Scale elo gained/lost with goal differential (per game)
Open
Droid37 messagesstarted May 23, 2026, 6:58 AM
Droid37 messagesstarted May 23, 2026, 6:58 AMgameplay
DroidOPlast week(correct me if this is already the case i looked through a few games and it did not appear so, unless it is very slight)
makes close losses less punishing, and rewards getting 13 significantly faster. while large goal gap is not 100% accurate representation of skill diff (losing player couldve taken same line but a bit slower), it approximates it reasonably.
gain/loss would still be calculated based on elo diff in conjunction with this. also, forfeits could be treated as if the winner got 13 to encourage playing out losses for a couple more goals
makes close losses less punishing, and rewards getting 13 significantly faster. while large goal gap is not 100% accurate representation of skill diff (losing player couldve taken same line but a bit slower), it approximates it reasonably.
gain/loss would still be calculated based on elo diff in conjunction with this. also, forfeits could be treated as if the winner got 13 to encourage playing out losses for a couple more goals
127
r0hkxlast week
this would add weird gameplay incentives and make close games less exciting
r0hkxlast week(edited)
a 13-12 win is a win as much as a 13-0 win. the main benefit to rewarding decisive wins more would be so the ratings converge sooner, but that doesn't really matter since you can always just play more games
9
DroidOPlast week(edited)
so the ratings converge sooner, but that doesn't really matteri do think it matters somewhat given the long game times, but i agree the effect of it shouldnt be so large that a 13-12 win isn't much better than a draw. it'd just be a 5-10 elo difference or smth from the usual gain (if players are equal elo)
chessmincraftlast week(edited)
It should be equal regardless of how close the game is. It makes the strategy quite strange if you change the goal from winning to getting the most goals you can.
chessmincraftlast week
Although maybe less elo change in general...
r0hkxlast week
another problem is that you don't really have full control over the goal difference. depending on the board and seed, you could easily have wildly different scores against the same opponent
DroidOPlast week
less elo change overall makes it very tiring to climb ranks
DroidOPlast week
this is true but in general a 13-12 game was probably closer than 13-2
r0hkxlast week
for example, if most goals are very quick and snowballable, even a very good player might lose like 13-3 if their opponent is just slightly better. on the other hand, if most goals are more difficult and don't chain very well, its easily possible for a player that is much worse than their opponent to only lose like 13-10
1
r0hkxlast week
the score alone doesn't tell you very much about the players' skills
r0hkxlast week
on the aggregate it can be an interesting statistic, but it just shouldn't be part of the rating system
r0hkxlast week
rr is not really the valorant rating system
r0hkxlast week
the actual rating is the hidden mmr, and that likely doesn't have any "performance boost" or similar
r0hkxlast week
rating systems in 1v1 games should pretty much only look at win/loss/draw unless there is a very very good reason why that is not enough
DroidOPlast week
fair
r0hkxlast week
for rating systems in team games, usually win/loss/draw is all you need as well, and if its a game like valorant where you're constantly playing with different players, your individual rating gets sorted out by looking at correlation of you being on a team and winning, rather than any individual stats like kills or damage
DroidOPlast week
team games its just too difficult to quantify your contribution to the win/loss anyway
1
DroidOPlast week
but also i was thinking about like
DroidOPlast week
itd be really bad or weird to do something similar in ranked where split difference was accounted for
DroidOPlast week
idrk what justification i have here other than the game length
r0hkxlast week(edited)
(i don't like just silently downvoting but i disagree but don't want to derail the thread)
1
blanket addictlast week
definitely do not do this decoupling incentives from always playing for the win sounds awful
blanket addictlast week
I think it just doesn't work well with elo anyway elo works best when you're just scaling off of winning/losing
Threadfinn5 hours ago
In valorant the mmr is almost entirely performance and score based as far as I know, win/loss barely even affects it in lower ranks. At the top it's weighted differently so winrate and score matter more than performance, but even then the score part still has a huge impact.
I think scaling elo by score works for draftout because there are discrete goals. If the final score is 13-12 then one player won on 13 goals and the other player won on 12. The player who got 12 still lost so they should lose a decent amount of elo, but it's functionally very different from a blowout game. Something like ranked or chess are different because there is only win, there are no intermediate goals (ranked splits don't work as intermediates because of pacebaiting).
If I just show you a game score and ask you to predict the result of a rematch, a 13-0 to 13-3 is almost always going to stay the same, while a 13-12 is far more likely to flip. The elo gain should reflect that imo. Yes it's possible the player who lost 13-0 just got unlucky, but it's also possible to go on a loss streak across multiple games due to bad luck.
A change could be instead of using a base K value of 60 like now, use a K of 40+2*(winner score-loser score). This leads to the same elo change at 13-8, but close games between two people at the same elo would be more like +/- 21 and a 13-0 would be 33 instead of the current change of 30. Impactful enough to notice, but not so strong it overshadows the real goal of winning.
I think scaling elo by score works for draftout because there are discrete goals. If the final score is 13-12 then one player won on 13 goals and the other player won on 12. The player who got 12 still lost so they should lose a decent amount of elo, but it's functionally very different from a blowout game. Something like ranked or chess are different because there is only win, there are no intermediate goals (ranked splits don't work as intermediates because of pacebaiting).
If I just show you a game score and ask you to predict the result of a rematch, a 13-0 to 13-3 is almost always going to stay the same, while a 13-12 is far more likely to flip. The elo gain should reflect that imo. Yes it's possible the player who lost 13-0 just got unlucky, but it's also possible to go on a loss streak across multiple games due to bad luck.
A change could be instead of using a base K value of 60 like now, use a K of 40+2*(winner score-loser score). This leads to the same elo change at 13-8, but close games between two people at the same elo would be more like +/- 21 and a 13-0 would be 33 instead of the current change of 30. Impactful enough to notice, but not so strong it overshadows the real goal of winning.
1
Threadfinn4 hours ago(edited)
this too. There's no reason to play when down 4 goals (right now at least)
Threadfinn4 hours ago
It's not just the ranked ff mentality imo. I think it's currently more beneficial to ff in draftout than in ranked. In ranked i usually don't ff when behind because there's that chance my opponent burns in lava or dies in end. In draftout once behind by enough goals, even if my opponent dies multiple times they'll still probably beat me on at least some of the remaining goals if I don't already see a good line for myself.
Threadfinn4 hours ago
And as for this, I think it isn't too bad in this regard as long as the impact of score isn't overinflated. Like with what I suggested the difference between a 13-0 and a 13-12 is 33vs21 elo. 2/3 of the elo gain is still a result of the win, so you could play a confident 13-12 line where you think you have a good chance of winning a close game, or you could risk pushing to try and win at all the early game goals to try for more elo, but risk a much higher chance of losing.
Threadfinn4 hours ago(edited)
I play Valorant so I'll use that as an example. If you win 13-0 there you get a lot more rr (elo) than if it's a close 13-11 game. Despite this, almost nobody always plays to win every single round. Everyone accepts that sometimes you have to sacrifice a couple of rounds you probably already lost anyways, because it can give you a better chance at winning those future rounds.
blanket addict2 hours ago(edited)
I think disincentivizing forfeits is not a good thing tbh
blanket addict2 hours ago
I think the downside of you get the win but the game just stops is a lot less bad than feeling trapped in a game you really really really don't want to play
blanket addict2 hours ago
and it'd make elo less accurate