What happens when I publish a commander?
Publishing makes your current commander available for ranked competition. Its brain and trained values determine what it does during those fights. You do not control it by hand in a ranked fight.
You can keep training locally. Those changes do not replace the published version until you publish again. Check which publication produced a result before attributing it to your latest training run.
How does Elo work in BotBellum?
Elo is a rating based on recorded ranked outcomes and the opponents’ ratings. It summarizes competitive results, not the size of a brain or its training score.
A win raises the rating and a loss lowers it. The size of the change depends on the expected outcome. Beating a higher-rated commander is a larger surprise than beating a lower-rated one. A draw can also move ratings when the commanders entered with different ratings.
Open Rankings to see Elo alongside wins, losses, draws and games played. A short record gives you less evidence than many fights across varied opponents.
Which activities change ranked Elo?
Only recorded ranked fights change ranked Elo. Training explores candidate versions on your device. Tests compares selected saved versions under controlled conditions. Play lets you spend energy and deploy units yourself against a published AI commander.
The training mechanism and player controls are being reworked. Keep using recorded ranked outcomes to judge competitive performance instead of treating a particular training score or algorithm as a promise of ranked strength.
| Activity | What happens | Changes Elo? |
|---|---|---|
| Training | Candidate versions are evaluated during a run. | No |
| Ranked fights | Published commanders make their own decisions. | Yes |
| Tests | Saved versions face matched opponent versions. | No |
| Play | You deploy units against a published commander. | No |
Why can win rate and Elo tell different stories?
Win rate describes outcomes in a particular set of fights. Elo also takes the opponents’ ratings into account. A high win rate against weaker opposition and a lower win rate against stronger opposition are not the same competitive record.
As an illustrative example, 6 wins and 4 losses is a 60% win rate. That percentage alone does not tell you whether the commander is stronger than one that won 50 of 100 fights. Opponent strength, game rules, dates and sample size also matter.
Watch available ranked replays to see how the commander spends energy, responds to a push and distributes units across lanes. For your own commander, open Analytics and select Ranked to scope published versions, dates and compatible game rules. Owner analytics are private and free. Public profiles show a more limited summary.
Why can training results differ from ranked results?
The candidate evaluated in training may differ from the version you published. The training and ranked records may also cover different opponents, dates and numbers of fights. Check those differences before concluding that a training change failed.
Seasons can introduce balance changes and other updates. Current-season commanders compete under the current rules, while earlier commanders can become archived. Compare compatible rules and check season context when reading a record.