New season please

[fluid crater]

The issue is that there is no regular new seasons, no os decay, no raising of uncertainty when on winstreaks etc...
So uncertainty system is fundamentally flawed.

[hybrid jackal]

play 300 games and retire thats how it is now

[dim tendon]

I agree something needs changing but going overboard would also be harmful

[hybrid jackal]

not doing anything is more harmful just raise it thats all there is to it if its tooo much you lower it if its too little you raise it

[magic crystal]

Valid point perfi but other threads or maybe this one discussed separate skill ratings for diff size teams which would solve that issue imo

[buoyant dew]

I agree!!!!

[hybrid jackal]

well now we have 0 impact

[buoyant dew]

With a silver lining still being that we also have close to 50% WR’s so at least ‘as designed’ the system is successfully balancing lobbies

[dim tendon]

And that's also bad, I agree

[hybrid jackal]

0.02 gains and loses are a joke it means you get forced to stay the OS you are doesnt matter if you improved or not

[dim tendon]

I think everyone agrees with that and we don't need to restate the point

[hybrid jackal]

so there is no point in this whole discussion and just raise it asap

[buoyant dew]

Part of the point of the discussion has been to try to generate a consensus which it feels like has happened over the past week or so, particularly based on all the data analysis done by Cow and Doushibag

[hybrid jackal]

eh its longer then a week this is an age old discussion since the start where the majority already thought it would be better to raise the min cap but it hasn't been done

[buoyant dew]

I think people absolutely have the right to ask for analysis or some sort of evidence that the proposed change won’t break anything and that it is the ‘right’ change to be making, and doing the work to provide that evidence took time and a lot of effort by those two

[hybrid jackal]

what to break what is already broken

[buoyant dew]

Since that’s quite poetic, maybe I’ll offer this:

What’s another few weeks to someone who has already waited a lifetime?

[hybrid jackal]

it means im uninstalling maybe no one will care but ive reach the end of my patience for this

[woven moat]

Do you want a refund? Come on, let people do the work.

If it's so trivial for you to raise, do it yourself, game is open source.

How freaking hard is it to understand that there isn't many people currently familiar with this area, it's all done in free time, and people are already working on it.

[dim tendon]

Since you've consistently been adding nothing but whine to this discussion despite warnings I'm giving you a targeted #1047081727291559986 mute

[knotty axle]

Being 48 OS and having to play the same strategy every game or be a detriment to my team was not fun at all. I stopped playing because I would have to int hundreds of games to get my rating to a point where I could play different things without griefing my team.

-remindme 14d check for new season

[lyric shore]

so multiply the entire OS value by 100 or 1000
0.02 becomes +2 or +20

There is a perception issue here. The OS gets rounded and displayed as an integer. Yet changes by fractions.
In addition; the algorithm assumes that fits players to a distribution.
The model will still deliver players to fit into that distrubtion. We are skewed in how we see that number; 30+ OS is actually 93rd precentile. Top 7% of players.
Anyone 25+ is a strong solid above average player.

The issue about being "stuck" on low OS due to uncertainty is a seperate issue; I can see why people want something to change about that.

[pale haven]

It doesn't matter if I win. Or lose. That's the thing. I have to win 10 or 15 times more often to achieve the same result. My uncertainty or whatever it is is 0.02-0.04. That means I have win 10x more just experience any significant rise. Nothing feels better than progress. Look the math is pretty impressive how all this stuff is calculated but surely something like an it should be proportial to an inverse quotient of current standing. Players below 10 should rank up quickly but have to achieve milestones. Before being able to surpass that ceiling. Play twenty games, don't stall on e/m for 5 min. Use the resource sliders. Don't build a converter before storage etc Just random basic learning mechanics. And so forth. Scale the achievements. 10-15. 15-20 etc. But progress should still happen that develops the players skill and challenges him. This is of course happening in conjunction with the uncertainty mechanics etc. But there is movement. As the player OS increases inversely the uncertainty decreases but the achievements or whatevers never let that stall out. It should be hard to get rank above 30 or whatever. But you shouldn't have a system where an OS 45 Player has an uncertainty or 0.4 and an OS 13 Player with an uncertainty of 0.02-0.04. I don't care if it keeps me at thirteen. But I want my wins to count. Also I'm totally not advocating a reset in OS. No ways. Absolutely not. And also you can't say the it was my win or lose rate. The team loses or wins. I could be crushing it and lose the game. And visa versa.

[bold zodiac]

You play for OS only? OS was only used for balance and only problem i see is if you got worse and correct your current state will take longer. If you are undervalued you will enjoy more wins.

[pale haven]

@indigo pine I held you on bismuth. Well. I got asked by bdbr to play with stardom and their team while they trained on industrial. I kicked some 30 OS ass. Ive kicked alot of high rank ass before. And get the " woah you played really well for a 13"...which is funny and irritating at the same time. And those wins didnt mean shit to me. Just a small fraction. It's the same for me when I have to play in low OS lobbies. It does my nut in with inexperienced players. And I invite anyone to go watch me play. God I hope you don't get a bad one. Cos let's be honest there some awful moments. But there's some good shit in there. Not something you'll see in an OS 13 Player. And that's the point. How can you say to a guy who's got I dunno 700 matches under his belt, that now he has to play with dudes who started today. But look. I've made my point. I love all you players. And maybe I like being the hardest OS13 in Bar 😂 Maybe I should change my name to TheBloody13..... No I don't . I really really don't. I would also be happy with 0.1. At least that gives me a fighting chance.

[cunning bear]

You can see a full explanation of Openskill (our rating algorithm) on the site: https://www.beyondallreason.info/guide/rating-and-lobby-balance

The site guide answers many questions such as "Why can't new players start with a lower rating" and "Why can't we rate based on in-game scores?"

[dim tendon]

Also final warning

[pale haven]

Smurf alert 😂

[pale haven]

Lol goaf. Sometimes I hate you 😂

I feel I've contributed actually to the conversation and not just whinged. I feel postive and reassured right now and that's a good thing. And if I've if said too much I apologise.

[unkempt raven]

Yeah, I don't even disagree with you that change is needed. Just making sure you understand the implications in your specific case, which you clearly do.

[haughty light]

No it would break becuase of renames having no the same name as a current or prior name existing
I have the same issue

[turbid sparrow]

I think the point is, if he improves his gameplay to improve his winrate, his OS will only very slowly adapt up from 13. A minimum uncertainty would accelerate this process.

[unkempt raven]

Yep, we got there in the end.

[buoyant dew]

Hey, simp alert but I’m just overall extremely impressed by the developers of this game so when they say they’re working on it, I believe them

[spare parrot]

i think everybody believes that people are working on it, just that people think they are working on the wrong thing for the timescale involved

doing a complex-ass analysis of data and reading 50 scientific math papers to find the perfect parameters is good long-term work, but before this is ready you could have 3 years' worth of seasons via simple uncertainty resets in the meantime

[river herald]

Well with all due respect, the uncertainty resets make the game balance universally crap for everyone for months, whereas the uncertainty floor will probably help a small number of players a marginal amount and give a temporary psychological boost for another minority

And I say that as someone who thinks it should be done 👍

[spare parrot]

psychological boost is half the point of having a ladder though. number must go up

[fickle gazelle]

Psychological trickery is the main reason ratings exist. Especially when you add in tiers like plat/diamond etc. Having actual balanced games is secondary, I'm sure everyone is well aware that even the best rating systems out there produce crap balance in team games about half the time

[river herald]

No one respects the chevrons 😜

[spare parrot]

the uncertainty floor will probably help a small number of players a marginal amount and give a temporary psychological boost for another minority

so what's the downside? remember you don't need to tie it to a season and stick to it

[river herald]

I hate to say it, but the urge to add play time and have some social signifier for the pve crowd leads to the devaluation of the other system of social status

All the PvP players know OS is the real metric and the chevrons are for participation awards and identifying noobs

[spare parrot]

limited (and possibly outright fake if it just identifies you as a lob in the eyes of PvPers) social status via mere engagement is good though. people playing the game is value in itself to the game

back onto OS though, remember that the more time you waste finding the perfect formula instead of just yoloing it, the fewer opportunities you will have of seeing the actual practical effects various parameters have. idk how seriously bar treats the "still alpha/beta" phase though

[icy wing]

actually chevrons can be disabled, so you don't even have to look at it

[river herald]

I think most of us are agreed that the main barrier at the moment with OS is technical wrt production changes

if that's not the case then 🤷‍♂️. I can make up recommendations (pretty much 3.0 floor id be happy with at this stage, and any questions anyone has to get past any doubts about what it'd do I think can now be answered quickly)

[spare parrot]

who would be unhappy with a floor?

[river herald]

It would shift leaderboard relative positions and relative match ratings more or less overnight

In my experience people can be suprisingly petty about that kinda stuff

[fickle gazelle]

Of course they don't, playtime is a meaningless metric, some people become amazing after 20 hours and some people never learn the basics after 1000

e; oh it seems that was your point 😛

[fickle gazelle]

Fear of a negative response from changing that doesn't make sense. We're getting a negative response now from tons of people, because their OS isn't moving. Avoiding fixing the source of that, because it might theoretically make like 10 people upset (most of whom have stopped playing because their OS isn't moving) does not make sense.

Anyone who loses a couple spots in the leaderboard will be motivated to fight to get their position back, once their OS can move again.

[icy wing]

how much more OS gain you guys are aiming to have compared to now? 50% more?

[charred lava]

My uncertainty is between 2 and 3 rn, and I move between .1 and .2 OS per 8v8 generally.

[icy wing]

i think keep it like right now but have the floor be around 0.1 gain

[river herald]

I don't think it's realistic we can place the explicit floor on the gain...

[tired hemlock]

I'm still of the opinion to activate the tau parameter, must less work than adding a manual change like a floor, or uncertainty decay with time.

[hexed night]

you don't have to reset it all the way to the max, just increase it, at least for the people with the lowest values

[river herald]

Tau slows down your decay, doesn't help with the people already there and doesn't solve the problem over the long term

[hexed night]

short term problem solving is good for the short term

[tired hemlock]

Then you also get weird things like, should a 'pro' really be able to earn 1OS stomping noobs every 10 games? Without actual matchmaking used the rating is ripe for giving meaningless results.

[river herald]

Well, tau doesn't imo solve it in the short term either

[tired hemlock]

tau solves it in the long term, but not short term, yeah, unless you bump things a bit now too

[icy wing]

actually the meta people are looking for is loosing ts fast with the OS rework, so people can chill at 35 ts while being real 50

[tired hemlock]

Either way, hard capping the rating change is a very unsatisfying solution.

[river herald]

Yes, but that problem comes up if you try to intercept and explicitly handle the rating as well. If you set a floor on the movement, then people can begin to boost and arbitrage by targeting games where OS movement should be less than the floor internal to the model

The other issue from a technical perspective is that rating movement is kinda internal to the OS library, whereas mu and sigma are external values expected to be fed into the model as a parameter

If I'm right, the actual code change is deceptively easy to implement an explicit floor: just a max(floor, sigma) function call

Simulations suggest the explicit uncertainty floor has very little effect on the overall distribution of ratings as well

[hexed night]

is there something stopping from doing a change like

if uncertainty < 2
uncertainty = min (uncertainty +1, 2)

for now?

[river herald]

I think the holdup now is how to make changes to the prod environment rather than the theory

[tired hemlock]

Also something to keep in mind, you already have different ratings for 1v1, team, FFA etc. So the parameters can be different for each. Maybe for FFA beta should be higher, for teams tau higher etc.

[fickle gazelle]

idk, people said that months ago but so far as I can tell all the people who can make that change are still running simulations for a whole new season, rather than figuring out how to deploy OS changes

I think the people here who are hesistant to make changes are not really aware of how bad the situation is right now. Tons of people have stopped playing completely, particularly high OS players. And the ones who still play are openly trolling more and more frequently, specifically saying that what they do doesn't matter, cause their os won't change anyway.

The hesitance to make change comes from a very reasonable worry that things will get messed up, or that the system will have flaws. But the severity of the situation calls for quick action - there's no point trying to deal with the stage 1 cancer while the patient is bleeding out from a blown up leg. First we fix the major problem, then we can figure out how to address smaller and longer-term problems.

[river herald]

For FFA its barely an issue atm because the vast majority of the community is no where near playing it enough to get near reasonable uncertainty floors to be honest (and those who might are probably accurately calibrated), for 1v1 the main problem is making sure you don't set it too high to make the variance too bad given the comparatively larger movements and we don't have to worry that much about balance or explicit randomness of outcomes from the mixture of teams, which primarily leaves teams balance and tradeoffs as the main issue.

Which all leaves me at the: choose some conservative amount somewhere probably in the 2.5 - 3.5 uncertainty window

[river herald]

and we can check what the basic movements are for those at the uncertainty floor if we want to explicitly target some number like average movement of 0.1 in 8v8s for some arbitrary reason

[elder parcel]

lol i played 50 games with 68% win rate and only went up 0.8

[charred stag]

I would point out that in the summer of '22 Teifion literally split OS into large team and small team OS, without any visible discussion that I saw(albeit BAR was smaller at the time), or running any simulations. He just did it as an experiment, and a month or two later switched it back to just team OS. Don't let perfect be the enemy of the good. Save your work, revert if something goes catastrophically wrong, and give it a try. Most players are not even going to notice anything different.

[crude eagle]

You can't really set a floor on the actual gain/loss itself, that has to be dynamic based on the matchup, but keeping uncertainty above a certain value with a floor on uncertainty means the values are generally going to be above a certain gain/loss, but not strictly.

running some simulations now with changing the Tau value and it doesn't seem to do much at all so far. If it has no real positive impact on my results, then I don't think it's going to alleviate anyone's current issues even if it was changed right now.

[fickle gazelle]

tau being uncertainty? How could that not change the results, wouldn't it immeditely change how much OS change people have per match?

[tired hemlock]

Tau is different to the uncertainty.

Tau is a 'dynamic factor', a tiny increase to your uncertainty after each match to slow it's decay, and effectively keep it from decreasing to 0. I think the default recomended is sigma_0/100, i.e. 0.083, a tiny tiny amount.

[woven moat]

@patent hamlet can we have please a public list of steps and blockers, decisions that need to be made for changing anything here we can point people at that are interested in this change? Likely the more granular the better. You can add such list in whatever format. Even just a new temporary GitHub wiki page. All state is lost in this discussion thread here very quickly, not clear what is blocker and what isn't and doubts about both understanding of severity of the issue and if progress is being made.

[pale haven]

Simplifying dramatically here. Turning the volume up and the tuning it down to calibrate is much easier than turning the volume low and then wondering how high the right listening volume is. Occams razor suggests 2 or 3 would be the easiest way to determine a viable working model in the shortest time frame.

[turbid sparrow]

Also players who didn't realise that playing when drunk would really crater their rating, and now are only slowly recovering by sober play!

[turbid sparrow]

What is tau, time-weighting?

[buoyant dew]

Is that what Tau actually is though? From my read of the python implementation, it is just a ‘de-buff’ to your uncertainty and changes in uncertainty.

teams[team_index][player_index].sigma = math.sqrt(player.sigma * player.sigma + tau_squared

So if your uncertainty was 3.0 and a tau value of 1.0 was used as a debuff, you would be treated as having an uncertainty of ~3.4

I don’t think Tau of 0.083 does anything meaningful

[tired hemlock]

I don't know anything about the python implementations sorry. Is that a temp variable or going to be carried into the update post match?

[buoyant dew]

I referenced the python implementation only as I found it more easily interpreted than the whitepaper, which is way more math oriented and put me to sleep trying to read through lol

Tau is a system-wide variable, once set would apply to all players for all games. We currently don’t use Tau (equivalent to using a Tau of 0), and some of the smarter-minded statisticians are in the process of investigating what impact increasing Tau would have on lobby balance

https://discord.com/channels/549281623154229250/1190016394666983566

[buoyant dew]

My simple take is that it’s a de-buff both to your uncertainty and your changes in uncertainty

[river herald]

However it's implemented, the simulations show it doesn't generally stop decay, just slows it...

The most extreme player in the community still ended up at relatively low uncertainty after 6 months play

And otherwise higher values of tau ended up having somewhat weird effects

And it seems all strategies I think requires some kind of prod mutation of current player values manually, can't just change the code

And the caveat of tau is that it marginally increases the amount of time the other players remain at high uncertainty

[glass robin]

Am I correct in assuming the main blocker to this feature is deployment? The code change itself is simple but you lost a dev responsible for deploying?

I also assume the change you want is setting min uncertainty not changing tau

[river herald]

Intellectually, I think the code change should be very simple

However I have no experience with elixir so I would have to figure out how to set up the Dev environment to actually test it. And also obviously, if I had a dollar for every time someone said "oh it should be an easy change!"

And I'll be honest, my motivation for doing elixir is very low (in the sense of setting up yet another Dev setup for a language/area I'm not really very interested in)

[glass robin]

Elixir is a programming language? And nobody knows this language currently?

[river herald]

It's not so much the language, I mean I can eyeball it and tell you what the change should be. It's more about deployment...

Making the distinction between writing code and actual eco system surrounding a language in prod

[haughty light]

It’s relatively new and isn’t as widespread as you think

It’s growing in useage though

[river herald]

Like even if I say "just wrap this call in max(floor, sigma)" how do I test result that I'm actually right?

And then the mechanics of rolling it out to prod, I believe the BAR team is relearning that with teifons departure

[alpine slate]

2 observations to estimate a trend requires that you know the rough form of the function that describes it

Or else any finite number of points can perfectly fit an infinite number of polynomials

[buoyant dew]

your mom can perfectly fit an infinite number of polynomials

[pale haven]

Lol. Classic. Some good dry humour to break up the math class.

[fickle bone]

Going back to the issue of inflation, I think it is important to distinguish the model's uncertainty resulting in awarding a higher skill improperly, vs. the raising of the uncertainty floor, to give players more OS mobility, resulting in higher ranks because it solved the original mobility problem. Those two things are not the same to me and I feel like this thread has multiple interpretations of inflation and we are all using it differently. For example, player X plays bar as their first RTS, goes on massive loss streaks for months. Learns the game and is playing at a level much higher than what his/her rank represents. The uncertainty floor is then raised. Player X gets more mobility, resulting in higher rewards for wins over less games and their rank increases. In that case I would not say their rank is inflated the model is just more accurately representing their skill

[ancient bear]

Imo resetting os will be so bad, maybe uncertainty to 2 at most

[warm ivy]

Resetting can be good for very low rank such as < 1.
For others players is enough acting on uncertainty.

Maybe uncertainty can be raised to high level for everyone.

[crude eagle]

https://github.com/beyond-all-reason/Beyond-All-Reason/issues/2611

[spare parrot]

this isn't the actual summary

it was made for the purposes of an internal discussion

to bait out the real summary which surely exists somewhere

[alpine depot]

can they really just run a sql command and fix it already

[low coral]

imo do something like move 1/5 of the way from current to starting sigma, to spread it out more evenly

[peak skiff]

Reminder for @buoyant dew

Reminder from YAGPDB

pray for new season

[amber lantern]

Keep praying

[buoyant dew]

🙏📿 🤲

[crude eagle]

"some gigachad made a proper doc. good!" Where is this doc you speak of? Internal for developers only?

[spare parrot]

yes

[alpine depot]

why though

[crude eagle]

So they can have an internal discussion without being distracted by others probably.

I'm sure if they feel they need any additional external input they'll ask for it.

[buoyant dew]

Just like how I hide my feet pic folder

[primal adder]

New season when?

[patent hamlet]

Being discussed in earnest. We'll update you all within the next couple of weeks

[pale haven]

Yeah I reinstalled... There are so many lostdeadman streams a brother can watch. 😂 💛.

[buoyant dew]

New season when?

[past dust]

This is usually because you resigned but your team won the match.

[shrewd coyote]

A new season (in terms of resetting ratings, or complete uncertainty reset) is pointless, it does not fix the underlying issue, and only delays it. The problem is uncertainty needs to change based on new info.

1. Skill is not static, and having a winrate consistently above or below 50% should increase your uncertainty.
2. Time should also increase uncertainty by a small amount (this has no impact on people playing regularly, but will add up for those taking break, as it should due to skill decay)... a small daily sigma increase toward default is actually suggested by the openskil creator btw.
3. Inflation is unavoidable in any online game, even moreso for a f2p one. If you want to have seasons, this is where you do your "reset", but at no point do you completely reset OS or uncertainty... you simply squish all ratings towards the point that would eliminate your calculated "inflation" based on all the users who signed up, played a few games and then never logged in again. This will mean all OS's will be shifted slightly lower, but rank order is maintained.
4. If your season refresh comes with large balance/gameplay changes, then you can also add uncertainty. (I'm personally against the typically discussed "chunk" of sigma, and think it's better to just decrease uncertainty delta for the first X weeks of a new patch.. but this stuff is all non standard customisation anyway)

The biggest issue with stock implementations of OS, is that uncertainty trends to 0, which does not aling with reality. The entire point of these systems is to get people's WINRATE to trend toward 50% asap. What we have now, is a situation where a player messing around early in their career, who then decides to get "serious", will just end up with a 70+% winrate ruining the balance of games, because the system is sure that based on his first 200 games he's trash, and is never willing to reassess that.

[buoyant dew]

yeah you'd be right usually - but our team did in fact lose

the reason it's -0.0 is because I have very low uncertainty (~0.7) and because the balance deviation was a bit higher than normal (don't have exact figures but probably ~4%)

[past dust]

0.7 uncertainty is quite an achievement honestly. Guess you can have 0.0 with that huh.

[crude eagle]

those are rounded numbers too, it's not actually zero

[past dust]

Yeah it's probably 0.004 or something. In any case a bit frustrating how little it changes.

[crude eagle]

yeah, hopefully they'll fix it soon
I did what I could to help with their decision in the short term at least

[past dust]

Me too 🙂. We'll get something in the coming weeks I'm sure.

[river haven]

I dont know if you will ever see anything here going fast or in time. Compared to the other lately (at least initially successful) indie company of BattleBit, who always reacted in time, here the things are discussed in vicious cycles over years and years again, threads getting derailed by the same people with the argument that is was always like this and why would we change it, good solutions of the discussions are mostly not noted anywhere and a few years/months later the same stuff is discussed again.
And even if a solution is finally found is has to be definetely the perfect one, so lets evaluate it for a few months before implementing.

I did not play for a few months and after coming back, I was surprised that we are mostly still stuck in the same issues and that they are STILL discussed with the same arguments lol (Does not mean I do not appreciate the effort of the developers, but I highly suggest implementing some management basics like priorities and deadlines for certain things)

[glass robin]

Is BattleBit opensource and run by volunteers? It seems to have it's own dev studio.

If people truly want to change things here they'll have to get a dev role and starting looking at the codebase.

I highly suggest implementing some management basics like priorities and deadlines for certain things)
There's no way someone working for free wants to be managed like a real company.

[hexed night]

This isn't an issue that would need corporate management to fix

[abstract cosmos]

Nothing gets changed because forever the same people keep blocking for reasons that it’s always been that way is a straw man argument

[woven moat]

That's not true at all. Point me please at those people that say that nothing should change.

Note that random players don't count as they don't control anything. I don't know about a single person who's option actually matters to say that current status is good state.

[abstract cosmos]

Direct your question to 1Daniel1 not to me

[woven moat]

Oops, sorry, I failed to read your message correctly

[spare parrot]

yea "(...) is a straw argument"

[dim tendon]

I think we're all understandably a bit on edge about this whole subject

[river haven]

The story is actually kinda similar to BAR, apparently most of them still had their "real" jobs and saw it (untill release) more as some kind of hobby:
https://www.gamedeveloper.com/game-platforms/behind-i-battlebit-remastered-s-i-seven-year-journey-to-become-a-smash-hit

And I am quite sure that here is a similar potential... (and I think even more ppl actively involved than in BattleBit)

[river haven]

that s not a straw man argument, funnily you are actually one of the persons I meant with this argumentum ad antiquitatem. Though in your case it is not always argumentum ad antiquitatem, not seldom you just dont understand the messages of the other people - as I have just seen in another topic, where the other guy got even triggered so hard by your responses that he began to personally insult you lol

[spare parrot]

if you meant him specifically then you're wrong because he is a literal nobody

[glass robin]

No I think that's different still than BAR. It's how companies usually start with a couple founders. It's still an actual team. Like a startup.

“In the beginning, it was more of a contract/freelance job since I specialized in what they needed at that time, but then the job led to long-term friendships,”

So they were actually paying people to do work. No different form a normal company. They weren't volunteers. The founders had the intention from the start to make a for-profit company.

If I'm paying you to work for me, I can give you a deadline and end your contract if you don't meet my expectations. With volunteers, it's really up to them when they want to do the work.

[abstract cosmos]

I am a person will you grant me that

[abstract cosmos]

So it’s not a group but only me? Please show proof that this is actually the argument I make as I’ve never said that

[spare parrot]

sure, you're cool. but for the purpose of defusing the argument you're still a white nick

[river haven]

"they were all working in their own fields, and making this game was a pure hobby for them." I think some of the modding guys, they hired later when it became mor ambitious - the article doesnt go too much into detail.

Also with volunteers you can do similar stuff, make a meeting, discuss together what the priorities are and until when you want to get it done, talk about who can solve it best. That is actually how it is done in many voluntary organisations, though the issue might be here that the people know each other mostly from the internet, thus the bonds are maybe not that strong as in the typical voluntary organisation.

[river haven]

"one of the persons" -> not only you

Ohh I remember for instance the energy converters / afus discussions quite well, where you were arguing that they are at a sweet spot and nothing has to be changed and it was not even clear yet about what exact numbers we were talking

[buoyant dew]

@river haven without knowing your intentions, which I assume are good, could you please keep this on topic no need to devolve into past history or attempt to create a slinging match

[elder parcel]

@buoyant dew @buoyant dew

[twilit oriole]

i feel that i am in this rating limbo where i win or loose only 0.05 every match, almost impossible to raise my rating, is it only me? why ir works that way?

[indigo pine]

it works the same way for everyone who's hit a low uncertainty. aka system has determined you are around 30-31os and now ur stuck here for until anything is changed gg

[river haven]

Can someone explain me the uncertainty algorithm? On the homepage it says it only depends on the number of games played.

Should not the uncertainty also be related to the winrate, I mean we want a person at his true elo, where he kinda wins 50% of his games or not?

@buoyant dew for instance has a winrate of 58%, thus he is underrated and it should be still possible for him to gain elo until he reaches the 50%
@twilit oriole has a winrate of 67% (extrapolated from the list on the screenshot, cant see total), so there should also be more movement possible.

[elder parcel]

read the page
uncertaintty just goes down more games you played. it's not to force 50% win rate over whole career but 50% win rate at the stable point

[pale haven]

Let's not forget I have even worse uncertainty than that and I've been stuck on 13 since forever and will be forever unless something changes. It's not fair to assume that the devs don't want to make change. It's their passion project and I'm sure making huge changes can be quiet distressing. But when things like this fester it becomes unpleasant for everyone. I'm sure when the changes are implemented. And they work. Which I'm sure they will the general overall postive sentiment for the change and the relief of having taken action on a draining issue will benefit every bodys moral, on both sides of the fence.

[indigo pine]

realistically though i dont see many people moving from their currently assessed skill level, assuming they're low uncertainty atm.
only a few will see massive gains if uncert were to get reset.
more inflation may happen instead, so for example I would go from 43 to 53, but everyone at 50 rn would just be 60.
and current players at 10os, would possibly move 10 > 20, but in actuality, relative skill levels would be the same.
i dont expect massive upsets in terms of OS.
The only exception is high OS people dropping OS on purpose, which i would probably do myself if im being honest.

[glass robin]

Isnt OS zero sum? If you gain OS someone is losing a similar amount.

[charred lava]

One example: A player can resign before 90% of the match on the winning team, meaning one player would lose OS and their mirror would not gain.

[glass robin]

That actually means OS would have deflationary pressure not inflationary.

[indigo pine]

after last reset everyone did infact increase by around 10os

[ancient bear]

We will see huge unbalance imo, for months , if we OS / uncertainty reset

If you are so low calibrated you should be able to win tons of games and be raising OS in no time

[floral lynx]

I don't think so, if you have high uncertainity and play against someone with low uncertainity you will move a lot more than your enemy

on second thought I'm actually not sure, maybe you're right. that would mean that people who are stuck should ideally play against people with high uncertainity

[glass robin]

@lone junco I know you've looked at the OpenSkill library. Can you confirm if skill changes is zero sum after a match is played?

[lone junco]

Individual matches aren't zero-sum, because players with larger uncertainty gain or lose more points. It's only zero-sum on average.

(Actually it probably isn't quite zero-sum even on average, because overrating of new players means they lose more games than they win, so they presumably lose more points than their opponents gain.)

[hexed night]

if that was to happen, which idk if its true, it would alleviate the downside of new players starting at 17.

if a current 10os player becamse a 20os (with enough games put in)

[pale haven]

That's like Paris Hiltons quote " Just don't be poor" 😂. I can't actually pull myself through this again. Cos it is requires empathy from you to understand that I have played hundreds and hundreds and hundreds and hundreds of matches. I'm tired of it not meaning anything. It would suck if I said okay you go play 250 matches against 10os players. You'll be up in no time. It also comes down to credibility. How can a player who's OS 30 still not know some of the most commonly played maps or is still like oh wow I didn't know that could do that. Like I saw tonight. I'm just like dude... How did you get to 30 and not know that. ( not referring to you obviously 😂) but I can't look at his OS as something meaningful either. I think we should stop worrying about the top of the pyramid and think about the vast player base. I don't know. I'm tired of trying convince people what it's like, players who get to enjoy the one aspect of bar many of us don't. Progress. 🤷

And you play awesome bro. So props

[hexed night]

talking about effects of a full reset is kinda pointless, thats not gonna happen and nobodys asking for that

[icy wing]

Listen to this man, he got it right

[elder parcel]

it's just QoL
to have your os adjust faster to changes in your actual play. like meming a bunch, or learning new role/factory

[tired hemlock]

No. That's one difference with Elo.

[turbid sparrow]

A resounding no. Neither is TrueSkill for that matter. 🙂

[heady ginkgo]

huh really? How does the population mean OS stay constant then?

Agh I should just go read the paper

[dim tendon]

It doesn't 😛

[heady ginkgo]

....That sounds fake but ok xd

[dim tendon]

A sizable portion of ACowAdonis's concerns about it come from that

[heady ginkgo]

It doesn't really make sense to me that you can have a relative ranking system without maintaining a fixed reference

[dim tendon]

Anyway, I think this thread has outlived its use, #announcements message is official confirmation that yes, a new season is coming

-remindme 24h unless anyone has any threads they'd like to wrap up here, I'll close this one

[peak skiff]

Set a reminder in 1 day from now (<t:1708535513:f>)
View reminders with the reminders command

[heady ginkgo]

I know how BAR uses OS currently is kind of busted because of the whole resigning before 90% and not gaining OS thing, and then balancing uses (OS - uncertainty) and uncertainty changes don't need to be the same

[pale haven]

I just want to say thank you.

Thanks for listening. And thanks for doing something. Means a lot. ❤️

[fluid crater]

Very op change indeed!!

[buoyant dew]

@woven moat very op ❤️❤️❤️

[haughty light]

@dim tendon I think it’s best to not close this thread yet

[alpine depot]

well yeah we knew new season is coming, but noone still knows when and how, and the announcement post explicitly says it's not a new season

[dim tendon]

Fine, then

[crude eagle]

Wasn't the biggest or primary reason for him asking for a new season to deal with the excessively low uncertainty he and others had? This is going to boost uncertainty up for people below the threshold, so it's addressing, at least in the short term, the primary desire for a new season and alleviates the immediate issue.

[dim tendon]

Probably good to still have a thread open for the not directly immediate aspect 😛

[crude eagle]

If people have other reasons for wanting a new season, perhaps a new thread should be made for that instead of continuing this one.

[pale haven]

Yeah i was only advocating for the uncertainty increase. Im more than happy to wait for a new season to be polished.

[raven jetty]

Exactly!!!! 100% agree!!! I was trying to say this earlier

I think resets are pointless, I am a fan of having some kinda of decay/compression method to achieve average OS ratings closer to what should be average

I also do not see the value in having the leaderboard rating differ from the match balance rating, since honestly if someone has 500 games balanced at 53os in lobby balance, and the other player has 1000 games at 52os in lobby balance, I bet you the player with 53os is actually better despite the 3 sigma vs 1 sigma delta

I could possibly explain this on discord and answer questions, also I have my own questions ect. But I do understand numbers and systems really well. If someone can walk me through how to setup the simulation that is already being used, i might even be able to run my own simulations to better help and tune things.

[peak skiff]

Reminder for @dim tendon

Reminder from YAGPDB

unless anyone has any threads they'd like to wrap up here, I'll close this one

[patent hamlet]

A new season and all it entails is being constructed as we speak. It will include changes to the OS rating implementation to ensure even the most dedicated BAR players do not see their rating stagnate, whilst also improving the balancing of lobbies using these ratings. Stay tuned

[glass robin]

What are these future changes? Are you talking about a min uncertainty or something else?

[calm kiln]

BAR is the only game where I've experienced rating gains decaying to infinitesimally small values. Should have some minimum threshold to prevent things from becoming stale.

[magic crystal]

Any plans for a larger balance patch to come out at the same time as the new season

[bold zodiac]

As expected, people are droping OS on purpose O_o. Thats why you asked for new season?

[fickle gazelle]

The only real solution to that problem is more players. Skill distribution in games like this is almost always extremely sparse at the upper limits, which means every player at 45+ OS will be essentially forced to carry every single game they are in, which is an understandably unpleasant situation to be in. For those over 50 it's even worse. Until there's matchmaking as well as the ability to fill a high-os lobby with similarly skilled players, the best players will continue to try and tank their OS to stay out of the alway-carry position.

I don't believe there is any change that can be made to how OS works that will solve that problem. There could maybe be other incentives (IDC suggested cosmetics or other perks for being above certain thresholds), but that would be mitigation at best.

[bold zodiac]

Yep i know, i had 50 OS before this sigma update but i don't know why i dropped 1,7 OS without playing. I think when you put game on steam it should make a player base bit bigger.

[buoyant dew]

Everyone dropped by something like that amount

[fickle gazelle]

match rating = skill - uncertainty. Everyone below 2.5 uncertainty was increased to 2.5, so you'd lose match rating proportional to how much lower than that you were.

[charred stag]

Setting aside the tanking players, which is annoying. I would expect this and any future uncertainty or OS reset to bring the top couple percentage points of players closer to the average. Season 0 started shortly after the mass influx of players from Winter and other streamers. Those people helped feed the OS inflation of the top players. BAR is still growing, but not the sort of mass surge we saw 10 months ago or so. So we have less brand new players for the higher players to drain OS off of.

The only way I could see a similar thing happening would be something like a BAR Steam release bringing in a mass surge of players, and inflating the top end players again. Which is good, we want a tighter distribution of OS

[crude eagle]

The simulations on the historical data actually increased, not decreased, the range of top and bottom players when there was a higher uncertainty floor. The increased variance causes people to swing higher and lower more easily and gives people more time at a higher uncertainty to move. When people stagnate in rating due to a high number of games played lowering their uncertainty it makes it hard to move, stopping them from dispersing as much. With a raised floor the top players were a good chunk higher and the worst players were a bit lower, but the upper range definitely shifted more.

[charred stag]

With a floor in place I would expect that to be right. But we don't have a floor, just a soft reset of what is it...1800ish players to 2.5? A floor would likely increase the distribution, a regularly occurring reset should decrease it.

[haughty light]

skills need to rebalance
that is partially why the uncertanty increase was done

tiny population of highly skilled players vs a large swarm of inexpeierenced players is going to cause severe balance quakes
which is why the OS inflation partially occured
that and the uncertanty reset originally didnt help with its timing i think

[crude eagle]

inflation or you mean some high OS people are overrated playing against noobs a lot?

[crude eagle]

That's how balance rating works, it's Mu - Sigma, so when the sigma was bumped up, the shown rating decreased, but the base Mu stayed the same.

That happened to everyone with Sigma (uncertainty) below 2.5

[haughty light]

yes that is something that partially happens

[dim tendon]

$text sigmareset

[cunning bear]

If you had below 2.5 uncertainty, your uncertainty has, as of 2024-02-21, been reset to 2.5. You will also see this as a slight decrease in OpenSkill rating (skill - uncertainty). This is not a bug.

See #announcements message for why. See the OpenSkill guide for explanation on what these terms mean.

This affects at most 2000 players. Congratulations! You grind a lot :stuck_out_tongue_winking_eye:

[haughty light]

its a stacking effect
the population change which occured in a single month basically caused a lot of issues
this would occur with TS btw

[elder parcel]

if uncertainty had a minimum value than that would decrease sandbagging
you lose X games you will only win X games later

[woven moat]

As we said in the announcement, this was a temporary mitigation, and more work will land down the road.

[sturdy cedar]

Take this idea with a grain of salt… Has anyone discussed a handicap feature for the players below the other team’s OS average? This incentive increases/decreases however far from the other teams average you are.

I wonder if this is something we can enable for the high OS players in a lobby setting, to help them not have to “carry every game”, which can get not fun.

Or the 50 OS player can take a self nerf to their economy to artificially drop their OS in that lobby.

Just an idea. Would be borderline impossible to balance.

[charred lava]

The incentive is that you lose less OS when you lose, and gain more when you win, if you're on the underdog team. The difference is larger in small teams generally, which is why playing the smaller formats moves the needle faster.

[turbid sparrow]

The uncertainty reset seemed popular. How often will there be new season uncertainty resets going forward?

[icy wing]

hopefully as soon as the balance is working properly again

[turbid sparrow]

Did balance ever work before?

[chrome quarry]

I'd like to see global LoS of player spawn locations. In our current situation I think it would be something that could dramatically improve game quality

It would allow you to see potential communism spawns as well as help with lane mismatch issues

[turbid sparrow]

Line of Sight? Wouldn't that give out information unrealistically that the other team could exploit?