positions within words.

Notably:

• The {S, 4} combination (a trailing "S") dominates, suggesting many plurals.
  • {S, 4}: ____S
• Other frequent positional letters:
  • {A, 1}: _A___
  • {E, 1}: _E___
  • {E, 3}: ___E_
  • {I, 1}: _I___
  • {O, 1}: _O___
  • {S, 0}: S____
  • {T, 0}: T____
  • {U, 1}: _U___
  • {Y, 4}: ____Y

Finally, by combining aggregate letter frequency and positional data, a hybrid scoring system emerged. This method offers a more balanced and nuanced approach, producing unique top words: AEROS, SOARE, REAIS, AROSE, and RAISE.

Even when you feed my script the same 2,309‑word official Wordle answer list that WordleBot uses, our rankings still diverge because of how we each value information: my hybrid metric simply adds up how frequently each letter—and, to a lesser degree (10 % blend), each letter‑in‑position—appears across all answers, then zeroes out any word with duplicate letters on turn one, so high‑coverage vowel‑heavy options like AEROS and SOARE dominate; WordleBot, by contrast, runs a full entropy simulation for every guess and keeps duplicate letters if they shrink the remaining solution space, which is why consonant‑balanced staples like CRANE and SLATE top its chart. In short, we share the same dictionary; the gulf comes from aggregate‑frequency math versus entropy‑driven feedback simulation, plus my harsh repeat‑letter penalty and modest positional weight.

Parting Thoughts

Choosing the ideal Wordle starter is about balancing letter frequency and positional insights. Popular starters like CRANE and SLATE remain consistently strong choices due to their strategic letter placement and high-frequency letters. Meanwhile, hybrid scoring systems, which blend multiple metrics, offer compelling alternatives like SOARE and AEROS, maximizing initial guess effectiveness.

Whether sticking with tried-and-true favorites or exploring data-driven options, the real fun of Wordle lies in its daily puzzle-solving and the friends and family your spend doing it with.

My Words

The words generated by my program rank first by a hybrid metric (10% blend), then positional, then aggregate letter frequencies. The metrics are calculated by a sum of the letter’s value, with the value equaling the number of letter occurrences / total words. Positional does the same over the individual positions.

Top Words

Top recommended words based on expert analysis. Check mark ✓ applies to words that have been Wordle words before.

Source Code

Want the source code? Find it here.

GitHub - IllyaStarikov/starikov.co: Source code from my blog.

Source code from my blog. Contribute to IllyaStarikov/starikov.co development by creating an account on GitHub.

GitHubIllyaStarikov