school winrate

[galehar]

@quinarbre: do you have access to the global winrate per school? That would be interesting.
Personally, I'm struggling with Sylvans. I think I have a much lower winrate with them.

[Quinarbre]

Here's what I have:

• Choice rate per school, among all players:

  Code: Select all

  Imperial 	0.39
  Highland 	0.22
  Sylvan 		0.21
  Everfrost	0.18

  Fairly balanced, with a big bias towards Imperial because of the duplicate deck, and a small one against Everfrost because it was added one year later.
• Choice rate per school, among winners:

  Code: Select all

  Imperial 	0.38
  Highland 	0.23
  Sylvan 		0.21
  Everfrost 	0.18

  There are biases, but since they match the previous ones perfectly, I'd say the factions are balanced.

That's over 100.000 games played; here are a couple of facts that could introduce a bias:

• 85% of games are played in High Form, 10% in Deathmatch and 5% in Melee (ok not a real bias, everyone knows High Form is how the game should be played )
• In 64% games the players choose their factions, and consciously or not, they are probably inclined to pick the ones they succeed most with. In the future this could be mitigated by random factions being the norm in the Arena.

[vagabond]

I'm curious if there are other statistics available, like win rate for each legendary? I'm particularly curious about Time Elemental's impact on win rate.

[Quinarbre]

Not directly, no.
When we first made the game public, I remember there were discussions with CGE to give them a full database dump of the games played, so theoretically they could have that kind of information available, but I'm not sure it's worth taking the time to fetch it. The takeback mechanic also makes it harder to keep track of the actual use of a given being in a game.

[butki]

Hi all,

I would be interested in updated stats about the school win probability with nethervoid and etherweave included, if you can still share it with us

thanks!

[Quinarbre]

Here are the current rates, over the last 1000 games:
Played with:

Code: Select all

Imperial   0.32
Highland   0.17
Sylvan     0.17
Everfrost  0.12
Nethervoid 0.11
Etherweave 0.1

Won with:

Code: Select all

Imperial   0.35
Highland   0.14
Sylvan     0.16
Everfrost  0.14
Nethervoid 0.12
Etherweave 0.09

General table options (random or player's faction choice, game form) remain in the same kind of levels as before.

All three expansions were marked as "unrecommended for beginners" and can be disabled in each player's options (about 17% people disable Everfrost, and 22% disable Nethervoid or Etherweave), which can account for their relatively low choice rate.
It it interesting that so many people are reluctant to use these factions, be it by default or when faced with choosing them in-game.

Win rate is still consistent with choice rate (the rates being now computed over 1000 games only make the numbers subject to a higher fluctuation, a 0.03 discrepancy is not significant).

[T72on1]

Nice, thanks for the numbers.

I personally struggle with Sylvan and Nethervoid. But that has everything to do with me not being experienced with them, and even with Tash-Kalar in general.

[Superluminal]

Hello everyone,

I just stumbled upon this thread while specifically searching for discussions about Tash-Kalar statistics. This topic is very interesting to me and could be beneficial for those looking to improve their game.

With 8 years of experience in development, including over four years in data science and analysis, I believe I can provide significant insights if I can access the data. I am very interested in this topic and am willing to share my thoughts, findings, numbers, and graphical visualizations with the community if it's possible to access the data.

I think there could be really interesting discoveries hidden within the raw data that can enhance our gameplay. The native statistics from BGA are quite poor. Quinarbre has provided some new insights, but I believe there is much more to be extracted. It would be very interesting to know numbers like:

- Win rates for each pair of factions. It's possible that you will have an advantage choosing a specific faction if you know the faction of your opponent.
- The potential impact of certain cards on win rates.
- Effectiveness of different schools and strategies across game modes.
- Expected win percentage based on Elo difference.
- Identification of which card combinations are most frequently used together, revealing popular strategies or synergies.

Additionally, we can categorize the results by game modes (high form, deathmatch, etc.), player Elo strength, game type (play, arena, tournament) and (real-time, turn-based), player count.
We can also track if numbers change over time. Moreover, I think the games need to be filtered based on the game termination (e.g., some games are terminated by time) and potentially other impoortant filters can be found.

Depending on what data is in the database, there might be more valuable information that can be extracted.

@quinarbre, I believe it is possible to transform the data, accounting for the takebacks to make the analysis much easier. I think I can take on that challenge.

What do you think?

Here is my portfolio for some credentials regarding my skills:
https://sergeygerodes.xyz/