What will your opening rack look like?
Seven tiles out of a full 100-tile bag. There are 16,007,560,800 distinct ways that can land, and every figure below counts them exactly rather than dealing a few million practice racks.
Face value only, before any board multiplier — and blanks count zero, which is exactly why the tile you most want to draw drags this number down.
The exact face value of an opening rack. The highlighted bar is the likeliest value, not the best one — a high-value rack is usually a hard rack to play.
Vowels and consonants
The bag holds 42 vowels and 56 consonants, so a seven-tile rack averages three vowels — which is also the split that scores best. Getting one or none happens 12.50% of the time — about one opening rack in 8. A truly vowelless rack, counting neither a blank nor a Y as a way out, is rarer still at 1.11%.
Vowels in your rack
Number of A, E, I, O or U tiles. Blanks count as neither.
Blanks in your rack
Only two blanks exist in the whole bag.
The racks people complain about
| Rack | How often | About one in |
|---|---|---|
No vowel at all Seven tiles and not one A, E, I, O or U. A blank counts as neither here -- it can become a vowel. | 1.9% | 53 |
No vowel, no blank and no Y What a player actually means by a vowelless rack: no blank to spend and no Y to lean on either. | 1.1% | 90 |
Exactly one vowel Playable, usually, but rarely well. | 10.6% | 9 |
Five or more vowels The other way a rack goes wrong, and it happens more often than an all-consonant one. | 10.9% | 9 |
The Q with no U and no blank The single most complained-about rack in the game -- and mostly a false alarm, see below. | 4.8% | 21 |
The Q with no U, but an I QI is good in both official word lists, so an I rescues the rack on its own -- though not in ENABLE, which is the public-domain list the two-letter word pages use and one of eleven two-letter words it lacks. | 2.2% | 45 |
The Q with no U, no blank, no I and no A The genuinely awkward case, once QI and QAT are accounted for. Much rarer than the folklore. | 1.3% | 79 |
At least one blank Two blanks in a 100-tile bag, and you see one in about one opening rack in seven. | 13.6% | 7 |
Both blanks at once Rare enough to remember. | 0.42% | 236 |
At least one S Only four S tiles exist. They are worth holding for what they do -- pluralise and hook -- not for being scarce; there is one J and nobody hoards that. | 25.5% | 4 |
Can you play all seven?
A bingo — laying all seven tiles in one word — pays a 50-point bonus, and it is the first thing anyone wants to know about a rack. 12.6% of opening racks can make one. That is the rack’s potential, not a prediction about you: it counts a rack as a bingo if some valid seven-letter word uses exactly those tiles, whether or not you would find it.
| Blanks on the rack | How often | Can bingo |
|---|---|---|
| None | 86.42% | 7.5% |
| One | 13.15% | 43.9% |
| Both | 0.42% | 81.1% |
This is the case for the blank, in one column. A rack without one bingoes 7.5% of the time; one blank takes that to 43.9%, and two to 81.1% — roughly 5.8× for the first one. It scores nothing and it is the most valuable tile in the bag.
Which word list? It matters: a bingo rate is a fact about a dictionary as much as about the bag. This is ENABLE (Enhanced North American Benchmark Lexicon), which is public domain. It is the open list most word games build on, and not an official tournament lexicon. Both official lists are larger and both are copyrighted, but not by the same margin: NWL2023 (North America) has about 14% more words than ENABLE, so its bingo rate is a little higher — the published figure for the 2014 list is 12.92%. CSW24 (Collins, used everywhere else) is larger again by roughly two thirds, so a Collins rate is meaningfully above this one rather than slightly.
Every one of the 3,199,724 distinct racks was checked against it, weighted by how often it comes up — nothing here is sampled.
The Q problem is mostly a myth
There is exactly one Q, so it reaches 7.0% of opening racks. Of those, 68% arrive with no U and no blank — four U tiles against one Q is not a fair fight. That is the figure the folklore stops at, and it is the wrong one.
QI is good in both official word lists, and an I is far easier to draw than a U — there are nine of them. (It is not in ENABLE, the public-domain list the word pages here are built on, which is one of eleven two-letter words NWL2023 has and it does not — so those pages will tell you a Q has nowhere to go in two squares, and at a rated game that is not true.) On 32% of Q racks the U never shows up but an I does, and the rack was fine all along. Allow for QAT as well and the genuinely awkward Q — no U, no blank, no I, no A — is down to 18% of the racks you draw it in: 1.26% of opening racks, or about one in 79, against one in 21 if you believe the folklore version.