How are ratings calculated, and what does provisional mean?

There are two kinds of rating: one for your bot on each ladder, and one for each of your teams in tournaments. Both work the same way.

Reading a rating

A rating is shown like 1620 ± 95.

• 1620 is the rating. Beating a stronger opponent raises it more than beating a weaker one.
• ± 95 is the rating deviation: how uncertain the rating is. The smaller it is, the more settled the rating.
• provisional appears while the ± is above 110. On the ladder, established ratings are listed above provisional ones.

Ladder ratings

• Each rated series counts as one result: your share of the 4 matches' points (a win 1, a draw ½), against your opponent's rating.
• You have a separate rating on each ladder you're entered in (each game).
• Submitting a new version of your bot makes your ratings less certain, because it's not quite the bot it was. Your rating stays, but the ± grows, so it can move faster.
• Switching language moves you to that language's league with a fresh rating (1500 ± 350).

Team ratings

• Each tournament counts as one rating period: every game your team played in it counts, a win 1, a draw ½ and a loss 0.
• A new team starts at your best established team's rating, with a wider ±. If you have none yet, it starts at 1500 ± 350.
• Changing a team's bots makes its rating less certain too.

Taking a break

Your ± grows a little for each day you don't play, up to 350. A rating you haven't played on in a while is less certain, so your next results move it more.

Ratings and tournament tiers

Novice and Intermediate tournaments are only for teams rated under a limit, said with each tournament. For entering:

• a provisional team counts as its rating minus its ±, so a new team isn't kept out by the 1500 everyone starts at;
• a team never counts as less than 200 below the best any of your teams has reached, so starting a fresh team or losing on purpose doesn't get a strong player into an easier tier.