May 12, 2021

Update on Matchmaking improvements


Dear Hunters, as you know, we recently introduced some matchmaking changes in Update 1.51. Since the matchmaking changes dropped, we have been monitoring your feedback, analyzing the game data, and examining how the changes have played out in the bayou.

You can get the full lowdown on the matchmaking changes from Update 1.5.1 in this blog. However, today, we want to explain how the changes have affected the matchmaking game data, what we have fixed already, and what we may change in the future to improve the experience for everyone.

So, after Update 1.51 rolled out, we, unfortunately, saw an increased amount of unbalanced matches in the data in the following days. We then released a number of fixes, some of which have already had an impact.

First of all, we modified the ratio for team-based matchmaking. The matchmaking system uses the group average PvP(player versus player) rating to find a suitable game. Within a group, the PvP rating for individual players can be fairly different. For example, within a group of two players, you could have one player with a very low PvP rating and one with a very high PvP rating or two players with two average PvP ratings. That's a balance to strike when it comes to matchmaking. The ratio is a hidden multiplier calculated into a team's rating to accommodate for their relative strength compared to other groups. By reducing this ratio figure, we have seen in the data that this has helped prevent groups, especially groups of three, from being pulled into higher rating matches, thus causing an imbalance.

Next, we decreased the strength of the ELO decay, another feature introduced in Update 1.5.1. ELO is a method of skill-based matchmaking drawn from the world of chess, and you can read up in-depth about how it is implemented in Hunt here.

In broad terms, using the ELO method, we generate a figure that should accurately reflect your individual skill level within the community by looking at your player versus player performance.

The ELO decay figure takes into consideration players who have not played Hunt for a while. We assume that if they have not played recently, they may not have the same PvP skill as when they were playing previously, thus lowering their PvP rating slightly over time. There are two coefficients for players who have absences from the game, set at seven days and thirty days, each of which slightly lowers the player's PvP rating. After the player has come back, they are no longer in the state of "ELO decay." We adjusted this figure by decreasing the strength of the "decay," but this tweak did not significantly reduce the number of unbalanced matches showing up in the data. We will keep monitoring the numbers and ratios implemented and may adjust them in the future as needed.



We then decided to increase the number of brackets in all regions, especially for high skilled players. We found that very high-ranking players ended up in matches with players with lesser skills due to the size of the higher brackets. We have now separated these extremely high-ranking players out into more brackets. This has resulted in more balanced games, lowering the difference between the PvP rating of different groups in a match. This was one of the largest areas of concern showing in the data after Update 1.5.1, and we can already see more balance come through in the current data.

We would also like to discuss the matchmaking system around random groups. Random groups do not use a PvP rating to find a suitable partner at this time. Currently, first, the system looks to see if a potential partner has Prestige or not. If you are a Prestige player or have a 100 rank, we will aim to group you with other players who have achieved Prestige or a 100 rank.

If you have neither Prestige nor a 100 rank, the system will aim to match you with players who also haven't achieved Prestige or such a high ranking either. However, if suitable partners are not found, the system will then look for anyone in the queue in the same region and same type of group as you, for instance, a group of two or three.

Because we prioritize finding a game quickly rather than waiting for a perfect match, it can mean that when you are playing with random players, you may match with a player who has a very different PvP ranking.

At the moment, when you get matched with a random teammate, or an invited one, you do not know the exact matchmaking ratio of your teammates. Because of the logic described above, your skill may be lower or higher than them. However, the enemy team's average will be on a similar skill level after the matchmaking system has performed its calculations. It should be noted that team invites have a slightly higher rating compared to random groups due to the possible pre-existing synergy and communication between teammates.

However, to address this, we may try to match random teammates with more comparable skill levels to achieve more balanced games in the future.

We do not currently expose player or team PvP data generally in the game. However, you may have seen a PvP rating on the death screen. This rating compares you and the person who killed you, and it does not consider the team's average PvP rating. For example, you could be killed by someone with a significantly higher ranking who is partnered with someone with an extremely low ranking. The overall ranking of the group explains why you might appear to have such a personal 'one-on-one' mismatch on this screen.

This is also an area where we plan to make changes. After a match, we intend to show you information about who was in the match and which team they were in. By doing this, you will see the rankings of all players and teams and how they compare to you and your team. We haven't yet decided how to show this in the most user-friendly way, but we think demonstrating the relative skill of all players and teams will help give you context for the results of each and every match and your achievements in them.

We may also remove the ranking comparison in the death screen, as we think this info might be better explained with the full context of the ranking data described above. Instead, in the pre-game lobby, you'll see ratings and rankings for both you and your teammates, and in a new post-match summary screen, you will see more information about all the players and teams in each match, including the person who killed you (or you killed!) We'll have more news on this when we decide what we think will be the most effective system.

We are also reviewing the group calculation to see if we need to put more weight on a higher PvP rating instead of averaging out the team's rating. This work is happening now, and if we find a reasonable calculation, it is a potential change that could arrive in Update 1.5.2.

We believe that transparency is important. Our goal with Update 1.5.1 was to improve the matchmaking system and, consequently, improve the Hunt experience for everyone. Ultimately, you never really know how things will play out until you can see the data. Hopefully, you have seen the effect of the tweaks and changes we've already made. Be assured that we will continue to analyze the data, listen to your feedback, and make further changes, including those outlined above, as we go forward. Game development is ultimately an art, not a science, even when the data heavily influences decisions. While we have seen some steps forward since Update 1.5.1, our work doesn't stop to make the matchmaking system the best it can be. And your input is vital on this journey.

See you in the bayou,

Your Hunt Team.

++

Don't own Hunt yet? Get your copy for PC on Steam or from the Crytek shop. For Xbox, head to the MicrosoftStore, and for PS4, find Hunt on the PlayStationStore.

Got queries or questions? Stop by our social media channels and let us know:

Discord + XboxClub + Facebook + Twitter + Instagram + Xbox + Steam



We use cookies on this website to improve your experience, analyse our traffic and integrate with social media. You may adjust your cookie preferences by clicking “customize”. Please find further information in our Privacy and Cookies policies.

CustomizeAccept