How it works

Upload a .StormReplay file and we read the players, heroes, result, map, game mode, awards, talents and end-of-game stats from it. You must have played in the game, so we check that your Battle.net name matches a player in the replay. The same game uploaded by several players is merged into one match, and uploading a file again only fills in details we didn't have before; it never overwrites what is stored.

Games with computer players are kept, but bots are never listed, rated or claimable. Players who opt out are hidden from other people's views.

The game mode (Quick Match, ARAM and so on) comes from the matchmaking id stored in the replay. If a replay has an id we don't recognise it shows as "Other mode".

• You can rate teammates from your own games in the last 14 days, and change a vote for 7 days after casting it.
• Each vote is a thumbs up or down, with optional strength and problem tags (hover each for what it means), a short comment (up to 280 characters) and 1–5 ratings for fun to play with and reliability.
• You can cast up to 30 votes a day, and brand-new accounts wait 24 hours before voting. Votes and comments can be reported and are reviewed by moderators.
• If the replay shows you gave a teammate an in-game end-of-game upvote, they are tagged "You upvoted in-game" and the thumbs-up is highlighted. Nothing is voted for you; click it to add a site vote. Replays uploaded before this was tracked won't have the tag.

• Overall score (0–100)Thumbs-up share of votes, with a starting nudge toward 50 so one early vote doesn't swing it. Votes lose half their weight every 180 days, so recent behaviour counts more. Upvotes players give each other inside the game count as a fraction of a thumbs-up and can never lower a score. A text label (Well liked, Liked, Mixed, Disliked, Poorly rated) appears after 5 votes.
• Fun to play with and ReliableThe average of the 1–5 ratings, older ratings counting less, pulled slightly toward the middle and scaled to 0–100. Shown after 3 ratings.
• Silence penalties and Fun to play withEach game where a player had a chat or voice silence penalty counts as an extra 1-star signal in their Fun to play with score, dated by that game. These signals pull the score down but never unlock the bar on their own. Games uploaded before silence was captured are not counted.
• Damage build (healers)A healer who took 4 or more talents from a hand-kept list of damage-focused talents (extra damage, basic attack damage, or offensive heroics, but not crowd control or mobility). It is a hint that they were building to fight rather than to heal, not proof they healed badly.
• Win impactWin rate in the games we have seen, pulled toward 50% because matchmaking aims for even games. Shown after 10 games. It is a record, not a measure of skill: a player on a winning streak with friends will look good.

Awards like MVP and Most Hero Damage are the ones the game itself hands out at the end of a match, copied from the replay. We add one of our own:

• Globe GrabberHad the most regeneration globes in the game (10 or more). Ties share it. The replay counts a globe for every hero close enough to benefit from it, not just the one who collected it, so this reflects who was in the right place rather than who grabbed them.

These are for fun and look only at numbers in a single game. They need a minimum number of deaths so a quiet game doesn't brand anyone.

• Top FeederDied the most of anyone in the game, with at least 3 deaths. Ties share it.
• FeederDied more than 1.25× the average for their own team (at least 3 deaths), without being the top feeder.
• Dead WeightLowest takedowns minus deaths in the game, only if that number is negative and they died at least 3 times. Ties share it. A support who rarely gets kills can earn this, so read it alongside the role rating.
• QuitterLeft the game more than a minute before it ended.
• AFKIdle for 2 minutes or more. Quitters get Quitter instead.
• Time Out ChairSpent the most time dead in the game, and at least 4 minutes and 25% of the game. Ties share it.
• PacifistLowest hero damage on their team among players who aren't tanks or healers, and under 50% of what their teammates in that group averaged. Needs 3 such players on the team and no tie for lowest.
• Paper Tank and Missing HealerGiven to a tank (Paper Tank) or healer (Missing Healer) whose role rating is under 20, the Poor band. See Tank and healer ratings below for exactly how that rating is worked out, including who they are compared with and what it can't tell us.
• Top DonorGave the enemy team the most XP through deaths in the game, at least 1,000 XP. Each death is worth 250 + 50 per level of the dead player's team, scaled up or down by the level gap between the teams, so dying while your team is behind costs less than dying while it is ahead. It is an estimate from death times and once-a-minute team XP, so it can be off, and games uploaded before this was added need a re-upload to get it. Ties share it.
• Odd PickA talent choice that makes little sense. Either it is on a hand-kept list of bad combinations for that mode (for example Anduin's Light Bomb in ARAM), or it is a talent taken in 3% or fewer of games on that hero in the same kind of mode (ARAM or not), judged once there are at least 40 such games, and the player lost. Rare picks can be creative rather than wrong, so winning games are never flagged by the rarity rule.

The game page compares the two teams on level, takedowns, towns destroyed, mercenary camps, lane XP, time spent dead, hero damage, siege damage, healing and regeneration globes (counted for every hero a globe's effect reaches, so a team's total is several times the number of globes actually picked up). A gap only counts if it is big enough to matter, and each gap is scored by how large it is relative to the totals, weighted by how much that factor tends to matter (team level and takedowns most, globes least). The biggest few become the reasons listed. If the winners were behind on something, it appears as "still ahead, but it wasn't enough".

The weights are judgement calls, not a proven model. Treat the list as a summary of what stood out, not a verdict.

The Comebacks column on My heroes shows, for each hero, how often your team won after trailing. A game counts as trailing if, at any point after 5 minutes, your team's total XP was more than 10% below the enemy team's. The column reads wins / trailing games, for example 3/8 means your team trailed in 8 games and won 3 of them. Games without XP data are left out, and a dash means the team never trailed that far. XP is a stand-in for being behind, since a team can trail on XP and still hold a stronger late game, so treat small samples with caution.

• Biggest contributor (winning team)Each player is scored on their share of their team's hero damage (35%), healing (30%), siege damage (15%), XP (15%) and takedowns (10%), minus a penalty for their share of the team's deaths. A healer's healing share is capped at 40%, since a lone healer would otherwise win every game. If the top two are within 10% of each other, nobody is named.
• Biggest drag (losing team)The player whose deaths, time dead, idling or leaving early hurt most: leaving before the end, being idle for 2 minutes or more, dying 4+ times while making up 30% or more of the team's deaths, or spending 2+ minutes dead with 30% or more of the team's dead time. If nobody crosses a line, nobody is named.

These ignore things the replay doesn't record, like positioning, shot-calling or who was targeted.

The replay doesn't say how well someone did their job, so we compare their numbers with other players on the same kind of hero. All numbers are per minute so game length doesn't matter, and games shorter than 5 minutes are ignored.

• Tanks: damage taken, time spent crowd-controlling enemy heroes, takedowns, and deaths (fewer is better).
• Healers: healing, teamfight healing, clutch heals, and deaths (fewer is better).

For each measure we work out the share of comparison games they beat, then average those. 80 or more is Excellent, 60 Good, 40 Average, 20 Below average, under 20 Poor. We compare with the same hero once there are 25 or more games on it, otherwise with every tank or healer, and show nothing until there are 30 games to compare with.

This ranks output, not decisions. A tank who soaks damage by standing in the wrong place scores well; a healer who is ignored by a team that never gets hurt scores low. Bruisers and other roles aren't rated.

Paper Tank and Missing Healer. These anti-awards use this same rating. A tank whose rating is under 20 gets Paper Tank, and a healer under 20 gets Missing Healer. Nothing else is different between them: the role decides the name. At least two measures must have data, so a game with missing stats gets neither.

• Who they're compared with: the same hero once there are 25 or more games on it, otherwise every tank or healer, and only when there are 30 or more games to compare with. A hero we've seen rarely is judged against a mixed group that plays quite differently.
• Because the rating is a ranking, some games will always land in the bottom band, even if nobody played badly. These awards are for fun, not proof of a bad game.
• The comparison group is made of the games stored so far and changes as more are uploaded, so a game's result can differ slightly depending on when it was worked out. Awards already given are never taken away.

A private list of everyone you've shared a game with. "With" counts games on the same team and the record is your result in those games; "Against" counts games on opposing teams and how many you won. You can filter by an award or anti-award to see who earned it in games you shared.

• Only games someone has uploaded are counted, so every number reflects the games we have seen, not a player's whole history.
• Silence penalties (chat and voice) are a snapshot taken at the start of the game.
• The replay has no health over time and no per-target healing, so we can't tell who a healer healed or whether they chose well.
• Thresholds and weights are tuned by hand and may change as we see more games.