Marvel Rivals and 106 Team-Ups: When Game Balance Becomes a Combinatorial Problem

Season 10 of Marvel Rivals has pushed The Hood onto live servers, carrying two new Team-Ups with it and lifting the total number of synergies in the game to 106. If NetEase keeps the hero release cadence of roughly one per month described in the current reference material, that list will pass 130 entries by the end of Season 12. Nobody can hold a lookup table that long in reactive memory, and that exact gap is reshaping how ranked players have to prepare before every match.

I spent several evenings with that list, not to memorise it, but to work out which kind of player the system actually rewards. The answer left me uncomfortable.

Marvel Rivals is a 6v6 hero-shooter operated by NetEase on the Marvel IP. Its biggest point of separation from the rest of the genre is the Team-Up system. Every hero owns exactly two synergy pairings, and according to the source material, no hero is ever released without a Team-Up.

The mechanic runs on two layers. The base layer is always available, granted regardless of which heroes your teammates pick. The enhanced layer only activates when the correct partner hero is present in the composition. Half of each pairing's power is therefore free, and the other half is locked behind a composition condition. The Hood entered in Season 10 with two new pairings. The reference material I used carries an update stamp of 14 September without a year, a small detail that was enough for me to flag doubt about how fresh the data really is.

A hero's value is now measured by the number of edges it connects to in the synergy web, not by its standalone strength. That is the core shift. In a normal hero-shooter, the drafting question is "which hero is strongest". In Marvel Rivals, the correct question is "which pairing web is strongest under the current patch". A small balance change to one Team-Up can ripple through multiple compositions at once, because every hero sits inside two pairings and every pairing drags two heroes with it.

That produces a combinatorial burden. With 106 pairings live, each new hero arriving on a monthly cadence adds at least two new edges to the graph. The balance surface grows faster than any tuning team can plausibly handle. This is not idle theory: once the number of pairings is large enough, the probability that at least one pairing dominates the rest approaches certainty, and finding it requires live performance data that the source material does not provide at all.

One design detail stands out: the source states new heroes will synergise with old ones. That is a backward-compatibility commitment, and it lets NetEase avoid the power-creep-by-obsoletion trap that turns old heroes into dead weight. But it also turns every new release into a rewrite of the power ceiling of a hero already in the game. The upside and the risk occupy the same slot, and they cannot be separated.

The base-and-enhanced split does ease the pressure of forced composition. Players keep baseline value even without their partner. But in solo ranked, where composition coordination is close to random, the enhanced layer becomes a dependency risk: whether you are strong or weak no longer depends on your own skill, but on whether a stranger on your team picked the right hero.

At 106 entries, system knowledge has become a competitive asset independent of mechanical skill. The source material recommends bookmarking the list for regular lookups. That advice sounds ordinary, but it concedes something: nobody can process this list through reactive memory any more. A player entering the game in 2026 faces hundreds of synergies they have never seen, while a player from the first season has accumulated them patch by patch. That distance is not about faster or slower hands, but about exposure time.

A monthly hero cadence also pushes the meta into permanent turbulence. One patch is never fully solved before the next arrives, and every tactical conclusion carries a very short expiry date. For professional teams, this shortens the adaptation window to the point where coaches must prepare several scenarios in parallel instead of optimising a single one.

Here I have to argue against myself. Every judgment above is structural, not data-driven. The source material supplies an inventory — which pairings exist — and not a single performance figure: no win rate, no pick rate, no ban rate. If I declared that a particular pairing is currently dominating, I would be making it up.

I could also be wrong elsewhere. The 106 figure may well have drifted, because a continuously updated document that stamps a date without a year always risks sliding out of sync between edits. The description of the base and enhanced layers may not match how the mechanic actually behaves in game, and if the real weighting leans further toward the base layer than the source describes, the whole concern about forced composition softens considerably.

Finally, the source mentions no tournament, no team, no player and no financial figure. So I refuse to draw conclusions about the professional scene. Marvel Rivals is a live-service game with an esports dimension, and the Team-Up system clearly matters to high-level drafting, but without tournament data in hand, any guess about tournaments remains just a guess.

My conditional prediction: if by the end of Season 11 NetEase has not published pick and ban data broken down by Team-Up pairing, the ranked ladder will self-organise around a small set of must-pick pairs, and most of the remaining 106 entries will exist on paper only. If they do publish, the story inverts — and I will be the first to rewrite it.

A hot take is not a hasty verdict. It is how I read a game using the reason of an outsider, and the Team-Up system in Marvel Rivals is the clearest example I have seen of a design betting that players will study more than they want to.