TFT shop odds, the champion pool and rolldowns
The shop is not random in the way most players think. It is a fixed set of probabilities over a shared pool, and both halves of that sentence are exploitable.
Shop odds by level
Every refresh costs 2 gold and rolls five slots independently. Each slot checks this table for a cost, then picks a champion of that cost from what's left in the pool.
| Level | 1-cost | 2-cost | 3-cost | 4-cost | 5-cost |
|---|---|---|---|---|---|
| 4 | 55% | 30% | 15% | — | — |
| 5 | 45% | 33% | 20% | 2% | — |
| 6 | 30% | 40% | 25% | 5% | — |
| 7 | 16% | 30% | 43% | 10% | 1% |
| 8 | 15% | 20% | 32% | 30% | 3% |
| 9 | 10% | 17% | 25% | 33% | 15% |
| 10 | 5% | 10% | 20% | 40% | 25% |
| 11 | 1% | 2% | 12% | 50% | 35% |
Two rows are highlighted because two rows decide nearly every rolldown in the game. Level 7 is peak 3-cost at 43%, and level 8 is peak 4-cost at 30%. Everything else on this page is a consequence of those two facts. The full table, with XP and pool, is on the shop odds page.
The shared pool, and what a unit really costs
Odds are only half the calculation. When a slot decides it wants a 4-cost, it picks from every 4-cost copy still in the pool — all 140 of them. Your specific champion is 10 of those 140, so about 7%.
Multiply it out and you get the number nobody quotes but everybody should: the gold cost of a copy.
| Gold per copy | 3-cost | 4-cost | 5-cost |
|---|---|---|---|
| Level 7 | 13g | 56g | 400g |
| Level 8 | 18g | 19g | 133g |
| Level 9 | 22g | 17g | 27g |
| …for a 2★ (3 copies) | 39g | 56g | 80g |
Assumes an untouched pool and a refresh costing 2 gold. The bottom row uses each cost's best level: 3-costs at 7, 4-costs at 8, 5-costs at 9.
Sit with the 4-cost column for a second. At level 8, three copies of a specific 4-cost carry runs to roughly 56 gold on average, and that is with nobody contesting you. This is why arriving at your rolldown with 20 gold is not "rolling badly" — it is not rolling at all. You needed two and a half times that much before you started.
It is also why the same table makes level 7 rolling for a 4-cost look absurd: 56 gold per copy against 19 at level 8. The 60 XP to level pays for itself in the first ten refreshes.
Being contested
Those numbers assume the pool is full. It never is. Every copy held by another player — on their board, on their bench, inside one of their 2★ units — is gone from the denominator.
One opponent holding a 2★ of your 4-cost carry has removed 3 of its 10 copies. Your per-copy cost goes from 19g to about 27g — a 43% tax, invisible, applied silently to every refresh you make. If two players are on it and one has a 2★, you are paying nearly double and you will conclude you got unlucky.
| Copies left of your 4-cost | Gold per copy at level 8 | Gold for a 2★ |
|---|---|---|
| 10 of 10 (nobody else on it) | 19g | 56g |
| 8 of 10 | 23g | 70g |
| 6 of 10 | 31g | 93g |
| 4 of 10 | 47g | 140g |
The lesson isn't "never play contested" — sometimes the contested comp is still your best line, and being second to a unit is survivable if you're first to the level. The lesson is that you must know before you roll, because the answer changes how much gold the attempt needs and whether it's worth attempting at all. That's a scouting job, and it gets its own chapter.
Executing a rolldown
A rolldown is a committed spend, not a mood. It has a structure.
Arrive at the right level with the right bank
Level first, then roll. Fifty gold at level 8 finds more than eighty at level 7. If you can only afford one of the two, level and roll next round.
Know your three targets before the first refresh
Name them. The carry, the second-best holder for your items, and the frontline unit you're missing. Rolling without named targets is how 40 gold disappears into 3-costs you sell two rounds later.
Buy every copy of every target, immediately
Even the fourth copy of something you already have at 2★, if the bench has room. Copies you don't take go back in the pool for the player contesting you. Bench space is cheap; a copy you passed on is gone.
Reassess at 20 gold, not at 0
At 20 gold you still have a choice. At 0 you have a board and no way to fix it. If you haven't hit by 20, stop, play what you have, and take the round — you can roll again next round with income restored.
Reposition before you hit "ready"
You just changed your board substantially and the positions are from two rounds ago. This is the single most common way a successful rolldown still loses the fight.
"I've already spent 40, I may as well keep going" is the most expensive sentence in TFT. The 40 is gone either way. The only question is whether the next 20 has better odds than saving it — and after a failed rolldown the pool is emptier and the answer is usually no.
Slow rolling
Nine copies make a 3★. For a 1-cost that's 9 of 30 and entirely reasonable; for a 4-cost it's 9 of 10 and basically the whole supply. So three-starring is a low-cost-unit strategy, and it needs a different rolling technique.
Slow rolling means holding a level and spending only the gold above 50 each round. You stay at maximum interest permanently, and every round you convert your entire income into refreshes. Over eight rounds that's far more refreshes than one rolldown, spread out so the pool has time to recycle from other players selling.
| Carry cost | Sit at level | Why |
|---|---|---|
| 1-cost | 4 or 5 | Highest 1-cost density. 30 copies makes 9 realistic. |
| 2-cost | 6 | 2-costs peak at 40% here. |
| 3-cost | 7 | 43% — the highest single-cost number in the game. |
| 4-cost | — | Don't. Nine of ten copies is not a plan, it's an outcome. |
The cost of slow rolling is that you are permanently a level behind, so your board is smaller. That trade only pays if the 3★ actually arrives. Give yourself a deadline — usually stage 4-5 — and if it hasn't happened, stop and level. A 2★ board at level 8 beats a 2★ board at level 6 every time.
Knowing when to stop
Three signals that the roll is over, whatever your gold says:
- You hit your main target. The rest of the board can be found next round at full income. Continuing to roll for the fifth-best unit is how you end up at 0 gold with one good unit.
- Two other players are visibly on your units. The pool maths above just changed under you. Bank it, play the round, reconsider the comp — see pivoting.
- You're at 20 gold without your target. Stop. This is a rule and not a judgement call, because the moment you're making a judgement call at 12 gold you've already made it badly.
Everything on this page is arithmetic you can do yourself, and nearly nobody does it mid-game — not because it's hard, but because there are thirty seconds on the clock and seven other boards to look at. The overlay keeps the pool counts and shop highlighting on screen for exactly that reason. The maths doesn't change; only whether you have it when you need it.
You now have the gold, the level and the units. Next: what to put on them, which is where more games are quietly lost than in any rolldown.