Creator Code
RoyaleAPI
Deck Search - Sort by Rating
Starting from today, our Deck Search will use Rating as the default method to sort decks. In this post, we will explain what this [...]
site new feature
Starting from today, our Deck Search will use Rating as the default method to sort decks. In this post, we will explain what this number is, and why we have decided to use it.
Background
We have traditionally had two main sort method: popularity and win rate.
- Popularity shows what is meta.
- Win rate shows the probability that you will win with the deck.
These sort methods work well when all the decks get played in lots of battles. However, as you drill down with our plethora of filtering options: card inclusions, battle types, date range, etc, you will soon notice that neither sort method works very well when the sample size goes down.
Rating - New Statistical Model
Today we’ve introduced Rating, a statistical model that we have developed that uses both Usage (aka Popularity) and Wins to calculate a unified number to denote the strength of a deck. The advantage of using this method can be easily understood when we illustrate it using some hypothetical data:
| Deck | Battles | Wins | Win % | Rating |
|---|---|---|---|---|
| A | 900 | 850 | 94% | 93 |
| B | 10000 | 9000 | 90% | 89 |
| C | 20 | 20 | 100% | 84 |
With these samples, you can see that while Deck C has a 100% win rate, it is not necessarily the strongest of the three, given how small the sample size is.
On the other hand, using Rating to sort these decks reveal a far more statistically meaningful result, and here showing that Deck A is in fact the strongest among the three, followed by Deck B and Deck C.
If you don’t understand this completely, just know that this new default sort order will rank decks in the most optimal way.
Implementation
The rating system is now implemented in Deck Search for now, but we will soon utilize it in other areas as well, such as Card Stats, Player Deck Stats, etc.
We will likely tweak our formula over time, but we have confidence that this new system will help you find the best decks within the game!
Appendix
Deck Search
Here are some sample searches where the sample size is small enough for you to see how Ratings make a difference:
- Time: 7 days | Cards: Lava Hound, Clone | Game Mode: Ladder
- Time: 28 days | Cards: Lava Hound | Game Mode: Grand Challenge
Data Samples
Here are more hypothetical numbers for you to compare the differences:
| Battles | Wins | Win % | Rating |
|---|---|---|---|
| 1000 | 600 | 60 | 57 |
| 100 | 60 | 60 | 50 |
| 1000 | 500 | 50 | 47 |
| 100 | 50 | 50 | 40 |
| 10 | 6 | 60 | 31 |
| 10 | 5 | 50 | 24 |
Feedback
If you have any questions or comments about this feature, please write to us on this Reddit post
Social