The **Elite Solo Ladder** (**ESL**) is our custom 1V1 leaderboard for competitive (or just statistically inclined) players of *Dawn of War II: Elite*. Since the built-in leaderboard of *Dawn of War II: Retribution* has broken due to design flaws and discontinuation of support, we have made a functional one.

As modifying the original backend is not possible, the custom system must be external. For this, we employ a background app that sends results, metadata, and the replay files to our server, and use this to calculate skill ratings and collect *Elite* statistics.

Our skill ratings are calculated using the Glicko rating system (Glicko-1) instead of the more common Elo rating system. Glicko seems more informative and suitable to our player population.

The Glicko rating system (also called Glicko-1 since the introduction of the more complex Glicko-2 system) is a skill rating system developed by statistics professor Mark Glickman as an improvement over the older Elo rating system.

The Glicko system gives a similiar mathematical skill estimate as the Elo system, but also gives a numerical estimate of how reliable the rating number is. When players compete against others and their skill rating is adjusted, the reliability indicator – RD (ratings deviation) decreases, narrowing the probable range where the player’s skill is. If a player does not compete, the RD slowly increases to reflect that we grow uncertain of what their current strength is.

- Rating
- The player’s estimated skill.
- RD
- How reliable the rating is right now.
- 67% CI
- The 67% credible interval (rating ± RD). We can be 67% sure that the “true” rating lies in this range.
- 95% CI
- The 95% credible interval (rating ± 2 × RD). We can be 95% sure that the “true” rating lies in this range.
- CR
- The
**credible rating**(rating - RD). By ranking players according to the 67% credible limit instead of the unadjusted rating, we can have a reasonable guess about their strength. This prevents high RD players from rising to top ranks too easily.

New players are entered into the system with 1500 rating and 350 RD. This puts the 95% CI range all the way from 800 to 2200 (1500 ± 2 × RD) which is pretty wild, as we have no idea how good the player is.

As players keep competing, the skill rating, naturally, rises or drops according to wins and losses. The RD decreases as the system gains more data about the player. In chess rating systems, an RD of 80 or less is often considered reliable and a sign that the player is competing actively. Players with a high RD are not shown on the ratings list until their skill estimate gets reliable.