Marvel Rivals Season 10: 106 Team-Ups, Two Per Hero, and a Balance Problem Nobody Can Solve in Time

On the night of September 14, in a small internet cafe on Tri Xuan Street in Haidian District, Beijing, I sat behind a twenty-one-year-old student. He locked in The Hood, Marvel Rivals' newest character, then turned and shouted at the friend beside him: "Take the other one, take it." The friend hesitated for about three seconds and locked. The Team-Up bar lit up on screen, and the match turned from the second minute.

That student is not a better shooter than anyone. He still misses headshots, still stands at the wrong angle, still reloads at the worst possible moment. But the system had just handed him something he could not unlock on his own. The entire design philosophy of Marvel Rivals fits inside that moment, between the ceiling fan and the smell of instant noodles.

I have been going to places like that for nearly half a year, not to play — I play badly — but to watch how people decide before a match begins. How they pick, how they bargain with teammates, how they drop a good option simply because nobody will take the other half. Season 10 has just pushed that whole system into a new state.

Context

Marvel Rivals is a 6v6 hero shooter developed by NetEase, positioned as a direct competitor to Overwatch. What separates it from the rest of the genre is not the Marvel roster; it is the Team-Up system.

The mechanics are fairly clear. Every character has Team-Ups, and no character is released without at least one pairing. Currently each hero has exactly two Team-Ups, and 106 pairings are live. Season 10 added The Hood and his Team-Ups, while keeping the rule that new heroes pair with old ones rather than standing alone.

The most important technical detail is how effects are tiered. Each Team-Up has a base effect that is always active, and an enhanced effect that only triggers when the partner hero is in the composition. Half the power of a pairing is free; half is conditional.

On release cadence, new heroes arrive roughly every month or a little longer. That cadence matters more than it looks, and I will come back to it at the end.

Analysis

The Team-Up system turns the genre's familiar question — "which hero is strongest?" — into a fundamentally different one: "which pairing web is strongest in this patch?" That is a shift from a pick problem into a graph problem. With 106 live edges and every new hero adding at least two more, the surface that needs tuning is far larger than the hero count suggests.

Combinatorial burden is the biggest structural risk in Marvel Rivals, not any single pairing currently accused of being overpowered. A balance team can review heroes one by one, but it cannot review edges the same way. Once the number of pairings passes the threshold a person can hold passively in memory, quality control shifts from monitoring to simulation, and simulation is always slower than reality.

Based on my own experience watching matches across several internet cafes, the clearest change is not in the fights but in the pick phase. Players used to lock what they liked and figure it out later. Now they ask first: "What are you on?", "Do you have the other one?", "Can you swap?". A small negotiation happens before every match, and whoever wins that negotiation usually wins the game.

This creates its own knowledge economy. With 106 pairings, nobody remembers them all reactively. The very compilation I used as a source advises readers to bookmark the list for reference — a sign that the game's knowledge barrier has risen high enough that memorization has become a skill. New players must learn not only how to shoot, but an entire lookup table of relationships.

At the player level, the system stratifies clearly. Players with wide hero pools have more pairing doors open, so their value rises without them shooting any better. One-trick players get pushed to the margins, because their enhanced effect depends on someone else agreeing to take the partner pick. In many matches I watched, the biggest loser was not the worst shooter but the player with the fewest options.

The base-and-enhanced split softens that pressure without removing it. Heroes retain minimum value when played alone, so you are not fully punished for lacking a partner. But the gap between having and not having a partner is a measurable advantage, and in ranked play a measurable advantage always wins. Every contract starts with a person before it becomes a number — and so does every Team-Up: it starts with a hero before it becomes a line in a 106-line table.

I think this is where the difference lies: at the system layer, not the content layer. A game can compensate for a content shortfall by shipping more heroes, but it cannot compensate for a balance shortfall by shipping more heroes. Each added edge does not increase the number of hypotheses linearly; it spreads them across a network, because a small change to one hero can touch several pairings at once.

A monthly cadence raises everything a notch. If a new hero arrives each month and each new hero adds two edges, the meta is never solved. For professional teams that compresses the adaptation cycle: by the time you have learned one patch, the next has landed. For ranked players it means a permanent sense that you are trailing a data table.

The Contrarian Angle

The compilation offers no performance metric whatsoever. No win rate, no pick rate, no ban rate. It lists the existence of pairings rather than measuring their strength. Every judgment about meta direction, including the one I just made, is structural inference rather than data inference. I say this because I once paid the price for reading a compilation and mistaking it for data.

The counterintuitive part is this: Team-Ups can make the game shallower, not deeper. If each hero has exactly two pairings, the pick screen stops being a creative space and becomes a lookup table. Creativity migrates entirely into execution — positioning, timing, coordination — while the single biggest strategic decision of the match is compressed into a table lookup. Designers tend to believe that adding relationships adds depth. That is not always true.

The commitment that no character ships without a Team-Up is a long-term design contract. It protects old heroes from obsolescence, but it also means every new hero can raise the ceiling of an existing hero in ways nobody anticipated at launch. That is power creep wearing the costume of preservation.

There is another layer the compilation never mentions. Dehumanization begins with the way we name a person using data. Here, heroes become nodes, players become "hero pools," and a match becomes a pairing problem. I once did the same thing to footballers, and I know how comfortable that feeling is — it makes everything look controllable.

Takeaway

The next domino worth watching is not which pairing gets nerfed, but whether NetEase shifts to a seasonal rotation of Team-Ups. Activating all 106 pairings at once is the fastest route to a balance surface beyond anyone's reach; rotation is the viable way to keep it manageable without discarding design work already done.

We — the commentators — are sometimes the ones holding the scissors that cut people into pieces. I wrote this from behind a student in Haidian who just wanted to win one ranked match before sunrise. If the system teaches him that victory lies in finding the right person to pair with, then the system has taught him a correct lesson — it is simply a more expensive lesson than it looks.