Stats in Gwent: Overview and IdeasPosted by Gwent July 5, 2020 in
Hello Friends of Gwent! Today I’d like to present to you the third article in the ‘stats’ cycle. In Nations of Gwent 2 we analyzed the Pro Ladder nationwise, and even put up Gwent World Map. In “Unofficial Faction Stats” we used data collected from public profiles to get some information on factions popularity and performance in the Season of the Magic.
This time I am going to describe in depth the general topic of stats in Gwent. What type of stats do we have in competitive Gwent? How relevant are these for players performance? What is ethical and what not? How stats could be used to improve player experience? I will try to answer all these questions and many more.
All the Data in Gwent
What type of data do you generate as an ordinary Gwent player? Obviously every game we play provides oceans of information. We start the game at a particular daytime (which enables who-plays-when analysis, like in Nations of Gwent 1). While queing we decide to play particular faction (popularity and pick stats, like in “Unofficial Faction Stats”) and leader. Our initial rank is determined (on pro ladder visible only to player himself). Then matchmaking process starts. Rating difference between matched players and time elapsed could be used to analyze the system itself. This way developers could improve algorithm, and players could try to find its secrets or avoid waiting too long. Game starts with coinflip, and we can see opponent faction, which determines the matchup. Coinflip itself is a simple random generator. However, it could also be more sophisticated to ensure that result are as close to 50% as possible in let’s say 200 games. The matchup is probably most important thing when it comes to competitive analysis, however given factions and leaders could play different decks which is another layer of data. Then comes the gameplay itself. This layer is the hardest one to analyze statistically; analyst needs be very smart to determine characteristic features for a given matchup (drawing crucial card yes/no and so on), which has to be decoupled from quality of play among other things. Gameplay analysis even with full possible information is rather human and artificial intelligence domain. When the game is finished, we can record match result and match duration. Match result is crucial for simple competitive analysis, as it is the only output for a big set of input values. Match duration data is not useful for players, but could be very important for planning tournaments or identifying a wintrader, whose opponents concede too early. After the game, fmmr changes are determined by a secret algorithm. Players could collect changes data to approximate this algorithm as well as possible. Developers could not really make more use out of it than validation. Each player profile changes, fMMRs, peak fMMRs and total MMR may change too, while number of matches of course has to change 😉
In principle, all the data is accessible to the developer, who decides himself on how much data is needed and useful. How about players? If recording own game data, player gets full information on everything, with few exceptions. Rating difference is not known (opponent fMMR is hidden), and could be only predicted based on fMMR changes. Neither opponent deck, nor gameplay (opponent mulligans…) are perfectly known, but great deal is visible.
Is self-recorded data useful? For self-development – with no doubt. But the sampling is very poor (let’s say around 200 games), and more importantly – slow. The meta state in Gwent could change very fast. Nilfgaard played a week ago is different from Nilfgaard played now. Stats quickly become useless. We tried manually collecting our own data in Legacy, but the work was tiresome and we had to give up on the idea fast. If medium-sized team is unable to put something like this to work, doing something similar by a single player is out of question.
As it is impractical to collect all the data accessible to a single player, some information has to be dropped. These information will never be retrieved or published without use of tracker. Alternatively, the data could be leaked by developers themselves. An example of such leak are leader stats.
If the data witnessed by player is not enough, other sources has to be used. There are three official sources: in-game leaders table, Gwent Masters Ratings and public profiles of players. In Pro Ladder player practice, data is often acquired also simply by watching streamers, what amongst other things gives some information on the temporary state of the ladder.
In-game leaders table gives information on wins/loses/draws, total MMR and best factions of players. Funny enough, best faction of players with private profiles could be accessed only in this, most simple way 😉
Gwent Masters Ratings site is poorer in data than it used to be. The main additional information are nations of players. National stats could be done based on this website solely (NoG1, NoG2)
Public profiles offer great deal of information. Especially, all the detailed faction stats are accessible only there in terms of total number of games played with a faction and the number of wins. The “Unofficial Faction Stats” article was possible only thanks to Pro Ladder players who kept their profiles public in spite of various risks which would be discussed later.
If going for big data analysis, there is no better way than using bot scrapping data from public profiles and Gwent Masters Ratings. If going for just a chunk of useful data for competitive play, it is worth to look at best players with public profiles (especially streamers) and try to yoink their successful decks ;-).
'Ethical' and 'Unethical' Data
According to test we did few months ago, profile data is live, which means that queue-sniping every player with public profile is possible. Just keep refreshing your victim’s profile, and when numbers change, start the lobby with deck countering opponent’s faction. There is big probability your victim queues at the same time.
Such an asshole behavior is especially possible for very high ranked players, who have great chance of being paired together. Obviously it is one, and by far the main example of possible unethical behavior. If you ever try to do something like this, do not excuse yourself by saying that your victim made profile public on own will. You should be smashed down to the ground.
Maybe it is lack of our imagination, but we don’t see other unethical possibilities for using data as accessible now.
Does data give competitive advantage?
Hard to believe it at first, but the answer is no, as long as it is used in 100% ethical way. Data gives some information and could help to identify what is played in general, and what options are successful.
We have few months experience of reading data, and we did not manage to use it to our advantage a single time (no, we don’t possess the bot Slama was talking about ;-)). The information granted by profile data is in reach of every pro player involved enough to watch streams and discuss meta with each other. Such information is even more complete, because involve leaders, not only factions.
What is really crucial for the ladder success is perfect knowledge of meta, matchups, and simply the gameplay skill. Be like TailBot – think on the game on yourself – and sooner or later you will be rewarded much more than by following the stats and data.
Data and Players Experience
A lot of time elapsed since the introduction of the “Hidden fMMR” policy, so we had a chance to get acquainted with the new situation. I see both pros and cons of this policy now, but I can be a bit biased towards pros due to building decks on my own and having access to more data than casual player. I think that any argument on this issue is rather pointless at the moment – a path was chosen and there is no reason to go back.
At the same time I think that the data should be visible in the old format after season ends. This would make all the afterseason analyses done by community much easier and richer. The argument that the data could still be relevant for the next season if there is no important changes in patch was lifted by top players themselves, contributing with their profile pics to Damorquis video on best players (no-changes patch was announced at this time).
We shouldn’t also trick ourselves into thinking that Pro Ladder is a swarm of wolves fighting for top spots at any cost. In reality there is around 100 players competing for Top64, and around 30 players fighting for Top16 each season. It is just 1% of Pro Ladder. The remaining 99% do not even think about esports, but wants to have as much pleasure in the game as possible.
A nice and unexplored so far possibility are faction ratings. I would really like to see a kind of Factions Hall of Fame, and maybe even separate ratings for each faction at the season end. It would add a lot of spice to the game. Fighting for best faction score is great fun for people who mains/loves one faction in spite of doing placements on four. It would also be a platform on which top players could compete with casuals, who have time only to climb with one faction.
With factions or not, Hall of Fame would be another quality of life improvement. I often want to see Top3 of past seasons, but there is no proper page. Putting all multi-seasonal information in one place, spiced with a pinch of glory, would be awesome.
Thanks for reading! I hope you are not going to change your profile settings to private immediately after reading this article 😉 Moreover, if you keep profile private during the season, please consider switching it for few hours after the season ends – this is the moment when I acquire afterseason data for stats articles.
That would be probably my last ‘stats’ article in this season. I’d like to hear your feedback on my work, maybe there are stats you are particularly interested in? Just let me know in comments or on Discord. See you on ladder!
Written by lerio2
Writing in-depth Gwent articles is fun, but time consuming. If you like ‘Alphabet’ or ‘Gwent for Geeks’ cycle or other articles, and fancy to support my efforts – here is the way to go: https://www.patreon.com/lerio2