Three Years, Six Titles: MSI Is Reading Worlds Better, But Not Well Enough to Trust

4:12 a.m. Incheon time. The MSI final ended and I closed a notebook that had filled four pages. No scores in it, no KDA lines. Only hurried notes: minute 14, red side gives up the second dragon to take two mid turrets; minute 27, forces a fight at blue buff instead of waiting on Baron; minute 33, flips the push to the bottom lane before blue side can rotate back. Three months later, in the Worlds quarterfinals, I opened that same page. Four of the six notes reappeared almost word for word.

The belief that grows out of that is comfortable: MSI is teaching us something real about Worlds. The problem is what it teaches, and that is not what most viewers want to hear.

Across 2026, 2026 and 2026, the six biggest titles in League of Legends — three MSI crowns, three Worlds crowns — all went to the LPL and the LCK. Four organizations reached the top step: JDG, BLG, Gen.G and T1. A six-out-of-six record sounds solid. Read more closely and it tells a different story than the one circulating on social media.

Context: MSI changed shape

MSI launched in 2026 as a high-level exhibition between spring champions. The old format was almost bare: six teams, one group stage, semifinals, final. Most players ran two or three best-of-ones before stepping into the tournament's only best-of-five. With a data sample that thin, MSI could barely separate a genuinely strong team from a team that simply caught the right moment.

Starting in 2026, Riot expanded MSI to thirteen teams, split into two brackets, playing double elimination. The number of best-of-fives a champion must survive rose from one to three, sometimes four. That is a hinge change, and I believe it explains most of the so-called improvement in MSI's predictive power.

Sân Cỏ và Bản Đồ — the name I chose for my blog at sixteen — keeps reminding me of one thing: to measure correctly, you first need enough measurements. The old format gave us one. The new format gives us four. Every conclusion about a more clairvoyant MSI must start there, not from the adrenaline of a five-game final.

The core: MSI reads the tier, not the throne

This is where the data gets interesting.

In 2026, MSI ended with JDG champion, BLG runner-up, Gen.G and T1 in the top four. That year's Worlds ended with T1 champion, WBG runner-up, BLG and JDG in the semifinals. Three of four names overlapped, only shuffled.

In 2026 the script repeated with unsettling precision. MSI: Gen.G champion, BLG runner-up, T1 among the leaders. Worlds: T1 champion, BLG runner-up, Gen.G in the last four. Again three out of four.

In 2026, Gen.G won MSI and T1 won Worlds. The name at the top changed, but the set of teams still standing in the final weeks of the season revolved around the same four organizations.

MSI does not predict the Worlds champion. It predicts the tier of the Worlds champion — and those are two entirely different problems.

To test that, look at the full history. Across ten MSI editions, only once did the MSI champion go on to win Worlds the same year: SKT in 2026. Every other case breaks the rule. RNG won MSI in 2026 and 2026; Worlds went to IG and then EDG. G2 won MSI 2026; Worlds went to FPX. SKT won MSI 2026; Worlds went to SSG. So when someone claims that winning MSI usually leads to winning Worlds, they are misreading history — and misreading the present.

What changed after 2026 sits one layer down: the ability to correctly identify the group of teams capable of going deep. The old MSI regularly pushed to the top step teams that were only good at one spring meta — a composition built around one carry champion, an early-push style, one perfectly timed individual play. Double elimination grinds that kind of team down, because it forces multiple best-of-fives against different opponents, including teams willing to impose a completely different tempo.

In other words, the expanded MSI did not create stronger teams. It created a better sieve.

Cross-ecosystem reading: lessons from Katowice and from the pitch

I keep the habit of comparing League of Legends to other ecosystems, partly because I came up through hosting major events, partly because I believe every map eventually shares one shape.

In Counter-Strike, IEM Katowice and the Major are two milestones a few months apart. Katowice was long treated as the midterm, the Major as the final. But the CS community stopped comparing them directly long ago, simply because it understands this: the Katowice champion is usually the team reading the current version best, while the Major champion is usually the team adapting fastest to the next version. Different abilities, different grading systems.

Football works the same way. The Champions League group stage and the knockout rounds answer different questions. The group stage asks who is consistent enough not to drop themselves across six matches. The knockout rounds ask who can carry the pressure of a single night, where one small error is magnified into a trophyless season. Plenty of group-stage winners fall in the quarterfinals, and nobody calls that a paradox.

Every arena has a map; the winner is whoever reads it before the ball rolls. The problem in League of Legends is that people are reading one map and assigning it the function of another.

The patch as invisible referee

There is one variable every MSI-versus-Worlds debate skips: the version gap.

MSI is played on the mid-season patch. Worlds is played on the late-season patch, after a major adjustment Riot usually ships around midyear. Between those points lie four to five months, and inside that window sits at least one large-scale change: minion experience tuning, turret strength, item group adjustments, or simply a rotation in the mid-lane champion pool.

Throughout my tracking career I keep a separate page for each version and mark three things: the pick rate of the jungle champion pool, the average time to first turret, and the number of full teamfights in the first ten minutes. Those three are enough to tell whether a team is playing on last season's instinct or has shifted to the new season's tempo.

Here is the crux. A patch wields the authority of a referee — it does not blow the whistle, but it decides who is allowed to run. When an MSI champion then fails at Worlds, most online analysis blames individual form or mentality. The truth is usually more mundane: that team's power curve was drawn on a version that no longer exists in October.

We habitually call patch-adaptation ability "strength." I disagree. Strength is what remains after the patch changes. Adaptability is only the condition that lets that strength show up in time. Blending the two is the most common analytical error among League of Legends viewers.

Why six out of six is still not enough

Now comes the part where I have to argue against myself.

The dataset of six titles in three years all going to the LPL and LCK sounds weighty, but it holds only six data points, and three of them are variants of the same cluster of teams. In statistics, this is a sample size any serious practitioner must call far too small for conclusions. Three years does not make a rule. Three years makes a trend worth tracking.

Worse, a counterexample sits at the very heart of that trend, and it is too large to ignore. T1 won Worlds three years running but never won MSI in that stretch. They did not even win their domestic title. If MSI were truly a strong indicator for Worlds, the team that dominated Worlds for three straight seasons should have appeared in the MSI winner's circle at least once.

Put another way, six out of six is a regional effect, not a team effect. It tells you the two largest League of Legends regions still produce every champion. It does not tell you which of them will win. And if you use it to place faith in one specific name, you have read a regional map and mistaken it for a match map.

There is one more possibility I am obliged to write down, dull as it is: the correlation between MSI and Worlds may not be causal. The expanded format ranks teams more accurately. But more accurate ranking does not create stronger teams — it only lets us see more clearly what already existed. If that holds, saying MSI predicts Worlds is just another way of saying we got better at measuring.

What to watch at MSI 2026

As a tracker, I will keep three columns in my notebook for the next MSI.

Column one: who wins and which region they come from. If a team outside the LPL and LCK reaches the final, the six-out-of-six streak ends and the whole story must be rewritten.

Column two: who finishes second, who finishes third. This is the most underrated part. A team that loses the MSI final in five games retains exactly the same signal about its tier. Historically, many Worlds semifinalists started from second or third at MSI. The champion merely collects the trophy; the last four are the ones answering the real question.

Column three, the one I value most: which patch MSI is played on and which patch Worlds is played on. If the gap between the two versions is wider than usual, every conclusion drawn from MSI has to be heavily discounted.

Closing

The map is only correct until the ball lands.

I printed that line on my wall at sixteen, after nearly forty people called me a keyboard coach for daring to propose a 3-4-3 for a goalless draw in the K-League. Years later, analyzing League of Legends matches at four in the morning, I understood that the principle never changes across sports. Every model holds until real data collides with it.

The useful question ahead of MSI 2026 is not who will win Worlds. The useful question is: once MSI closes, how many teams are still standing in the building in October? For three years the answer has been four, and always from two regions. If that holds again, we have no rule — only a better habit of reading maps.

And a habit, unlike a rule, can be broken in a single night.