Missing the point of what Iv said because you focus on the ‘pointless meaningless numbers’ that you view them as.
As you miss the point because it means nothing to you, you are never going to understand the frustration many are experiencing. Therefore its redundant going in circles with you even though its been explained rather clearly but you are choosing to ignore so you can just keep saying ‘it doesn’t mean anything’.
Read my other post again, let it sink in and if you really can’t grasp the issues around just the ranking structure then kindly spare the time of replying. I explained how the gaming portion is no longer fun for me outside of ranking and even Arena mode.
Today my last 2 matches out of 6 was +24 and -26 because i finally played with Silvers.
Bottom line if you play with you rank you will gain/loose +/- 25
But if you are being matchmade with bronze while you are silver you will get only 1-2-3 and we all know why.
Unfortunately nothing we can do here.Playerbase is small so you will either have to wait like 40 minutes in the Search to find Silver players or after 5-6 minutes of not finding looking for lower.I don’t think there is much to be done
Firstly, thanks for the comprehensive write-up. Secondly, thanks to everyone for their patience with us despite our radio silence. We’ve been doing a lot of monitoring over the past (almost) week, and we’ve also received reports internally from folks who are experiencing some of the same issues y’all have been. Let’s see if we can answer or further elaborate on the systems in place to give you a better understanding of how things work. Settle in, I can already tell this is gonna be a novella.
Let’s start here. To begin, we use a modified Glicko-2 rating system. The Glicko system itself is a modified version of the Elo rating system that many of you are probably familiar with. If you’d like to read more about the rating system, here are a few helpful links:
It ends up being a lot more complicated than victory = points earned, loss = points lost, when you peek under the hood. To break it down as simply as possible, here are the factors that are taken into consideration when a match in Evolve concludes:
Your current rating (AKA “score” AKA “points”)
Your opponent’s score (if you’re a Monster player, it would be the Hunter team’s collective average score)
The Glicko formula’s confidence level (known as a Rating Deviation – RD) in the outcome of the match based on the history of your and your opponent’s matches
The Glicko formula’s volatility variable, which is based around how consistently you perform in your matches, or if you’re all over the place as far as wins/losses against opponents the algorithm expected the opposite result of.
*Quick side note: I’m by no means a mathematician. My explanations above may be a little off-base, but is what I’ve gathered based on comprehensive reading of the materials I’ve linked, and what’s been explained to me internally. If there’s any players that do have a background in this kinda thing and I’ve butchered it, I encourage you to correct me so we’re all more aware!
What’s not taken into consideration are the following:
How fast the match was won
What Stage you or your Monster opponent were at when the match was won
How the match was won (Monster death, Hunter team death, wildlife OP, relay destroyed, etc.)
Which character participated in the match
Anything else not explicitly listed in the “here are the factors that are taken into consideration list” above
Now that this is out of the way, let’s go back to Placement Matches…
Placement Matches are the “curtains” we use to hide the beginning stages of the Glicko system, as it’s trying to figure you and your patterns of wins/losses out, and what your rating should be. If you saw it, it’d look like it was jumping all over the place and be pretty stressful to witness without knowing the inner workings of the system. I’ve been told that at about 15-20 matches, the algorithm should be less swing-y, meaning that the Rating Deviation and volatility factors should be much more accurate and you aren’t losing/gaining a massive amount of points at the end of every match, if you performed as the algorithm expected. However, we thought putting players through 20 Placement Matches felt cruel and like a large barrier of entry. It was 5 internally (to help QA test it faster), but we opted for 10 at launch, in hopes that it would cover up most of the jumping around so it wouldn’t be as stressful to the player to watch themselves launch up and then drop back down before finding their “settling point”.
When you start your first Placement Match, you are set at what we anticipated would be the middle-ground of the playerbase, 1500. This is the tipping point right at Bronze Destroyer and Silver Skilled. Every Placement Match played results in your rating jumping up or down, based on the outcome, and after 10 matches, it’s presumably found a relatively accurate division you’re supposed to be in, before you start trying to climb up to the top.
We had opened up the matchmaking in an attempt to alleviate wait times for players. We found that, for Hunters, this seemed to work a bit, however Hunter teams were a mish-mash of players, and playing with Hunters of a lower-skill level usually led to the team losing, which would drag the higher-rated Hunters down with them.
For Monsters, we have found that there are actually more Monsters queuing to play than there are Hunter teams available, resulting in longer wait times for the Monsters. This also meant that by the time that Monster was “next in line”, their matchmaking pool was wide open, so it’d take whoever was first available. As a result, Monsters in the Silver/Gold divisions were frequently getting matched with Hunters rated far below them, making the climb upwards tedious and frustrating (and for the Hunters they were rolling over, it was not an ideal experience).
We’ve started to slowly restrict the matchmaking once more for both Hunter teams and Hunter-Monster matchmaking with the hopes that it will give you a fun, challenging match. There was a slight error in the matchmaking that was affecting match times for about the past 24hrs, but we found that and have corrected it, so please give it another shot!
One additional issue we’ve been seeing that you didn’t mention (but I want to, as I figure others might come by this post, looking for comment about it), is that they’re a higher-rated Monster or Hunter who gets matched against an opponent that is of a similar rating and they don’t earn as many points as expected. Conversely, they get matched against an opponent who is far below them and sometimes they earn a really large amount of points (I believe one report was something like 400+). We’re continuing to investigate these issues, as this is not intended behavior, and falls outside of our expectations. We did make alterations to the Glicko-2 algorithm, so it’s possible that something isn’t calculating as it should. We’re going to be running additional simulations in various scenarios internally to see if we can figure out what’s going on there.
As for the rest, I don’t have any further information to share at this time, but we are currently working on a hotfix build that I hope will add further clarity for when we need to balance our Hunter to Monster queues, in addition to resolving some of the other points you brought up.
Thanks again for your feedback and continuing to give Hunt 2.0 a shot! We’re working on it every day and will continue to keep a close eye on things so we can make it more fun and rewarding.
Firstly thank you for taking the time to write that all out! (Go get yourself some hot choco, sit back, and enjoy the gratitude that will surely be bestowed upon you for answering many of the communities queries in one fell swoop )
This answers a lot of questions and confirms some of what I suspected was going on behind the scenes of the rating system.
I’d just like to be sure, but what this means is that if you have a history of beating players a rank lower than you (say you won 90/90 games against Silver skilleds), and then lose a single game against them, you will lose fewer points [e.g. -12] than if you had say a 20/90 games won against the same team? Therefore as you play more games, the Glicko formula takes into account your history.
Basically, if you lose against a lower rank lots of times, you will lose lots of points so that you are brought closer down to a rank that the Glicko formula believes you will perform more consistently in [~50% W/L ratio]? Is that accurate? :s
Again, thank you for the knowledge drop. I always love when such posts are made. I love nothing more than knowing how the little things behind the scenes work.
First, thank you for finally posting some inside knowledge of how the ranking system works.
BUT, I don’t know… They say it’s more complicated than win = gain points, loss = lose points, but it really isn’t. The amount of points you gain or lose is calculated with an algorithm based on your past performance and the score of your opponent. Big deal.
If it was an ideal ranking aystem, it actually would be complicated by taking into consideration your personal performance. If the ranking system was effective, members of a hunter team could gain or lose points all during the same match regardless of winning or losing.
Remember that time you dealt 13k damage to the monster while heal bursting your team to good health as Slim only to lose because your assault wandered into a tyrant and your support thinks it’s his job to cloak and run away when the first person goes down? You should gain points and your assault and support should lose points so you don’t have to get matched up with people like them again.
Winning/losing should count for something, but it shouldn’t be the end all deciding factor between gaining or losing points.
Since they’ve already set the ranking system up based on wins, we’re not going to see individual stats incorporated into the ranking now. That’s disappointing. The new system is still better than no ranked play at all, but it’s flat out dissapointing and was poorly chosen from the onset.
If I win a game within 3 minutes, I won’t have a high amount of raw damage. Same for a medic healing. If the Support does well shielding and no-one loses much health, the Medic won’t gain much, or may even lose points, simply because they didn’t have the opportunity to heal as much. -.^
I’d love for personal performance to be incorporated to a larger degree, but that’s not easy to measure when a match can last anywhere from between 3 minutes to 25 minutes (Although technically the maximum is 40 minutes). Then you may say, take the damage/healing etc relative to the length of the match. The issue then is, what if the monster flees till 3? What if you aren’t able to catch the monster because the Trapper doesn’t dome, therefore you can’t damage much? A lot of variables can be present in a single match. I’m no expert in ranking systems and the algorithms available/used. Is it possible to factor all of these things in in numerical values that are accurate? -.^ I imagine it would take math, a lot of trial, more math and many many tweaks to get it right.
Oh, sure I agree 100% with you. I couldn’t possibly make up a system I’m describing that would accurately take into account game length variables and a million other variables for your individual score. However, there are people who get paid to design such systems. Those are the people needed here.
I imagine the system wouldn’t need to look at EVERY aspect of a player’s performance. TRS would pick out the vital objectives of each character and find a couple stats that accurately reflected that. Here is where secrecy to the details would be appropriate.
We wouldn’t want trappers to know that a cheap way to boost their rank would be to dome a monster, drop the dome after 2 seconds and redome the monster 45 seconds later over and over again. That’s counter productive to winning to boost your ranking.
Exactly what I was thinking about the vital objectives. I think the system in place with how @Miel described it is effective. It’s simple, sure. But it’s effective. By taking into account past performance, the algorithm predicts how you should do. If you exceed it (presumably) or match it, you get points based on that. If you don’t, you lose points. If you consistently do poorly compared to the algorithm’s assessment of your skill, you lose greater amounts of points, until you start to do better. Eventually you’ll settle into your skill bracket.
Unless you play PUGs. Then you are rolling the dice to find a good team. In PUG games, I was winning 31% as hunter. One of those games matched me up with a group of 3 that had mics and knew what they were doing (not pros or anything, just good silver tier players). We got on a roll and my overall win % is sitting at 51% for hunters after playing roughly 20 games with them.
Without a premade… This win or lose system really struggles to separate the good from the bad. Again, just in PUGs.
I stopped playing Hunter with randoms now =I I only play with a premade. I understand where you’re coming from though. It’s a nightmare for sure. All I can recommend is finding some people to add and play with. It changes the game on so many levels.