Level 4 is where Streakle runs end, and the maths says so

How we got these numbers

We built a greedy solver that always plays the guess minimising the expected number of surviving candidates, breaking ties toward words that could themselves be the answer. It played all 601 unique words in the live answer pool, twice per opener: once with a four guess budget and once with three. The candidate guess set was the 600 strongest openers plus every remaining live candidate at each step. Word lists were read from the shipped game source on 3 August 2026.

The rule almost nobody notices

Streakle's guess budget is not constant. Levels one, two and three give you four guesses. From level four onward you get three. The rule lives in a single line of the game's source, it is never announced with a banner, and most players discover it by losing to it.

It is also, by a wide margin, the most consequential rule in the game. We wanted to know exactly how consequential, so we built a solver that plays optimally and made it grind through the entire answer pool under both budgets.

What one fewer guess costs

Read that table again with the streak rule in mind. One miss ends your run. If even flawless play fails 22 to 25 percent of level four and higher puzzles, then a long streak is not primarily a test of vocabulary. It is a test of how much information you extract in your first two guesses, plus luck.

This also explains a feeling most regular players have: that runs die in a cluster around levels four to seven rather than tailing off gradually. They do. The cliff is real and it is at level four.

Solve rate is the share of the 601 answers the solver cracked inside the budget.

A perfect solver fails roughly one Streakle puzzle in four once the guess budget drops to three. A human will do worse. Level four is not a difficulty bump, it is a different game.

Why the opener matters far more at three guesses

At four guesses, the gap between the best opener and AUDIO is about two percentage points of solve rate. Almost nothing. At three guesses the same gap is over 16 points. That is the single most useful thing in this article.

The reason is structural. With four guesses you have slack: a mediocre first guess can be repaired by a strong second and you still have two chances to convert. With three, guess one and guess two have to do all the narrowing, and there is no recovery move. A bad opener does not just cost you information, it removes your ability to survive having made it.

So if you only change one habit, change it for the back half of your run, not the front. The opener you use at level one barely matters. The opener you use at level six decides whether the run continues.

How the solver actually spends its guesses

Guess three is where most puzzles are won: 63 percent of them. Guess four accounts for another 23.5 percent, and that entire slice is exactly what disappears at level four. The 141 answers the solver needed a fourth guess for are the puzzles that end runs.

Notice how few solves land on guess two. Only 12 percent of answers can be pinned down after a single opener, no matter how good the opener is. Trying to solve on guess two is almost always the wrong instinct. Spend guess two widening coverage and convert on guess three.

Playing the back half of a run

• Treat level four as a hard mode switch. Slow down. The clock is not the constraint, the guess budget is.
• Do not attempt a solve on guess two unless the surviving candidate list is genuinely down to one or two words. A speculative guess two costs you the whole run about a quarter of the time.
• Use guess two to test letters you have not touched, even if that means guessing a word you know is wrong. With three guesses this is counterintuitive and correct.
• Repeated letters are concentrated in the higher levels by design: the game's difficulty score adds 2.4 points per duplicated letter, more than any rare letter except X, J, Q and Z. If your normal approach stalls at level five or higher, test a double before you test another new consonant.
• The 15 hardest words in the pool by the game's own scoring are JAZZY, FUZZY, BUZZY, FIZZY, DIZZY, JIFFY, BOBBY, YUMMY, FLUFF, ZIPPY, PUPPY and KAYAK. Every single one has a repeat. That is not a coincidence, it is the scoring function.

Is the wall too steep?

We are publishing a number that does not flatter our own game, so it is fair to say what we make of it. A 77 percent per-puzzle ceiling under perfect play means the expected streak for a flawless player from level four is roughly four more solves. That is a tight, high-stakes design, and it is deliberate: a streak that anyone can extend indefinitely is not a streak.

It does mean the leaderboard is measuring something closer to sustained accuracy under pressure than raw vocabulary, and we think that is the more interesting thing to measure. But if you have been feeling that your runs end abruptly and unfairly around level four to six, you were reading the game correctly.

Frequently asked questions

Exactly when does Streakle drop to three guesses?

At level four. Levels one to three give you four guesses, level four onward gives you three. Your level equals the number of puzzles you have already solved in the current run plus one.

Does the word itself get harder as well?

Yes. Streakle sorts the whole answer pool by a difficulty score built from letter rarity, repeated letters and vowel count, then draws each level from a progressively harder band. So level six hands you both a harder word and one fewer guess.

Can a human realistically match the solver?

No. The solver evaluates every legal guess against every surviving candidate at each step. The useful comparison is that the solver's 77.5 percent is a ceiling, and the gap between your rate and that number is the part you can train.

What is the single highest value habit change?

Stop trying to solve on guess two. In our simulation only 12 percent of answers were pinnable after one opener, so a guess two solve attempt is usually a coin flip that costs the run.