MZ's Lessons: The Coinflip Survey
OtakuMZ unpacks an extensive experiment of data collection surrounding the coinflip and lays out the results for use in the ongoing discussion. Let’s dig in!
The Coinflip is one of the most discussed topics in Gwent and the Gwent subreddit attracts numerous posts regarding ideas to fix it. As with any other card game, there is a difference in win rate (WR) observed by many players depending on who goes first or second. Most games have a system in place that diminishes the advantage (or disadvantage) to an acceptable level. Depending on the CCG you are looking at, reported differences in WR varies. At the lower end, a small percentage seems acceptable, but 10% and above is definitely not. If you believe the voices of the community, the problem is particularly accentuated in Gwent as there is no compensation system in place for the coinflip as some other games possess. Going second and especially being able to play the last card in the game is very important in Gwent. The system of only playing one card per turn enables the player with that “last say” to deliver the final counter, e.g. killing off a pair of buffed Dol Blathanna Protectors with a decisive Scorch in contrast to just one. So how to tackle this debate with hard numbers? Let’s get to it.
To date, there is little data available that proves or invalidate this thesis for Gwent with the biggest dataset reported numbering below 450 matches. Due to recent tournament data, one could believe that the problem is non-existent after the Gold immunity patch. Before that particular patch, data from the GwentOpen #1 correlated with a large advantage in favor of going second with a 72% winrate in this tournament. The sample size was low, though, with only a bit over 30 games in the whole tourney. Match data from the competitions (GwentSlam #1-3, GwentOpen #2) after the aforementioned patch suggested that the problem of the coinflip ceased to exist for the most part with nigh-exactly-balanced 50:50 WR. The sample size of all four tournaments was reasonable with around 125 matches in total.
Aim of This Survey
This study should help to shed some light on the state of the coinflip.
· Is the coinflip still a problem?
· If it is, how big is the difference in WR compared to other CCGs and TCGs?
· Are there faction and leader specific differences?
I collected match statistics from patrons of the GwentUp tracker and streamers that have access to the the statistics tracking tool as it can deliver all the data needed (see screenshot below). Shoutouts to all the players that were kind enough to share their data (in order of data delivered): unlucker_dog, Fewof, Shasta, Airballshooter, praedo, thekeyer, Oceanmud, N.O.D., dTitan, glamps, Vanitas Cabal, Nirv, ShadowplayRed, Gheed, Gravekper, and Snake.
Data from 4632 matches was collected with 2335 matches played first (blue coin) and 2302 played second (red coin). There was no significant difference between blue/red coin data subsets (p> 0.05; two sided t-test). Further subsets were created for every faction. There were 500+ games in every faction. First and second played games inside each faction are also balanced with no significant difference.
Regarding leader distribution, most leaders were represented sufficiently well to analyze data with the exception of Dagon, John Calveit, and Harald the Cripple. Those three had very small datasets with 10, 20 and 16 matches respectively. Hence, win rates of those three leaders cannot be statistically evaluated and they were excluded from interpretation.
Overall WR of the matches collected was 59.11%. Optimally, it would be 50%, but as data has been collected from players that are invested in the community and patrons to GwentUp those tend to have a higher skill level than the average one. Due to the overall positive win rate, there were more games won that were analyzed than those lost.
Looking at winrates, as expected, going first (blue coin) was associated with a lower win rate than going second (red coin). On average, the difference was 14.43% with the highest represented in Monsters (17.65%) and the lowest in Scoia’tael (6.67%). Scoia’tael was a real outlier and the only faction that showed an almost-acceptable win disparity.
Looking at leaders, Unseen Elder (18.36%) presented the highest dependency on the coinflip followed by Crach an Craite (18.08%) and Henselt (17.85%). The lowest impact on the going first or second could be observed in Eithné (4.46%), followed by Francesca (7.61%) and Foltest (10.05%). Unseen Elder had the smallest dataset apart from the three excluded leaders with “only” 261 matches total.
You can see, that the problem IS indeed massive.
The data presented here is by far the biggest collection of match data published so far. In a previous report by Reddit user Luciferrrro, 238 matches (blue/red – 127/11) from various streamers were analyzed leading to a disparity in win rate of 15.2% (37.0%/62.2% for blue and red coin respectively). The overall winrate was 48.7%. Garrett_O23 commented in the same post a bigger set of 423 games with WRs of 58.5% and 70.0% respectively resulting in a difference of 12.5% which is a bit lower than the number from previous mentioned dataset. Even though the studies were a lot smaller than the data presented here, the outcomes speak the same language with 5.5%, 12.5%, and 15.05%. Before judging this to be fair or unfair, let’s take a look at other games.
Hearthstone (HS): I chose HS as a comparison as I know it best besides Gwent. First of all, as in most of the CCGs/TCGs other than Gwent, it is considered favorable to play first, not second. The balance in most TCGs like MTG is done by “draw first or play first”, granting an additional card to the player going second. HS has an additional compensation is in place. About a year ago, Vicious Syndicate analyzed 200k games on that matter (51.65% vs to 48.35%). I have also written a review on that topic regarding different game modes. TL;DR, the WR disparity is about 3.3% for HS which is considered very fair. Depending on the matchup, it can even be favorable to play second, as the deck/class you use can abuse the compensation mechanic - The Coin - which provides additional mana as well as the additional card. In general, aggressive decks want to play first, control-ish decks tend to be neutral or slightly favored when playing second, and combo decks usually prefer to let the other player start.
Chess: If we look outside of card games, Chess as one of the oldest and renowned competitive games has a 52-56% win percentage for the player going first. [Source]
From this example of a newer CCG and a very old game we can see that something around 2% to 5% can be considered balanced. It lies in the nature of turn based games, that there WILL BE always a slight advantage or disadvantage depending on the coinflip. According to Rethaz from CDPR, the difference between going first and second is ”often largely over estimated. […] there’s a reference to a 10% or even up to a 20% difference, the difference (at the high end of ranked and also everywhere else) is far below this.” [Source]
The data presented here suggest something else, but the (very few) games from recent tournaments suggests that the truth of the matter may still be on Rethaz’s side. It is open for debate, if a game that is balanced for top 100 or so players, but that is heavily skewed for the rest of the player base is really balanced. Although tournament data suggests a good balance, recently CDPR introduced a fixed alternating coin distribution. Critique to this system is, that knowing when you go first or second you can select a specific deck that is better in this circumstance. On the positive side, this adds another strategic element to the game - and at a tournament level, removes a significant amount of variance that is controlled be neither of the players in favor of letting player skill have the greatest impact possible on the outcome.
I am a bit torn. When I dug into this topic with analyzing GwentOpen and GwentSlam data I was very optimistic that the problem is overhyped. Now that I concluded this survey, this optimism is diminished. A near 15% disparity in win rate just due to the luck of going first or second? I feel this is too much - and I know I’m one voice among the chorus to express this sentiment at various points across Gwent’s history. Like many players, I think there should be a balancing mechanic introduced to the game that lowers the impact of going first or second.
How this should play out, I leave to decide to the developers or people who know a lot more about game design than I do. Regarding the timing, I hope they do it before the beta ends and rather sooner than later. In any case, I am confident in CDPR and the Gwent team that they find a solution. After all - they created this fantastic game in the first place, and every update has proved to be a step in the right direction.