was a suggestion to make min rd 75 but turned into a discussion ig?
Open
wolfeee0119 messagesstarted Oct 7, 2026, 1:14 AM
wolfeee0119 messagesstarted Oct 7, 2026, 1:14 AMgameplayother
wolfeee0OP2 weeks ago(edited)ive been thinking abt the elo stuff a bit bc i keep seeing people complain about how slow it feels to climb. personally i actually like glicko but i do get where people are coming from.
So I started messing around with the match data and ran a bunch of simulations just to see if there was some middle ground. the one i liked most was basically putting a small floor on RD, so once someone has played heaps of games glicko cant become too sure of their rating and barely move them anymore.
in close games the current system was around +11/-11, with an RD floor around 75 it was more like +15/-15, and era 1 was around +30/-30. so it kinda sits nicely in between without going back to the massive swings from before.
obviously theres still a downside bc losses get bigger too and ratings become a bit more volatile. especially since higher rated players already struggle to get matched with people around their elo, so they end up playing down a lot and getting barely anything when they win.
So I started messing around with the match data and ran a bunch of simulations just to see if there was some middle ground. the one i liked most was basically putting a small floor on RD, so once someone has played heaps of games glicko cant become too sure of their rating and barely move them anymore.
in close games the current system was around +11/-11, with an RD floor around 75 it was more like +15/-15, and era 1 was around +30/-30. so it kinda sits nicely in between without going back to the massive swings from before.
obviously theres still a downside bc losses get bigger too and ratings become a bit more volatile. especially since higher rated players already struggle to get matched with people around their elo, so they end up playing down a lot and getting barely anything when they win.
6
GracefulBeauty2 weeks ago
i kinda feel like the low rating gain/loss is an actual problem due to rounding
GracefulBeauty2 weeks ago
you could get +14 for a win that should've been +14.4, and that's actually a pretty significant difference right now
wolfeee0OP2 weeks ago
Hmm that’s interesting never thought abt that
GracefulBeauty2 weeks ago
especially for top players playing lower rated players, gaining +3 instead of +3.4 is a huge difference
wolfeee0OP2 weeks ago
Ye I like that
GracefulBeauty2 weeks ago(edited)
on average fractional rating lost/gained due to rounding should even out towards zero, but it also might accumulate for a while, and knowing that feels kinda bad
wolfeee0OP2 weeks ago
Maybe every time it accumulates to plus 3 or something it adds on
GracefulBeauty2 weeks ago
i think it'd be fine if the fractional part was hidden (maybe it already is!) and accumulated correctly without rounding
wolfeee0OP2 weeks ago
I’m curious to see how much top rated players lost bc of that
GracefulBeauty2 weeks ago
also this is slightly tangential to your OP so my bad 
wolfeee0OP2 weeks ago
No it’s good it’s good to discuss other ideas
wolfeee0OP2 weeks ago
we could also make glicko more responsive based on performance, like if someone keeps doing better than the system expects it temporarily lets their elo move more, then settles back down once it catches up.
GracefulBeauty2 weeks ago(edited)
that's what i initially thought volatility did, but apparently it has the opposite effect (i'm not entirely sure)
wolfeee0OP2 weeks ago
if we accumulate the fractional elo, on avg every 100 games it would add only about 0.2 extra elo so it wouldn’t be a lot
GracefulBeauty2 weeks ago
oh you simulated it ig
GracefulBeauty2 weeks ago
i imagine it depends on the range of players you tend to queue and which games you win/lose
wolfeee0OP2 weeks ago
yeah sorta, higher volatility actually generally makes the rating more responsive, not less. its just indirect and glicko normally changes it pretty slowly. what i meant is more like deliberately increasing uncertainty when someone keeps outperforming expectations, then letting it settle again.
wolfeee0OP2 weeks ago
yeah exactly, thats what the sims showed too. on avg it basically cancels out, but for an individual player it can drift a few elo either way depending on who they play and which games they win/lose. i’ve got a database with all the saved matches so i can test quite quickly
sugaarush2 weeks ago
iirc marlin has said they’re looking to tweak the elo changes
sugaarush2 weeks ago
i think looking at ur sims is a rly good idea
sugaarush2 weeks ago(edited)
it’s a problem bc winning a 40 min gives 11 elo which is like 9% of one tier
wolfeee0OP2 weeks ago
ye we could also facotr in time spent in the game there are endless possibilities
sugaarush2 weeks ago
ig but my point is that it doesn't matter if the elo system is perfectly judges someones skill to the nearest 0.1%; its more important that its inviting and rewarding to play more games
1
wolfeee0OP2 weeks ago
yep couldnt agree more
Threadfinn2 weeks ago
I think the issue with that is the longer games will usually be the ones with close scores, so it wouldn't really make sense to gain for for a 13-12 win than a 13-0 win. The only way to counteract that would be to consider match score for elo gain, which they've already said they don't want to do.
2
wolfeee0OP2 weeks ago
ye ive never really explored that alley of increasing elo gain accoridng to match length but it could maybe be a tiny part of the equation if proven to actually be reliable
noura 🜛2 weeks ago
what is the suggestion
GracefulBeauty2 weeks ago
raising min RD to ~75 sounds like
noura 🜛2 weeks ago
i think having a clearer title would be helpful
11
Malox2 weeks ago
Could be a solution, also I feel like it should depends on the score. Winning 13-12 is not the same as winning 13-5
Malox2 weeks ago
You should win more from stomps and lose less from close game
0200
Malox2 weeks ago
Would make it less annoying too if you lose because of 1 RNG goal in the end
7rowl2 weeks ago
The issue with this is that downsides outweigh the upsides. People would rather win more elo than lose less. Really close wins and comebacks become a lot less satisfying and when you do get stomped it feels even worse
Malox2 weeks ago
That’s true, but I feel like it would represent better your actual elo
Malox2 weeks ago
It’s how it is done in most of the ranked games and although it would sure feel bad to get a -30 or so from a stomp, it would be « deserved » and it would also incite people to FF less
pinkpuff2 weeks ago
Another problem is Playing for goal dif is a different playstyle than playing to win, and you could then get "rewarded" for a close game but never have actually had a chance at winning
5
Malox2 weeks ago
Indeed… probably not that good of an idea then, I was wrong !
1
wolfeee0OP2 weeks ago
was a suggestion to make min rd 75 but turned into a discussion ig?
7rowl2 weeks ago
Im pretty sure that is not how its done in most ranked games tho
Malox2 weeks ago
Isn’t it ? Riot Games work that way at least, most battle royal I played too
Malox2 weeks ago
But I may not have a scope of 1v1 ranked games wide enough
Threadfinn2 weeks ago(edited)
Threadfinn2 weeks ago
I was typing this and it was getting too long I'm kinda just rambling here
Threadfinn2 weeks ago
Basically though, I think score should affect elo in draftout, since winning the match would still be far more important than winning every goal in terms of your overall elo gain.
Threadfinn2 weeks ago
And having score matter would make 13-12 losses feel a bit less crushing (me dying to a sniper skeleton + creeper combo and losing 400 elo because it was my first match and I lost 13-12 to another unrated player)
Threadfinn2 weeks ago(edited)
Also since dying is a thing, even if you have a strategy that should guarantee you a close win, you can die and still lose. If you're already up 12-2 a death probably won't matter, but if you're up 11-8 dying can make you lose your winning line. Point just being that score still matters even when you should be super behind.
Threadfinn2 weeks ago
I know this idea has been rejected but I feel like it could be reconsidered.
GracefulBeauty2 weeks ago
another problem with taking score into account is that tournaments would NOT take score into account. it's win or lose
GracefulBeauty2 weeks ago
and the type of gameplay/decision-making that you learn from playing competitive would not apply in tournaments
GracefulBeauty2 weeks ago
like if you're up 12-5, you've been in the nether for 5m, and enter end is on the board, in tournament the correct play would be to go for enter end and let your opponent waste their time collecting overworld goals until you get your guaranteed win 13-12
GracefulBeauty2 weeks ago(edited)
but in competitive the correct play would be to leave the nether and try to quickly steal one of the overworld goals so you gain more rating
GracefulBeauty2 weeks ago
also a lot of times the goal diff doesn't depend on the skill difference between players, it depends on the board / seed
GracefulBeauty2 weeks ago
if there is one obvious very linear line of goals, then whoever is ahead on that line will win with a huge goal diff
GracefulBeauty2 weeks ago
if there are two strong lines, then whoever plays their line better / snipes the other's line better will win with a very small goal diff
GracefulBeauty2 weeks ago
it becomes even more complicated when you factor in forfeits
GracefulBeauty2 weeks ago
say you're up 6-4 but your opponent enters nether first, and the nether line is VERY strong and linear and will take them to 13 goals. do you forfeit immediately to avoid losing more rating than if you were to play out the game?
GracefulBeauty2 weeks ago(edited)
and would forfeiting early there actually make you a better player? (because you would lose less rating on average than playing out the game, so a player that does this would be rated higher than one that does not)
GracefulBeauty2 weeks ago
or is it gaming the rating system
Threadfinn2 weeks ago
But doing this, you risk losing 30 rating to try and steal 5.
Threadfinn2 weeks ago
It would have to start counting forfeits as the winner gaining 13 to avoid this. That's how it works in other games, like in valorant if you ff at 1-4 your rating loss is the same as a 1-13.
1
GracefulBeauty2 weeks ago
it's a very small risk and you can always pivot back to enter end, so it would 100% be optimal
Threadfinn2 weeks ago
But you might lose your advantage on enter end by going back to overworld
GracefulBeauty2 weeks ago
rating systems incentivize players to play for the highest expected rating gain
GracefulBeauty2 weeks ago
so if there's a 95% chance to gain +10 rating for the play and a 5% chance to lose 20 or w/e
GracefulBeauty2 weeks ago
that's an expected rating gain of 8.5
GracefulBeauty2 weeks ago
so it's optimal
GracefulBeauty2 weeks ago
in tournaments the calculus would be completely different
Threadfinn2 weeks ago
To get +10 with what I was saying you'd have to win 13-2 vs winning 13-12 going for enter end. Even feinberg would currently play for the faster game because if you're already up 12-2 your chance of losing all 11 remaining goals is lower than the chance of dying in the fort going for enter end.
GracefulBeauty2 weeks ago
your chance of dying in the fort when you're feinberg and you're ahead by like 5m on enter end so you can go as slow as you want is basically 0%
Threadfinn2 weeks ago
Yeah but your chance of losing 11 goals in a row as feinberg is also 0%, better to save 15 minutes
GracefulBeauty2 weeks ago
the thing is the game is about managing probabilities and incentives
GracefulBeauty2 weeks ago
goal diff perverts the incentives
GracefulBeauty2 weeks ago
it affects decision-making for the entire game
GracefulBeauty2 weeks ago
splitting lines but claiming the stronger line so you win a close game becomes the wrong play for some reason
Threadfinn2 weeks ago
No it's still the right play, because the other play risks losing
Threadfinn2 weeks ago
I'm not suggesting making goal diff the only factor
GracefulBeauty2 weeks ago
but we said that the rating system isn't binary anymore
GracefulBeauty2 weeks ago
the players that play to maximize rating will be higher on the lb
GracefulBeauty2 weeks ago
risking losing so that on avg you gain more rating is optimal
Threadfinn2 weeks ago
If win vs lose is +15 instead of -15 and goal diff makes up for +/-5, then winning the game is still 83% of the game.
GracefulBeauty2 weeks ago
GracefulBeauty2 weeks ago
what i'm saying is just mathematically true idk
GracefulBeauty2 weeks ago
you could argue that it's not a big deal but it just is an actual effect that would happen
Threadfinn2 weeks ago
I think what you're skipping here is that while the "safe" play gives a 95% chance of winning, but you're in a position where the "risky" play is a 90% chance and saves you time.
Threadfinn2 weeks ago
I think you are right that it becomes slightly less competitively correct
Threadfinn2 weeks ago
But I think that effect is minor, and it would make the system feel much more rewarding
Maloxlast week
tbh, we're already only wining +10 even when stomps, we might as well add +1/goal diff
Maloxlast week
and tbh, you'd probably still win 13-8/13-9 if you decide to go for the end line, ur opp isn't doing 8 goals while you're doing any%
Maloxlast week
or if he can, then it's slop goals and it's better for you to exit anyway
577last week
finding a way to guarantee a win using your goal lead, rather than just going for a snipe to end the game faster, is a skill that should not be penalized imo
577last week
goal diff is like not really correlated at all with a particularly large gap in performance per game anyway
577last week
you can lose 0-13 because you were 1 tick behind on the same 13 goals
577last week
in fact ending the game faster is already enough of an incentive to throw a lot of ppl off from this form of optimal play
577last week
it just sort of decouples the elo reward system from what is Objectively better play
Threadfinnlast week
While that is true, you probably shouldn't play the same line as your opponent if you know you're slower. If you keep losing goals by a few seconds you really need to think of a way to save some time. Losing 13 goals by just a couple seconds means you did the same thing as your opponent, but we're worse on every single step of the way, and that probably should be punished.
Conversely, it is rare to have a 12-13 game where the player who lost by 12 never had a chance. If one goal had flipped due to rng, the other player might have won.
Conversely, it is rare to have a 12-13 game where the player who lost by 12 never had a chance. If one goal had flipped due to rng, the other player might have won.
Threadfinnlast week
I'll admit I'm partly just annoyed since all my losses have been 12-13. But every single one of those had multiple chances for me to win if just small things went differently
577last week
you dont know what your opponent is doing, and even if you somehow did (or guessed correctly) there are still cases where the objectively best chance you have of winning is saving the time u lost back on mechanics
577last week
its like if youre 5 seconds slower into the nether and theres 6 fast nether goals you kind of have to go there to win, but you end up in a situation where either you have a chacne at winning or you'll lose by like 6 goals
577last week
with no real middleground
Threadfinnlast week
If I'm playing against someone who beat me on wood tools and then beat me to the first structure, I already know I'm going to lose on mechanics. Trying to do the most obvious line I'll just lose, it's better to play to get lucky sometimes I think.
577last week
Conversely, it is rare to have a 12-13 game where the player who lost by 12 never had a chance. If one goal had flipped due to rng, the other player might have won.also i think this is Just Not True, ive had many games with like 1-2 goal diff where either me or my opponent was basically guaranteed to win for the majority of the game
Threadfinnlast week
I had a guaranteed win here but I fucked it up twice on die to game design
Threadfinnlast week
If they'd only had 3 goals then that mistake wouldn't matter
Threadfinnlast week
Because I'm ahead by so much I can pivot
Threadfinnlast week
But since they had 11 also, it gave them an opening
577last week
im confused what the point is with this example
577last week
it only makes sense to reward goal diff if it's an indicator of skill gap in every match, otherwise the incentive of "having more resources to pivot if something unexpected happens" is already enough of a reward in that subset of cases
Threadfinnlast week
On a 12-3 lead I can do whatever I want and win, even if I throw a bunch of goals. On a 12-11 game even if I have both goals secured, my opponent could still win if I make a mistake.
577last week
so why should the elo system disproportionately incentivize you to make decisions that are more likely to throw a bunch of goals
Threadfinnlast week
?
Threadfinnlast week
Wdym
577last week
if you have 12-3 and you get extra elo for goal diff then you are incentivized to go for whatever your opponent is doing instead of a guaranteed goal that you know they cant do
577last week(edited)
which allows them to take more goals if you do it badly (or if they get luckier, have some information that u dont, etc)
577last week
its objectively a bad play if your goal is "win as often as possible"
Threadfinnlast week
I don't see how potentially throwing a game is incentived?
Threadfinnlast week
If one player goes +14, +14, +14, -20, -20 they're up 2 elo. All their wins were close because they got for enter end, nqnl with no other nether goals, and breed 8 instead of contesting anything. Their losses they just lost.
Someone else goes +20, +20, -13, -20, -20 and is down 13. They force the easiest goal to try and win by a lot, but it cost them a game because their opponent remembered how to play.
Safer plays would still be rewarded over time. It would just make close matches like 12-12 XP bottle punish the loser slightly less.
Someone else goes +20, +20, -13, -20, -20 and is down 13. They force the easiest goal to try and win by a lot, but it cost them a game because their opponent remembered how to play.
Safer plays would still be rewarded over time. It would just make close matches like 12-12 XP bottle punish the loser slightly less.
Threadfinnlast week
Sorry for delay I'm at work and had to work.
Threadfinnlast week
Keep in mind +14 would be a 13-9 win and +20 would be a 13-3. The player going safe let their opponent get 6 goals for free to secure 1, which inherently is probably more risky that just challenging those 6 goals.