Loading...
  OR  Zero-K Name:    Password:   

Forum index  > News   >

Zero-K v1.9.6.5 - Handicap and Matchmaking Experiment

28 posts, 3228 views
Post comment
Filter:    Player:  
Page of 2 (28 records)
sort


2 years ago

Income multipliers can now be set for each team under Options to serve as a form of handicap. This is a commonly requested feature for coop games, but to take it a step further we're trying it out in a new matchmaking queue.

The '1v1' queue now generates even matches up to a certain skill gap, beyond which it creates unranked games with an income boost for the lower-ranked side. To stick to even games use the new '1v1 Narrow' queue, and for the best queue times use both at once. The '1v1' queue allows a slightly wider skill gap than the previous queue, while '1v1 Narrow' allows a slightly narrower gap.

Handicap and Matchmaking


  • Per-team income multipliers can now be set under Options -> Multipliers in battle lobbies.
  • Added a '1v1 Narrow' matchmaking queue that creates more evenly matched battles than the previous system.
  • The '1v1' matchmaking queue can create either handicap or non-handicap games. If a match would qualify for '1v1 Narrow' then a normal match is played, otherwise the match does not affect ratings and the lower skilled player receives a small income boost.

Behaviour


  • Added anti-structure overkill prevention for Lance, Lucifer, Ravager, Crab and Cyclops.
  • Reef and Shogun try to stay at max range against all targets.

Other Tweaks


  • Added NW vs. SE startboxes for Archsimkats Valley.
  • Calayo is back to just being NW vs. SE.
  • Tweaked metal spots on Battle for Planet XIV and DeltaSiege Island 8 Way.
  • Tweaked the sun and water parameters for a few maps.
  • Tweaked welcome text.
  • Better support for modding a few things related to the economy (fewer files to override).
+7 / -2


2 years ago
This update happened unexpectedly quickly, but with half of it out I decided to go the full distance. The numbers are fairly rough and there is a lot to tweak so be sure to give feedback. The rating difference for 1v1 Narrow is about 270 elo and corresponds to a 17.5% win rate. This number is easy to adjust. It is also easy to adjust the maximum rating difference for handicap games. The actual handicaps themselves are harder to adjust but there is always '1v1 Narrow' if their initial values prove to be too far wrong.
+0 / -0


2 years ago
Non-Steam users will need to redownload the launcher again since it no longer self-updates.
+0 / -0


2 years ago
Initial thoughts on handicap:


* Love being able to add a handicap to games. My brother has gotten interested in zero-k but it's not possible to play a fair match with him, so I've been balancing with a brutal AI on his team. It is way better to be able to just ratchet the resource multiplier up or down to make the game fair based on our skill level difference.

* The handicap system seems to assume a linearity in the resources required to balance different skill levels. I don't think this is likely. At high skill levels, the difference in 'resources' between players is likely to be much less. Consider a 2000 vs 1600 elo player -- the 2000 elo player will just expand and the 1600 player will do this a lot less. To offset, the 1600 player requires a lot of resources. Godde vs top 5 player is more about slight efficiencies that accumulate over time. Similarly in chess, small differences in material advantage mean much more at a high level than a low level of player. In other words, the required resources to balance should be sublinear with the ratings of the players.

* Philosophically, I think the handicap system kicks in too high at too low a rating difference. The goal should not be to make the game even (50/50), but to make the games more even. E.g., weaker player can still win 20-30% of the time.

* It would be nice to have the potential to opt out of receiving a handicap, as it makes it harder to improve when the timings and the mexes required to do different things is not constant between games. I don't mind someone getting a handicap against me, but I don't personally want a handicap against anyone.
+4 / -0
I like the handicaps. I suggested it 5 years ago.

How does the ressource factor depend on rating difference? Does it only apply to metal rate or also energy / build power?

Edit: In the last match, my opponent got 120% metal, energy and build rate.
+0 / -0
2 years ago
quote:
Non-Steam users will need to redownload the launcher again since it no longer self-updates.

Care to elaborate?
+2 / -0
USrankRiposteR if there is a handicap then it is set as follows:
if lowerWinChance > 0.20 then
    autoHandicapValue = 1.1
elseif lowerWinChance > 0.15 then
    autoHandicapValue = 1.15
elseif lowerWinChance > 0.1 then
    autoHandicapValue = 1.2
elseif lowerWinChance > 0.05 then
    autoHandicapValue = 1.25
else
    autoHandicapValue = 1.3
end

Some of the values aren't used, but they were there in case the maximum lower win chance for a non-handicap game needed tweaking above 17.5%. The absolute elo of both sides is available in this part of the code so it could be bought into the calculation. Doing so crossed my mind, but the numbers are so uncertain already that I thought it best to get some feedback now and refine the equation in future versions.

People don't seem to think in terms of win probabilities when thinking about whether they'd play a game so I often don't receive much information when I ask about them directly. The initial numbers are gut feelings based on years of occasional conversations/threads about how far up/down the ladder people like to fight, and then working backwards from there.

I would really like it if those that want to go deeper into this feedback would communicate in terms of functions. Like the one above. Functions are good for pinning down an approach - they don't have to be perfect first try.

quote:
It would be nice to have the potential to opt out of receiving a handicap, as it makes it harder to improve when the timings and the mexes required to do different things is not constant between games. I don't mind someone getting a handicap against me, but I don't personally want a handicap against anyone.

Something like this could be hacked into the system. I think it would be more complicated than a simple toggle though as I think there could also be a preference against playing ranked non-handicap uneven games from the position of the higher skill player as well. Adding options to the MM is not free as it can split it or reduce clarity.

I considered the issue of timings being thrown off by the nature of the handicap and it could be a big issue. Whether to apply the handicap to the higher ranked player, or even both players, was a bit of a toss up. I went with the current system because it makes for rounder numbers, seems to have the least impact on the importance of starting resources, and doesn't slow the game down. Perhaps the fairest system would be a somewhat symmetric system around 100% for both sides, so they both have the issues related to not having a normal economy. Something extra to consider is the relationship between handicaps and constructor speeds.

DErankBrackman I must have missed that, perhaps because it seemed to be about FFA. Given the points USrankRiposteR raises I don't see a multiplier as a function of ratings difference working.
+0 / -0
2 years ago
quote:
* It would be nice to have the potential to opt out of receiving a handicap, as it makes it harder to improve when the timings and the mexes required to do different things is not constant between games. I don't mind someone getting a handicap against me, but I don't personally want a handicap against anyone.


I agree with all of this basically, except I don't want others to gain a handicap against me either. Either way its giving my brain false signals that confuse the improvement process. If I run into Godde, or I run into the top 50 person, now suddenly the game is totally useless/meaningless in every way. At that point I will just spam fleas or something.

One could argue then that I should just go straight to narrow, which seems much harder to get a game. I actually don't see any reason for me to play MM with the handicap situation in place, it's offensive to my sensibilities. It's a nice setting for casual games, but I don't see how it has any place in a "Ranked Matchmaking" queue.
+5 / -0

2 years ago
If you want to crush total noobs without giving them the bonus you can use the 1v1 room.
+0 / -0


2 years ago
quote:
It's a nice setting for casual games, but I don't see how it has any place in a "Ranked Matchmaking" queue.


Says it all.

I honestly thought this update was a late April's fool joke.
+1 / -0
2 years ago
i didnt understand what is handicap and what it do?
+0 / -0
quote:
Tweaked the sun and water parameters for a few maps.

Azure Rampart seems to no longer have unit or map shadows. Other maps seem OK, but i haven't tested any of the other ones affected by this change.

UPD: looks like its new settings don't have a groundShadowDensity parameter. Setting it to anything non-zero reinstates shadows. Does the widget set it to zero if it's absent?
+0 / -0

2 years ago
I think a casual 1v1 queue with lower waiting times is great to have. It will probably be impossible to make this handicap balanced for serious games. Hopefully this doesn't come at too much of a cost for the "narrow" queue.

If we keep the handicap values stored in a way that's easy to associate with matches we could run some statistics later to refine it.
+1 / -0


2 years ago
This is the first I'm hearing of groundShadowDensity. I don't think the widget sets it and other maps with the widget work. Perhaps it is something particular to Azure Rampart. It can be set if required and you figure out how.

DF where should handicap be stored?
+0 / -0

2 years ago
Don't like the changes to 1v1 MM at all. I've ran into two out of the top 3 players on the ladder before and beaten them in the past and been rewarded for it; this system means that either I can be in 1v1 narrow queue and never run into them, or get a bonus against them that I didn't ask for and void the results when I can beat players ranked well above me when we're on an even playing field. It also would have slowed the dramatic recent rise of USrankRiposteR, who repeatedly beat players the system thought he had a single digit win chance against.
+4 / -0


2 years ago
quote:
This is the first I'm hearing of groundShadowDensity. I don't think the widget sets it and other maps with the widget work. Perhaps it is something particular to Azure Rampart. It can be set if required and you figure out how.

Done!
+1 / -0
2 years ago
quote:
If you want to crush total noobs without giving them the bonus you can use the 1v1 room.


So currently, there are 4 options to find a 1v1 game.

1v1 Narrow -

For someone at my elo, there are only 17 possible players that I can match with.

1v1 MM -

If I get one of the 17 players, I have an interesting game. Otherwise, I have this totally useless "handicap" game. Additionally, you are putting the onus on me to figure out if this game is handicapped or meaningful, as it COULD be either one depending on the ratings.

1v1 Casual-

This room does not use MM elo, its a completely different animal, as it uses casual/team/ffa elo.

1v1 Pro-

Probably the best option at this point, but there are plently of players that are not "total noobs:" who cannot play here, but would have a very good chance to beat me.
+1 / -0
I have watched the 1v1s of the past few days and tweaked the handicap numbers. Games at the top of the ladder were interesting in that sometimes they would have economic parity while the player with the bonus built up an army advantage, and sometimes the player with the bonus would run away economically. In both cases the bonus seemed far too large and has been halved.

Games at the lower end of of the ladder were quite chaotic, mostly depending on whether the skillset of the players lined up with using the bonus. The bonus here has also been slighted reduced. Most of the halving of the bonus happens between games where the lower player is rated from 1500 to 2000.

I also widened the definition of "1v1 Narrow" from an effective gap of 270 to 330. To put this in perspective, for games from this year:
  • 62.6% were matches within 270 elo.
  • 12.8% of matches outside 270 elo were won by lower rated player.
  • 73.2% were matches within 330 elo
  • 11.1% of matches outside 330 elo won by lower rated player.

The UI also auto-joins people who click "1v1" into "1v1 Narrow" as the former should be a superset of the latter.

Zero-K v1.9.6.7


  • "1v1 Narrow" range 17.5% win chance -> 13% win chance (matching range 270 -> 330).
  • Handicaps tweaked as per the lists below. They are now dependant on the rating of the lower rated player.
  • Clicking "1v1" also selects "1v1 Narrow".
  • Fixed Azure Rampart Shadows.
  • Gunship descriptions now refer to Locust rather than Banshee.

The previous win prediction to handicap function was as follows:
  • 15% to 17.5% -> 1.15 handicap.
  • 10% to 15% -> 1.2 handicap.
  • 5% to 10% -> 1.25 handicap.
  • 0% to 5% -> 1.3 handicap.

Now when the lower rated player has at most 1500 elo:
  • 10% to 13% -> 1.15 handicap.
  • 5% to 10% -> 1.2 handicap.
  • 0% to 5% -> 1.25 handicap.

The winrate thresholds are scaled down to 2/3 as the elo of the lower rated player rises from 1500 to 2000. This uses the 15% -> 20% bracket that technically exists at 1500:
  • 10% to 13% -> 1.1 handicap.
  • 6.6% to 10% -> 1.15 handicap.
  • 3.3% to 6.6% -> 1.2 handicap.

The thresholds are then shifted down by 2% as the lower rating rises from 2000 to 2500, ending up at:
  • 8% to 13% -> 1.1 handicap.
  • 4.6% to 8% -> 1.15 handicap.
  • 1.3% to 4.6% -> 1.2 handicap.
+1 / -0
My ideal for the handicap is as follows:
  • People that prefer it to a fairly onesided 10%< win chance game enjoy themselves and more people use the matchmaker.
  • People who want unmodified games get at least as many opponents due to MM activity, but with better balance.

Whether this can be brought about is unclear, and depends on where the win chance thresholds are set. The recent games and feedback indicated that the numbers were too high so I reduced them. It is hard to prejudge this so I'm planning to iterate for a bit. There are also a few different ratings reported by infra that could be introducing noise ( CHrankAdminDeinFreund may be looking into it).

quote:
Probably the best option at this point, but there are plently of players that are not "total noobs:" who cannot play here, but would have a very good chance to beat me.

I only became aware of this recently. I'll set the min rank to Giant (yellow) because there are a few giants on the top 50.

+3 / -0


2 years ago
I guess we could also add a "1v1 Wide" queue that has the same match range as "1v1" but doesn't apply a handicap. It would require a menu and infra release and we might be running out of queue space. Upvote if this is something you'd use over or in combination with the current options.
+5 / -0
Page of 2 (28 records)