How to Balance Teams When Skill Ranges Herald to Divine

Updated 2026-07-13

How do you balance a Dota 2 team when skill ranges from Herald to Divine?

Balance wide skill range teams by spreading the strongest and weakest players across both sides instead of clustering them, so each team ends up with a similar mix of skill levels rather than one team getting every high-rank player and the other absorbing every low-rank one. A strength-sorted snake draft does this automatically: rank every player from strongest to weakest, then deal them A-B-B-A-A-B-B-A across the two teams, so each side alternates between higher- and lower-skill players as the list is dealt out. The result on a truly wide-range group, Herald to Divine, is two teams that each have a strong anchor player and a player who's still learning, rather than one team of anchors facing one team of learners.

Pairing high and low players deliberately, rather than hoping a random split lands evenly, is what makes a Herald-to-Divine lobby playable at all. Two anchor-tier players on opposite teams effectively cancel each other out at the top of the strength range, the same way two Herald-tier players on opposite teams cancel out at the bottom — the total gap that's left over is far smaller than the raw spread between the highest and lowest individual player in the room.

Why does one Immortal warp the lobby?

One Immortal warps the lobby far more than a simple average shows, because the real strength gap between an Immortal and a Herald or Guardian teammate is enormous — often the difference between someone who can solo-kill lanes and someone still learning last-hitting. In Valve's ranked matchmaking there's an actual enforcement mechanism for this: a party that includes an Immortal player has every other party member's rank adjusted up to the Immortal's for matchmaking purposes, specifically because pairing an Immortal with much lower-ranked teammates would otherwise break matchmaking fairness. Custom lobbies have no such enforcement — nothing adjusts anyone's effective rank for a private game, so if your 10-stack includes one Immortal and several Heralds, the responsibility for not letting that Immortal single-handedly decide the game falls entirely on however you build the two teams.

The practical fix is to treat the Immortal's strength score as what it actually is — usually several times a Herald player's — rather than trying to compare medals directly. A balancer that converts rank tier into a strength score already does this: each medal tier spans a wide approximate MMR band, and by the Elo-style expected-score curve, every few hundred points of rating gap compounds the stronger player's win odds — stack seven tiers of it and the practical gap between Herald and Immortal is far bigger than counting medal steps suggests, and any split needs to price that in before totaling.

Does role coverage still matter in a wide-skill lobby?

Role coverage matters even more in a wide-skill lobby, because a low-rank player is much more effective in a simple, well-defined position than being forced into an unfamiliar role under pressure. Put your least experienced players in support slots with clear jobs — warding, stacking, following the offlaner — rather than handing them a solo lane against a much stronger opponent. Covering all five positions on both teams, adjusted for who's actually comfortable where, does more for a wide-range lobby's playability than any amount of extra MMR precision.

Can a balance score confirm a Herald-to-Divine lobby is actually fair?

Yes — a balance score of 85% or higher means the two teams' total strength is close even when the individual skill range inside the group runs the full ladder from Herald to Divine. The score doesn't hide the spread or pretend everyone is closer in skill than they are; it confirms that the spread has been distributed evenly across both sides. Dota 2 Groups' Auto Balance handles this the same way it handles a tighter-skill lobby — pull rank tier, win rate, and role history for all ten players, seed with a snake draft, then run up to 1000 swap iterations — the process doesn't change just because the range is wider, only the size of the gaps it's working with.

Frequently asked questions