New ranking system to stay or?

[Nore]

To be fair, I think only people actually playing(cough cough) should be able to talk on this subject.. just saying.

[archol]

Unfortunately i often see scores consisting 29,50 vs 0,50 elos, though the random teams are something refreshing indeed

[aadhar]

Hahaha. I wish I had ever played and saw this new and old ranking system

[pewpew lasergun]

To be fair, I think only people actually playing(cough cough) should be able to talk on this subject.. just saying.

i think his opinion matters, he is a pioneer lihl mod since lihl was created, and he knows very well the elo system and lihl.
u cant tell inactive players like fiji, pata, eld .... that their opinion doesn't matter, they made lihl.

[Nore]

Welll, at least all those people you named have played within the past season or so.. whereas this particular person has not in ages... going on 4-5 seasons if not more

[Dong]

imo, it is a shame that it feels like alot of players are dodging lihl games atm, because idk, they dislike the new beta system, or because they are afraid to lose elo, who knows. But fact is, that players that have weakness regarding calls get punished more, from what i have seen, i have played about 300 games now, at it kinda seems that atm, it is the same people playing over and over, " farming/gets farmed " and when you have a low player base, with a no limit elo to take from, this season will set a couple of records but yea, random teams are nice, i just feel that i have noticed a lot of games where the " score is somewhat close to 29/1 or close to that, which i find kinda sad it is funny to beat a superior team, but fact is, that you will lose more than you will win, against such a team

My point, was, i would like to see more players play, so we can get a better overall picture of this new system.

Dong out-

[archol]

Maybe doing captain mode (or a similar one where you can at least choose 1 teammate and get 2 randoms for example) will be benefital in that case

[Jamo]

I still like the random teams, I guess everybody agrees on that (from what I read so far). The only thing I still don't get is the 15/15 elo games. Very often I heard "In a league it should not be a too big difference in elo". But I don't see why tbh. Now it seems like Dong already pointed out, many people are just not playing lihl, maybe because of the reason he mentioned. Okay, the top players don't dodge now because they don't fear losing 28 elo. But as I already wrote once, for lower elo players this is completely demotivating. I mean, it's obvious that it's harder for the low team to win, so why not giving them the elo they deserve?
Maybe, one could make it a step function such, that the !scores are checked, and then it's either a 15/15, 20/10, or 25/5 game (or vice versa), depending on the range the actual scores are in. To avoid 29/1 games, which I can understand do suck.

[pewpew lasergun]

here is the best answer

This can be achieved by randomising teams and retaining the elo scaling of the previous system. This will result in greater differences in elo which can be counteracted by scaling the elo formula to be less harsh.

maybe supersexyy can give us some scaling example.
i think he meant a 29/0.5 will be capped at something like 20/10 rather then 15 both sides.
so it is random but also takes elo in consideration

[elektro]

I like new sistem (15 15 elo)

[supersexyy]

here is the best answer

This can be achieved by randomising teams and retaining the elo scaling of the previous system. This will result in greater differences in elo which can be counteracted by scaling the elo formula to be less harsh.

maybe supersexyy can give us some scaling example.
i think he meant a 29/0.5 will be capped at something like 20/10 rather then 15 both sides.
so it is random but also takes elo in consideration

The elo gain/loss is supposed to reflect the % chance of winning/losing. For example using a generic chess elo formula:

A player whose rating is 100 points greater than their opponent's is expected to score 64%; if the difference is 200 points, then the expected score for the stronger player is 76%.

Hence, if player A is 1100 and player b is 1000, player A is expected to win 64% of the time and the elo will reflect that. So if we run the test 100 times we will get 64 wins for player A and 36 wins for player B. The elo gain from winning 64 times for player A would be equal to the elo loss for player A for losing 36 times. Same, for player B but inverse.

So assuming both players are their 'true' elo then neither player will change after 100 games which is the point of the system. Obviously if a player is not their true elo they will gravitate towards that number.

Only problem is that 1100 vs 1000 in lihl does not have a 64/36% win ratio. Same could be said for a 200 and 300 elo difference. If we say 1100 vs 1000 elo has a 55% win ratio then the elo win/loss will more equal.

Obviously we should assume all players are at their true elo when estimating win %s. And before anyone says 'but players often aren't at their true elo hence the elo formula is broken', that is completely irrelevant to the mathematics of the formula.

Tldr elo formula can be scaled to reflect win %s of lihl as opposed to chess.

Hope that makes sense...

[dweiler]

I see a lot of misunderstandings about the current system here, and since I was the one proposing it, I will try to make some things clearer.
The random teams thing – I won’t go into that, seems like many people like it.

The Elo. Actually, there is no Elo this season. You should rather see it as a points-system. You get a point (in this case +15) for a win and a –point (in this case -15) for a loss. The reason it is still ‘looking’ like Elo is to not change the system too much and to be able to apply dc-penalties. So looking at +29/-1 games is wrong, every game is a point for a win, a point off for a loss (or in here, +15/-15).

There are several reasons Elo is taken out. 1. In old system, a good player would receive less points for winning with same allies vs same opponents as a player who is worse than him. I think it’s wrong to ‘punish’ good players. Same other way around, I don’t think it’s good to boost bad players. It’s the most fair if everyone gets a point for winning and loses a point for losing, no need to punish people for being good (and other way around).

2. Elo does not encourage activity. If we look at top 3 this season and last season, the amount of games played is pretty much equal already with still 1 full month to go and the bots being offline for about a week this season. Also in lower ranges, activity is encouraged (mainly in the >50% area). This is not because of 'dodging' but because more games don't matter with Elo. There is just nothing to play for when you reached your Elo, whether it be 1010 or 1500.

3. With encouraging activity in the >50% area, the general level of the league will go up because of more games between players on the higher ends of the league.

What you could see as a disadvantage of points-system is that activity for <50%-players is discouraged. Someone with 100 games with 46% is higher score than someone with 200 games with 47%. I think it is up to the players to decide if the advantages outweigh this disadvantage.

[pewpew lasergun]

what others are saying mickey is that ur point system 1 point for win 1 for loss is flawed.
because it doesn't reflect the probability of winning/losing.

take for example tinsoldier/1007/beep/donk_illuminati VS dong/ace_of_spades/dodoyum/nabo.

giving 1 point to the first team for winning ( assuming a 1 point loss if lost ) is wrong.
the probability of the first team winning is something like 95%.
and it happens all the time, thats what players meant by 29.5/0.5, means they have virtually no chance to win.
the risk reward ratio 1:1 is not beneficial to the weak team in this case.

thats why i think random teams with scaling like supersexxy suggested is the best. its the best of both worlds, u get random teams and the probability of winning/losing is accounted for. ^^

one more thing to remember is that players love elo, its their reward for winning, same goes for other rewards like colored spoof,its human nature, disrupting elo is not good for activity. imagine lihl without any elo at all, what will happen to the activity? management knows this very well.

if ratio was 1:0.5 everyone happy

WC3ScrnShot_033116_165743_01.jpg (253.67 KiB) Viewed 610 times

this is how u ask when game ends:

WC3ScrnShot_033116_171550_04.jpg (369 KiB) Viewed 599 times

they changed the game to 4s and used !sp or votebalance

the game hosted just after that:
we ended up using !votebalance at the request of second team

WC3ScrnShot_033116_172337_07.jpg (227.45 KiB) Viewed 599 times

[Diablo_]

Good players will win more, bad ones will lose more (theoretically, LTD is pretty luck based anyway). Does that mean the system is flawed? It definitely can be seen from both sides. Good players being dragged down by the system and bad players being pushed isn't necessarily fair either.

Like with the old system bad players could just relax and let the balancing do the job to not go too low, now they actually have to fight on their.

Imo both options are viable and generally I would prefer auto balanced games, but in LIHL I like the random teams a lot (since we are already the "top players" random teams don't result in ultra unfair teams anyway). Whether the Elo is 15/15 or an updated Elo function (like max 20/10 and the curve being less steep) I don't care too much.

[supersexyy]

If you're saying low players are protected by the old system through getting larger elo gains, you adjust the scaling to be more reflective of the game.

Essentially allowing the rubber band to stretch a bit more.