SvR May Need a Matchmaking Adjustment

[shut juniper]

At the moment matchmaking is level-based (not win-rate based).

While its precise formula has never been shared - I am going to presume that it literally sorts all players into a list of level numbers - and tries to stack the numbers between the factions in a way that results in as even a total level as possible.

One would think that this is enough, but this does present some problems:

First and foremost this presumes that a level 200 player is about as strong as two level 100 players, and that a level 2,000 player is therefore about as strong as ten level 200 players.

This represents a worst case scenario, and I actually don't think that Pixile opted for a linear player weighting system. I also wouldn't be surprised if the current system works rather well when no higher level players are in the lobby.

However for higher level players, things probably aren't working out so well from a matchmaking standpoint.

The recent discussion on Medic level 1 helped me to realize that I truly cannot afford to play the class if I don't want my abysmal win rate to dip even further.

Regardless of 'why' I might be underperforming, even if one insists on explaining it away as a "skill issue", 'something' about matchmaking isn't adding up.

The base premise of this suggestion is that more nuance may be desirable, beyond player level. Player level encourages alt accounts and ironically makes events like double exp weekends a double-edged sword.

So let us look at alternatives. Spoiler - the next post covers pros and cons of simplistic substitution, and I've concluded that this would not do enough, or would trade one issue for another.

So let's look at simple swap-out alternatives:

a - Games Played (in SvR)

The downside of this is that a level 1,000 player who tries out SvR would be considered as a very low weight player for matchmaking purposes.

On the other hand, number of games played should theoretically be a better indicator of experience in the mode that Matchmaking is matchmaking for.

b - Win-rate (in SvR)

The downside of this is that it promotes 'winner and loser brackets' where matches are preordained before the start based on the highest win-rate players being pitched against the lowest win-rate players.

On the other hand, this could be highly responsive and pick up on alt accounts within a handful of matches.

c - Kills per Death, Heals, Captures or Twinkle Slays

A downside of this is that it focuses on a single dimension of SvR.

A further downside is that different class-based stats don't exist and so stats are diluted.

d - Player level

The current system. A downside being that player level does not necessarily equate to skill level. This could result in a situation where playing a lot is punished with ever harder matches.

On the other hand a player who picks up SvR after a thousand games of Royale would be given some weight in matchmaking.

… I'm not feeling it, and most of these seem to replace one problem with another.

I have concluded that none of these approaches seems correct. The following is a basis of a more complicated approach:

Matchmaking Weight = (((Base Character Weight) + (Highest of 'Player-Level or SvR Games'-Based Weight)) + (Win Rate /10 Modifier)) + 10% for each pre-set party member)

Where:

Base Character Weight = 100

Player-Level-or-SvR-Games-Based Weight:
-25 = 001-010
000 = 011-025
025 = 026-050
050 = 051-100
100 = 101-250
200 = 251-500
300 = 501-1,000
500 = 1,001-2,500
700 = 2,501-5,000
900 = 5,001-10,000
1200 = 10,001+

Taking the above for a spin:

Four relatively new players in a party who are level 50, have between 101 and 250 matches in SvR, and a 50% win rate would each amount to:

(((100 + 100) +5%(win rate) = 210) + 30%(pre-party)) = 273 each

For me:

(((100 + 900) +6.63%(win rate) = 1066.3) + 0%(party)) = 1066.3

When I look at this, I hold a couple of concerns from a fine-tuning standpoint.

Playing SvR matches will relatively quickly bring up the weight of a player, as compared to level. Player level would help to reduce the smurf effect of a royale player trying SvR.

Faster weight gain does make alt accounts less worthwhile - so that could be a positive.

I also like the existence of a base weight for characters regardless of level. This allows for easier fine-tuning.

I quite like the added weight of win-rate in the formula, as it could also go a way towards less improperly weighting an alt account while also providing a slight relief for lower win-rate high level or game players.

So, how are these adjustments looking?

(placeholder - just in case)

I am thinking that a 20% weighting for win-rate might work out better, overall. Not higher, I don't think.

[lavish sparrow]

is the current matchmaking system based on the summation of each player's level, or some function of each player's level?

[shut juniper]

That is unknown. I would be surprised if it is just summation.

[uncut vigil]

using my stats:

(100 + 1200) + 7.198% = 1,481
yeah, this seems good

[glad elm]

Win rate should maybe be rolling average (over like 20-50 games depending how long you want it to last) and that way you can weight it more
The player level and games played bit should probably be a bit more granular (maybe 1 bit between most existing, so -25, 0, 25, 50, 75, 100, 150, 200, 250, 300, 400, 500, 600, 700, 800, 900, 1050, 1200)
pre-set party weighing is a good idea for sure
Overall quite good, I do feel SvR should have more of an indepth matchmaking system

[tawny charm]

imo, even if this gets implemented-- It wouldn't really change the fact that older players have been driven away with how SvR 2.0 is "shoot at choke for 6 minutes to complete dailies" with various 'half-baked' bandage fixes that inflates the issues of the mode doesn't know how 'casual / serious' it wants to be.
..or the fact there isn't really enough players to fill or balance the lobby.

While I don't really care about win rate, I still don't really find any of the available solution on this to be any fun.
In fact, I'm convinced that it might have been better to balance players out based on

Objective-Related Involvement / Accuracy,

(But they NEED to address the core issue that, SvR at core is very problematic.)

[tawny charm]

Also to add a bit of tl;dr of what you mentioned above,
a) Playtime-based.. is no better.

• Some players are only active when bot fills the lobby, some only plays during peak hours; there is a masive skill gap.
• Punishes casual NA & EU players /w high play time.
• Players that have low hours but actually decent at the game, wouldn't get filtered out.
  ^also smurfs.
  b) Winrate issues.. -- too punishing
• You get punished for winning too much.
• This thing has a chance to amplify the alr boring "Spam skunk gas, bcg, etc on choke" META. Which makes the mode even more boring.
• imo squad stacking would be an issue. <- players may also literally dodge games that they won't win.
  ^ DeadbyDaylight alr proven this to be incredibly toxic; if Matchmaker only cares about W/L metric.
  c) K/D.. -- worthless in coop PvP
• as what I mentioned in point a), some players K/D are going to be inflated by bot kills.
• It encourages griefing / idling behaviour (players don't want to do anything else beyond 1v1 each other and shooting bots)
• Doesn't reflect what that player does. They could be the same player that would rather farm xp and kill instead of blocking the flag.
• Very biased vs support med players.
  d) Lv Based - awful
  We already saw this thing not worky.
  High Level =/= Good players. Also smurfs.

sorr for text wall. :3

[shut juniper]

A number of interesting or otherwise valid points have been raised - thank you. ^_^

Sampling a recent win-rate would allow for faster adjustment in SvRMMR.

An alternative approach here might be a function of streak sensitivity.

-40 = -25%
-30 = -20%
-20 = -15%
-10 = -10%
-05 = -05%
+00 = +00%
+05 = +05%
+10 = +10%
+20 =+15%
+30 = +20%
+40 = +25%

Where a streak interruption doesn't break the streak but rather reduces the streak value by 1 + 10% of the lower threshold of the band, rounded down.

Thus, a streak-reinforcing result increases the streak number by 1 while a streak-interrupting result reduces it by, for example, 5 in the extreme brackets represented above. This serves to preserve any compensatory matchmaking weightings in play.

This would replace win-rate in the equation. The disadvantage here being that it would not be transparent unless streak were exposed to players.

I never experienced SvR 1.0. However the observation about choke-shooting to achieve dailies seems to be just as much a challenges problem as a chokes problem (and certain map areas do have notorious chokes, to be fair).

In terms of there not being enough players to fill or balance a lobby - there 'are' solutions that I am in favour of. For one thing, if you know that there is a numerical deficit of X between teams, then there theoretically is nothing preventing the weaker team from being granted enough extra AIs to mitigate that issue.

Any improvement in AI would also mitigate dependency upon a 'critical mass of players' being achieved for the game to be sustainable. The existing AIs do a lot of heavy lifting already in this regards.

I had considered flag captures as an element - but it turned out too niche. It also doesn't help that most flag-raisers aren't human (because most humans prefer to shoot things than to raise flags).

Interesting observations on the different stats. I was surprised to hear the view that K/D is unimportant - but I understand where that is coming from.

It's a little late for me to consider the chokes issue - but I'll give it thought in the coiming day. Thank you all for your feedback thusfar!

[tawdry briar]

I played a decent amount of SvR 1.0 and whilst some people say it was more fun it still its differences:

-Having the whole map as opposed to a singular area meant that there more skirmishes or 1v1 or just random people walking around rather than the "big" teamfights in SvR 2.0
-Random loadouts of whatever you find for both your enemies and teammates (same as BR) which meant some people still had no gear and others were stacked
-4 man pre-queued squads could potentially dominate as they could pick out strays due to sheer numbers/co-ordination
-People often would cap a point when the gas was near it and sac themselves which sometimes made a team comeback
-(At least in my matches) few people met around the flags it was more like a beacon
-rip the countdown music

[shut juniper]

Has the countdown music changed? I would figure that the current 'final 100 reinforcements' music would have been the same (or an update).

What you describe does help to identify issues - and I am going to guess that SvR v 1.0 did not have respawns - which kind of puts it closer to Mystery Mode territory.

I think that there may be room for adjustments to SvR that could help to alleviate the chokes issue - but that would likely require a less static flag placement - perhaps even spawn - situation.

And knowing the skewing effect that pre-partying can have is precisely why I include it in the above suggestion. ^_^ Alt-accounts would remain an issue - but hopefully shortening the path to weight gain would make it less worth-while an exercise for them - and therefore less impactful on general play experience.

Oh, and thank you for taking the time to describe how SvR 1.0 worked.

[thick sluice]

I personally know that my mother who's level 1600 is no where near the skill level of players in the 100s it is a nice reflection of play time however does not reflect skill level. It can but not always hence why we still get one sided matches

[shut juniper]

This is a fair observation, and thank you for sharing.

So, in a situation where one puts 'less' weight on player level (or where one maybe flattens the added weight at an earlier point, or flattening the curve as a whole), what should take its place?

One 'could' look at kills per match.

In a sense one advantage that this could have is that if a player favours non-assault or other kill-oriented classes, then perhaps that will reflect in their overall kills per match and reduce their overall weighting.

If one goes down this route then looking at heals and flag captures per match should probably follow.

That does complicate the equation a little.

[tawny charm]

The music when only about 100 RP is available seems to be some variation of the old SvR music when the final flag is held down.

Here's an easier comparison.

Reimplementing v2.0 with some of the features that makes 1.0 exciting might have been the way, imo (Since there is truly nothing you could really do unlike in the past that wouldn't hurt your team harder if you are already losing RP majority).
e.g;

• Instead of doing Mole retake, simply end the game via countdown if all flags are held for 60s and enough players are present in the game.
• And genuinely reopen PTB again to iron out the poor loadout balance.

While yes, I understand well that the old system has similar issue of teams being unbalanced -- at least you are never forced to sit through some random game that is clearly unwinnable for 6 minutes. :3

I don't have the interest to open a new thread for this, since I'm aware that they are busy with S.A.W and the mode is pretty much dead to me.

[shut juniper]

I am sorry to hear that SvR 2.0 as it stands today is a mode that is not enjoyable to you.

Thank you again for providing insights into what SvR 1.0 was like.

SvR 2.0 is pretty much the only mode that I play today. I do occasionally royale but it stretches the definition of the word occasionally. So I effectively play a number of matches in SAR every two weeks, taking a week or so of break in-between.

I have a history of enjoying cycles for longer than most people might. I have not gotten tired of SvR 2.0 as a format.

SvR, in theory, is an excellent mode for beginners.

Unlike Royale, where you get tossed back to the main menu and then a pre-match lobby after each death, death in SvR 2.0 is an opportunity to learn.

Dying 10 times in SvR in a period of 7 minutes is more 'fun' than playing 10 matches over 30 minutes where 20 of them are spent between or prior to matches.

That is part of what makes SvR 2.0 valuable.

Having a selection of classes with pre-determined loadouts is also helpful for newer players, as well as players who don't want to concern themselves with chance from scrounging for resources.

Although the complete elimination of such can also leave out an important element for beginners learning the game.

It 'might' be interesting to experiment with a diamond configuration of two mid flags towards the top and bottom middle of maps - with Twinkle showing up in the center. This would encourage Mid-to-Mid switches, as well as interaction with Twinkle.

I'd need to give it some thought.

I also made the Mystery Mode comparison on purpose - because it seems to me that SvR 1.0 could fit better into that format.

I understand not being inclined to open another topic. Of course this is a fair ways outside the realm of matchmaking.

[shut juniper]

What a diamond flag configuration might look like on existing maps:

Bamboo: Mid flags: One above the stage and one either where the present one is or further down. Twinkle would find himself near the center circle structure.

Security: One flag ends up above the north wall of Security. The other ends up south of where Twinkle spawns. moves up below the secret garage.

Lab labs: Current mid point moves down. Another mid-point is introduced above the lab labs wall (may require one ore more breaks - or a north entrance in that wall for flow).

Sahara: - A new mid-point north of Twinkle's current location. Current mid-point moves down a little. Twinkle ends up inside the pyramid, a little south for more equal access.

Superite: A new point ends up around where Twinkle comes in, perhaps further north. Existing point moves down towards south of the mountain. Twinkle ends up near where the current mid point is.

Farms: Mid point up at or above Twinkle's current location. Current mid point moved down into structure below. Twinkle moved up into the building.

Tundra: Mid-point moved up. New mid-point south of Twinkle. Twinkle moves above the entrance to the palace.

A fourth flag would naturally require an adjustment to how points are deducted. I'd say that owning 3 flags should be less punishing but owning 4 should be even more punishing.