How to Top 4 in TFT Consistently: The Habits That Hold Every Set
TL;DR: A top four means outlasting four players, not winning every fight. In our game length figures, from ranked games played August 25 to October 4, 2026, the median 5th and the median 4th go out two player rounds apart, so one more survived fight can be the whole gap. The habits that hold in any set are simple: count your losses left, spend before your last two, field your strongest board, finish cheap upgrades first, roll for more than one name, finish your items, and count who can go out before you. Then be patient: a true 55% player still lands under half about 1 time in 7 over 100 games.
This page is reasoning from Riot’s rules and plain arithmetic, not a measure of what each habit is worth. Riot publishes no such figures and neither do we, so the habits aren’t ranked. The worked examples are hypotheticals. Gold figures are averages on the shop odds Riot last printed. For what counts as a good rate, see senseos.gg’s average placement page. This page is about how to get there.
What does a top four actually take?
Four players going out before you do. That’s the whole job, and it’s a different job from winning fights.
Riot’s Ranked FAQ (March 10, 2026) rules out losing LP at 4th or better and gaining it at 5th or worse. So the step between those two places is the one that changes which way your LP moves. Riot prints no table of amounts, so we can’t price that step exactly.
How wide is it? Our game length page, from 110,328 ranked Set 18 games played August 25 to October 4, 2026, puts the median 5th place’s last round at 30 and the median 4th place’s at 33. These are medians across different games, not one game. By the game’s round list, you count 4 for Stage 1, 7 for each later stage already finished, then the round’s position in its own stage. So 30 is 5-5 and 33 is 6-1, and the rounds after 5-5 are 5-6, a PvE round at 5-7, and 6-1. So that’s two player rounds. A lost round is a flat amount by stage, and each enemy unit left adds 1. Adding Riot’s 14.9 and 16.1 lines, at 6 enemy units those two fights cost 16 and 18 HP, or 34 together. At those medians, the gap between never gaining LP and never losing it is about one or two lost fights.
That’s why this page leans on small things. You don’t need a better comp to cover that gap. You need a little more HP at 5-5, or a little less gold left in your bank.
Which habits hold from set to set?
Each of these rests on a rule, not on a unit, and each has its own guide with the full sum.
Count your losses left every stage
Divide your HP by what a loss costs, round up, and take away 1. At 38 HP in Stage 5, with 16 HP a loss, that’s 2. Do it when each stage starts, because the same HP buys fewer losses every time: our HP guide has the table. A number you’ve worked out takes the panic out of the next decision.
Spend before you run out of losses
A bank only earns while you’re alive, and it buys nothing once you’re out. With 50 gold and two losses left it can earn about 15 more by sitting there, as the HP guide works out. Our rolling guide says where to stop when you do press.
Field your strongest board, every stage
Each enemy unit left costs you 1 HP, so a narrow loss beats a wide one even when both are losses. A unit on your bench kills nothing. Our flexible play guide has the habit, and our positioning guide covers keeping your carry alive long enough to make those kills.
Take the cheap upgrade first
The prices aren’t close. For one named 4-cost at level 8, rolls and buying together, on the rolling guide’s sum and the odds Riot last printed:
| What you’re after | Average gold |
|---|---|
| 2-star from a pair | 30 |
| 2-star from one copy | 56 |
| 2-star from nothing | 81 |
| 3-star from a 2-star, with nobody else holding one | 340 |
So hold your pairs, finish them, and treat a 3-star 4-cost as a bonus for a game that’s going well. Star levels has more on which 3-stars are worth planning for.
Roll for more than one name
With a pair of each at level 8, 20 gold finds one named 4-cost 55% of the time and either of two 80% of the time, as the rolling guide’s table shows. That’s about what 40 gold does for one name. A board that can use two carries hits more often.
Reach your level with gold to use it
A level with an empty bank leaves you on about 5 gold a round, which is 2.5 rolls at 2 gold each (“Reroll 2”, seen on the Reroll and Buy XP buttons in a patch 18.5 game on 29 October 2026, with Riot’s 17.2 line in support). Our fast 8 and fast 9 guides say how much to land with.
Finish your items
A component in your bag doesn’t fight. Our slamming guide has the sum for when holding pays, and it’s a shorter list than it feels.
Change lines on a sum, not a feeling
A lost round isn’t a reason to leave your line. Holding a pair of a 4-cost, finishing stays cheaper than a fresh 2-star until only 2 copies are left, as our pivoting guide works out. Moving twice in a game pays the bill twice.
How do you count the race to 4th?
Work out how many have to go, then who’s likeliest. With five players left, you need one more out before you. You need a look at their health and the same sum you did on yourself: anyone whose HP is at or under one loss can go out this round.
Then compare. Here’s the chance you outlast one rival, by the losses each of you can still take. In the first table you each win half your fights. In the second you’ve spent enough to win 60% of yours:
| You can take | Rival can take 0 | Rival can take 1 | Rival can take 2 |
|---|---|---|---|
| 0 losses, winning 50% | 50% | 17% | 6% |
| 1 loss, winning 50% | 83% | 50% | 24% |
| 2 losses, winning 50% | 94% | 76% | 50% |
| You can take | Rival can take 0 | Rival can take 1 | Rival can take 2 |
|---|---|---|---|
| 0 losses, winning 60% | 57% | 24% | 10% |
| 1 loss, winning 60% | 88% | 61% | 36% |
| 2 losses, winning 60% | 97% | 84% | 64% |
These are plain coin-flip sums, not a model of real fights. Your rival wins half of theirs in both tables. Each round each of you wins or loses separately. If you both go out on the same round, we assume it splits evenly and count it as half. That’s our assumption, because we found no Riot page that says how that tie breaks. The tables look at one rival only, so with two players below you, your real chance is better.
Two things stand out. One loss of room is worth a lot. It takes you from 50% to 83% against a rival with none left. Spending moves an even race most. Winning 60% instead of 50% lifts you about 11 points at 1 loss each and 14 at 2 each, but only about 2 points at 2 losses against none. These lifts are worked before rounding, so they can sit 1 off the tables. So when you’re far ahead in the race, your gold has better uses. When you’re level or behind, spend: the lift is smaller if you have no losses left, about 7 points against a rival with none and 5 against a rival with 2, but gold buys nothing once you’re out.
A hypothetical. Five players left, Stage 5, 16 HP a loss. You have 24 HP, so 1 loss left. The lowest rival has 12 HP, so none. You hold 40 gold. You’re at about 83% to outlast them already, and winning 60% only makes it 88%, so hold what you can use for the final fights and spend only what your next fight clearly needs. Swap the two HP numbers and you’re at about 17%, or 24% winning 60%. That’s still a long shot, but your gold is worth nothing if you go out, so you spend all of it.
What changes once you’re in?
Everything about the risk. Four or fewer players left means your LP can’t go down, so stop protecting it. Our late game guide has the break-even sums and what to spend on. If you’re near 0 LP, our demotion page goes through that spot.
How long before a better habit shows?
Longer than it feels. Our page on whether TFT is rigged has the basics, including that 50% to 55% flips 5 games in every 100. Here’s what that means for what you’ll see. These are chances of reaching a 60% top 4 rate or better in a stretch of games, from a binomial, for players whose true rate is 50%, 55% and 60%:
| Games | Top fours | True 50% | True 55% | True 60% |
|---|---|---|---|---|
| 10 | 6 or more | 38% | 50% | 63% |
| 20 | 12 or more | 25% | 41% | 60% |
| 50 | 30 or more | 10% | 29% | 56% |
| 100 | 60 or more | 3% | 18% | 54% |
Over 10 games, a true 50% player shows you 6 or more top fours 38% of the time. So a good evening says very little. Now the other side: how often a true 55% player finishes under half.
| Games | Under half means | Chance |
|---|---|---|
| 25 | 12 or fewer | 31% |
| 50 | 24 or fewer | 20% |
| 100 | 49 or fewer | 13% |
| 200 | 99 or fewer | 7% |
| 400 | 199 or fewer | 2% |
So a player who has really improved can still look worse for 50 games. Change one habit at a time, and judge it by whether you made the decision right, not by the placement that game. Decisions you can check after one game. Placement takes a couple of hundred.
Which habit should you fix first?
Note the round you go out on and the gold you held, for your last few bottom fours. Then read it against the medians above and in the late game guide. These are starting points, not diagnoses:
| Where you keep going out | Against the medians | Look at first |
|---|---|---|
| Before 5-1 | Earlier than the median 8th place, on round 26 | The HP budget in our mid game guide |
| 5-1 to 5-5, gold unspent | Among the bottom four’s medians | When you spike, and when to stop bleeding |
| 5-1 to 5-5, no gold left | Same | Your level and roll timing in the fast 8 guide |
| Around 5-5 to 6-1, narrowly | Right on the line | Your last fights: positioning, scouting and items |
| You make top four and stop at 4th | At the line, from the good side | Whether you switched to playing for first |
For more on what to change after a run of bottom fours, our demotion page has a checklist, and why you keep losing goes through the gold and level habits to check.
What changes when Set 19 lands?
Very little here. The LP rule sits in Riot’s FAQ, not in a set’s notes, and every habit above is a rule about gold, HP and copies. Riot’s schedule lists 19.1 for December 1, 2026, and warns that dates can change. When it lands, check player damage and odds in its notes, since 16.1 changed them at a set launch, and remember the median rounds come from Set 18 games.
A new set is also the best time to practice. At Set 18’s launch, Riot’s Enchanted Wilds: Part I page (August 26, 2026) gave everyone five provisional games, where a bottom four cost no LP and a top four paid bonus LP. We found no Set 19 page yet, so check that before you count on it.
What should you do in your next game?
| Your spot | The call |
|---|---|
| Each new stage | Count your losses left |
| Two losses left | Spend what your next fight needs |
| Pair of your carry on the bench | Finish it before you chase anything bigger |
| About to roll | Have two names, and a number to stop at |
| Components in your bag | Slam them unless you can name why not |
| Five left, someone below you | Compare your losses left with theirs, and keep what you can |
| Five left, you’re the lowest | Spend all of it |
| Four or fewer left | Stop protecting your LP, and play for first |
| A good or bad run of ten games | Don’t change anything because of it |
If you want a second opinion while you play, TFTSense gives econ and leveling calls and tells you what to slam and who should hold it.
Sources
All read November 4, 2026. Riot publishes no figures on what any habit is worth, so each number links to the Riot page that sets it.
- Riot Games Support, Teamfight Tactics Ranked FAQ, dated 2026-03-10, for LP by placement.
- Riot Games, patch 10.6 notes, published 2020-03-17, for base income.
- Riot Games, patch 14.8 notes, published 2024-04-16, for each surviving enemy unit counting 1.
- Riot Games, patch 14.9 notes, published 2024-04-30, and patch 16.1 notes, published 2025-12-02, for player damage.
- Riot Games, patch 17.2 notes, published 2026-04-28, for the line that supports the reroll price.
- The game itself: “Reroll 2”, seen on the Reroll and Buy XP buttons in a patch 18.5 game on 29 October 2026.
- Riot Games, patch 18.6 notes, published 2026-11-03, checked for changes to player damage, odds and ranked.
- Riot Games Support, Teamfight Tactics, Enchanted Wilds: Part I, dated 2026-08-26, for provisional games.
- Riot Games, Patch Schedule, Teamfight Tactics, for the 19.1 date.
- The game’s data files, read November 4, 2026, for the round list.
- Our game length page, for the median last round by placement, from 110,328 ranked games, and our rolling guide, for gold per copy and the find chances, with its own Riot links.
The chances in the race tables and the rate tables are our own arithmetic. The rate tables are binomials, and each “or more” figure counts that number of top fours and every higher number. The race tables work one player round at a time, with a 50% or 60% chance for you and 50% for your rival. Counting a same-round exit as half is our assumption.
Update log
- 2026-11-04: First version.
Frequently asked questions
How do you top 4 consistently in TFT?
Play to outlast four players rather than to win every fight. Count your losses left each stage, spend before you're down to your last two, field your strongest full board, finish cheap upgrades before chasing big ones, roll for more than one name, and finish your items. These are reasoning from Riot's rules, since Riot publishes no figures on what each habit is worth.
Is top 4 or first place better in TFT?
Top 4 is the line Riot's rules let you be sure of. Its Ranked FAQ says LP goes by how well you did, that 4th or better can't lose LP and that 5th or worse can't gain it, and it publishes no table for the amounts. So protect top 4 until you're in it. Once four or fewer players are left, your LP can't go down, so play for first.
What top 4 rate should you aim for in TFT?
Above 50%, since four of the eight places in every lobby are top fours, so the average player gets exactly half. Going from 50% to 55% turns 5 games in every 100 from a finish that can't gain LP into one that can't lose it. What ladder players actually reach is on senseos.gg's average placement page.
How many games does it take to see if you've improved at TFT?
More than feels fair. A player with a true 55% top 4 rate still lands under half about 31% of the time over 25 games, 13% over 100 and 2% over 400. So one good weekend proves little, and neither does a bad one. Change one habit and give it a couple of hundred games if you can.
Should you spend all your gold to reach top 4 in TFT?
If you're the one who can go out next, yes. Gold buys nothing once you're out, and 5th can't gain LP. If someone below you is likelier to go first, you may not need all of it. With one loss left against a rival with none, a plain coin-flip sum, which assumes a same-round exit splits evenly, gives you an 83% chance of outlasting them before you spend, and winning 60% of your fights instead of half only lifts that to 88%.
How this guide was made: researched, drafted or edited with AI tools from MadeByKeo's notes, Riot's own sources and our own data. MadeByKeo decides what is published. How we make guides
Put this into practice
- Set 18 comps tier list: every live comp ranked by tier, with carries, items and the full board.
- TFTSense features: what the coach calls during a game, read from your screen.
- TFTSense compared with other TFT apps: Mobalytics, MetaTFT, TFT Academy, Blitz and more, feature by feature.
- Download TFTSense: the Windows installer.
Questions, or want a hand with your climb? Come hang out with us.
Join the Discord