Jump to content

Recommended Posts

Not sure if there already is a thread about this but searching elo led to an error so here I am.

 

The current elo calculation is imo not what it should be and I'm here to make some suggestions that will hopefully be met with mutual consent and help in improving it a little.

 

Atm the "farm" hero will most likely get the highest elo of a team.

This is due to the fact that creeps kills (woods & regular) give way too much elo compared to hero kills/assissts.

        -> this leads to players that did not contribute to the team`s success at all getting almost the same,

             often more elo as the player(s) who deserve it and therefore falsifying the overall elo score.

 

So basically I'm suggesting giving lower elo per creep and add more to hero -kills & assissts.

 

What I also thought of is that rule breaking (intentional feeding, item stealing, flaming, afking and such) should have a negative impact on that players elo (e.g. -20 elo for game ruin).

        > Maybe even show how many offences a player has made on her/his profile/stats page.

 

The leaving offence should be dealt with separately imo, as it is often unintentional (dc, powerloss, blackout,...).

 

 

Edited by FatmanKROUM
  • Upvote 1

Share this post


Link to post
Share on other sites

Everything of the game that is being recorded is used to calculate the player efficiency.

Farming in the woods or on the lane is often the key to win a game. This is the reason why i have influenced it on that rate.

Currently one kill weight as much as 20 creeps. If you count that too much, well lets compare it.

When you get killed you will loose around 200-750 gold and gain the same on a kill. This depends on the streak. Mostly it is around 300 gold.

Compared to the the creeps that are giving you areound 37 gold * 20 sums up to 740 gold.

The XP does not change a lot on 20 creeps vs a kill.

To sum it up, i think the rating between creeps and kills is fair enough. 

But i was talking on normal creeps. The weight on neutralkills is 1 kill to 30 neutral creep kills.

So if you are kill heavy but unable to farm, you will not gain more compared to a good farmer which might dont get a lot of kills. Which i think is totally fair. Kills are not the only factor that decide a game.

 

I dont know why i repeat it like each second day, but i will not punish anything that manipulates the elo of a player, nor do i add 'booster packs'.

The script itself does work well with some minor tweaks of course, but adding a negative impact on disconnect would kill the algorithm.

If you want i can add a simple +5/-3/-10  win/loose/punishment script that you can see aside. It is bullshit having an alogrithm that punishes a player with reduce the elo points. That is only done by people that are not aware what an algorithm is.

But as said, i am tinkering another script out currently that may is a better option. Not sure yet.

Greetings

  • Thanks 1

Share this post


Link to post
Share on other sites

Well I also think that elo system is weird in this way how described fatman, like just now we finished game, if you checked the elo in game is was like 29.70/0.30 , the whole team fought hard for the win and we all deserved a great elo for that, we fought with the motivation that we got a lot of elo if we gonna win it, but in the end only i got the high elo, and the others got shits like 5 or 7elo .... mostly its due bcs i got X-times more creeps then they

https://ohstats.net/en/game/61009

It can be disgusting for the ppl who fought hard  .... 

Question is: why I should take risiko and play vs stacked teams for 5elo, if I can play with stackers and get the same elo, and everything what i need to do is take some creeps in wood

Not even talking about that the whole enemy team lost only avg.14elo, while they should be close to -29elo ..... 

Share this post


Link to post
Share on other sites
11 minutes ago, xvideos.com said:

Well I also think that elo system is weird in this way how described fatman, like just now we finished game, if you checked the elo in game is was like 29.70/0.30 , the whole team fought hard for the win and we all deserved a great elo for that, we fought with the motivation that we got a lot of elo if we gonna win it, but in the end only i got the high elo, and the others got shits like 5 or 7elo .... mostly its due bcs i got X-times more creeps then they

https://ohstats.net/en/game/61009

It can be disgusting for the ppl who fought hard  .... 

Question is: why I should take risiko and play vs stacked teams for 5elo, if I can play with stackers and get the same elo, and everything what i need to do is take some creeps in wood

Not even talking about that the whole enemy team lost only avg.14elo, while they should be close to -29elo ..... 

Did you ever calculate on things you have said?

 

In fact, you have twice the amount of kills the rest of your team has, same goes to creep kills and towers. So the reason you get sucha  big cake is simply because you were the one contributing the major factor on winning the game.

 

Also it is against stacked players, your team has got 70 elo points overall, where you got fairly the half.
The enemy team lost 70 elo points.

If they would have won, they would have gotten only 30 elo points, and your team would have lost only 30 points.

The ingame "scores"  command is not up to date. It is using the old elo algorithm with the new scores. Need to update that.

If you want to read a bit into details how the rate is defined, i have announced it a while ago here:

 

  • Haha 1

Share this post


Link to post
Share on other sites
4 hours ago, GriefCode said:

Did you ever calculate on things you have said?

 

In fact, you have twice the amount of kills the rest of your team has, same goes to creep kills and towers. So the reason you get sucha  big cake is simply because you were the one contributing the major factor on winning the game.

 

Also it is against stacked players, your team has got 70 elo points overall, where you got fairly the half.
The enemy team lost 70 elo points.

If they would have won, they would have gotten only 30 elo points, and your team would have lost only 30 points.

The ingame "scores"  command is not up to date. It is using the old elo algorithm with the new scores. Need to update that.

If you want to read a bit into details how the rate is defined, i have announced it a while ago here:

 

i think the ELO Change message with !scores needs to be adjusted, so ppl can somehow see more clear how much they can gain. or take the elo change away if thats not possible,

i'm thinking ppl get confused or think they can lose 0.67 elo or gain 29 which is not correct :)

Share this post


Link to post
Share on other sites
13 minutes ago, l2p.com said:

i think the ELO Change message with !scores needs to be adjusted, so ppl can somehow see more clear how much they can gain. or take the elo change away if thats not possible,

i'm thinking ppl get confused or think they can lose 0.67 elo or gain 29 which is not correct :)

 

5 hours ago, GriefCode said:

The ingame "scores"  command is not up to date. It is using the old elo algorithm with the new scores. Need to update that.

 

Share this post


Link to post
Share on other sites
On 11/17/2017 at 11:52 PM, GriefCode said:

Currently one kill weight as much as 20 creeps. If you count that too much, well lets compare it.

When you get killed you will loose around 200-750 gold and gain the same on a kill. This depends on the streak. Mostly it is around 300 gold.

Compared to the the creeps that are giving you areound 37 gold * 20 sums up to 740 gold.

The XP does not change a lot on 20 creeps vs a kill.

To sum it up, i think the rating between creeps and kills is fair enough. 

 

First off, it's obviously really hard to be fair with the algorithm (fitting each individual game), but what I'm trying to say is that it's not just about how much gold you get for one hero kill compared to X creeps.

 

To showcase what a hero kill means for a game; a hero kill will in fact:

                > increase the morality of your team

                > decrease the ENY`s morality

                > create a hole in the ENY defence

                                    (leaving it vulnerable to a push or simply exposing a lane to a free farm)

                > the killed ENY can't get any exp during respawn

                > give X amount of gold and xp

 

To showcase what multiple creep kills will do on a lane:

               > push that lane (can be countered by 1 hero kill)

               > give X amount of gold and xp

woodcreeps:

               > give X amount of gold and xp

 

So to sum it up, it's definitely not just about the gold but rather about the psychological damage you inflict on the ENY and the huge morale boost you give to your team.

 

Thusly, how would i compensate for all that extra stuff you get from hero kills over creep kills?

Well I'm really not to sure myself but.... I guess 1 hero kill should equal roughly  28-35 normal creeps or 40-45 wood creeps.

 

As always, thanks for reading.

Edited by FatmanKROUM

Share this post


Link to post
Share on other sites

An algorithm is never about feelings and ingame actions, thats why it is an algorithm. It just gets a resultset which is used to calculate a players efficiency.

 

9 hours ago, FatmanKROUM said:

Well I'm really not to sure myself but.... I guess 1 hero kill should equal roughly  28-35 normal creeps or 40-45 wood creeps.

Explain me how you defined these numbers? Or is it just a guess for you? Algorithms are always based on facts, not on feelings. Having the creeps count 20 times less than a kill is by far a lot. Not only from point of calculating these and sum it up, also by seeing the in game point.
 

Usually to kill a player you need to hunt him and spent time when you cannot farm correctly. This does make you loose around 10 creeps on a good, and down to 5, maybe, on a bad farmer.
Normally this should be the rate to the kill, so one kill should actually count like 10 creeps. You see that i have calibrated it a lot down, going further down is simply wrong in my opinion.

Maybe its caused by EM, it appears that no one is aware of how valueable a high amount of creeps is.
Not meant offensive here ;) Just my questioning from what I've read.

  • Upvote 1

Share this post


Link to post
Share on other sites
4 hours ago, GriefCode said:

An algorithm is never about feelings and ingame actions, thats why it is an algorithm. It just gets a resultset which is used to calculate a players efficiency.

 

Explain me how you defined these numbers? Or is it just a guess for you? Algorithms are always based on facts, not on feelings. Having the creeps counts 20 times less than a kill is by far a lot. Not only from point of calculating these and sup it up, also by seeing the in game point.
 

Usually to kill a player you need to hunt him and spent time when you cannot farm correctly. This does make you loose around 10 creeps on a good, and down to 5, maybe, on a bad farmer.
Normally this should be the rate to the kill, so one kill should actually count like 10 creeps. You see that i have calibrated it a lot down, going further down is simply wrong in my opinion.

Maybe its caused by EM, it appears that no one is aware of how valueable a high amount of creeps is.

and there-in lies the problems, in em, the heroes are treated as the creep =]]]]

Share this post


Link to post
Share on other sites
On 28.11.2017 at 1:03 PM, caphalor. said:

https://ohstats.net/en/game/62560  .. fits to the topic

its true

https://ohstats.net/en/game/83723

LC has 1 kill, less assists, same amout of towers and the same amout of creeps than i have, just with neutrals and not normal creeps.

still gets more elo :D

cant understand this^^

@GriefCode should it really be like this? just wondering.

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

  • Recently Browsing   0 members

    No registered users viewing this page.

  • Who's Online   0 Members, 0 Anonymous, 3 Guests (See full list)

    There are no registered users currently online

  • Forum Statistics

    1007
    Total Topics
    6749
    Total Posts
  • Member Statistics

    1108
    Total Members
    138
    Most Online
    Newest Member
    Dustmuz
    Joined
  • Similar Content

    • By seofl
      Game-link:https://ohstats.net/en/game/90239
      Offending player(s) lobby-number or color or ign: Kukhode (pink)
      Time occured: ~21 minutes replay time, ~19 minutes in game time
      Details(optional): Fed my courier to the enemy team, tried to kill my items (I was Gyro). Proceeds to AFK farm the rest of the game without helping the team.
    • By marcipan032
      Game ruinned, suicide etc...game name: 8lllD
      game: https://ohstats.net/en/game/174597
    • By FatmanKROUM
      I would really like it if draws were counted as draws and not as losses.
      If a game draws, you will get a loss in your stats and draws stays at 0.
      This draw/loss will also negatively impact your overall stats (for those who care) i.e. your kills/deaths/assists per game will suffer if for example a game is auto-drawn at the beginning due to 2 people leaving the game.
       
      Tl;dr: Make draws count as draws and draws only and do not record the k/d/a of that game to your stats < would be my suggestion
      this would make the stats of each player more accurate as more draws currently mean more losses & "false stats"
       
      as always, just an idea
      cheers.
  • Members

    • kakapo 91

      Moderator
      Joined:
      Last active:
×

Important Information

We have placed cookies on your device to help make this website better. You can adjust your cookie settings, otherwise we'll assume you're okay to continue.