ESOCommunity

Thanks

No one to play with on ESO

EAGLEMUT wrote:

zoom wrote:

Show hidden quotes

n0el wrote:

zoom wrote:Please link me the seeding system rules.

Top 16 from last tourney. Rest are filled by ELO.

Next event we will improve it.

Thanks. Are the rules published anywhere?

It doesn't seem too bad, but you should never have ignored TWC completely. Do like 50% of placements there, or something.

The technical implementation currently only allows top16+Elo, or do all seeds manually. And honestly, I'm glad we even got that to work properly
Hopefully we'll find some time to improve the process.

Manual seeding should not be that much work. When I proposed my system a couple of days ago and presented what the outcome would look like it didn't even take me 5 mins. I would volunteer to do it for the next tournament with any system you come up with

I vote alligator for tourney rule committee.

deleted_user wrote:I vote alligator for tourney rule committee.

who is that alligator guy supposed to be? and why did he choose a name so similar to mine?

aligator92 wrote:

deleted_user wrote:I vote alligator for tourney rule committee.

who is that alligator guy supposed to be? and why did he choose a name so similar to mine?

that would be me

zoom wrote:

n0el wrote:

zoom wrote:Please link me the seeding system rules.

Top 16 from last tourney. Rest are filled by ELO.

Next event we will improve it.

Thanks. Are the rules published anywhere?

viewtopic.php?f=399&t=12517#p264225

TWC wasn't weighted because it was a minor tournament. We only use results from major seasonal tournaments to determine preseeds. I don't think this will change anytime soon since minor tournaments often have different rules. In this case, I wouldn't say winning the TWC tournament takes the same amount of skill as winning a seasonal tournament, and awarding half seeds will get very messy and obscure.

alligator_guy wrote:

aligator92 wrote:

deleted_user wrote:I vote alligator for tourney rule committee.

who is that alligator guy supposed to be? and why did he choose a name so similar to mine?

that would be me

What a piece of shit!

Mr_Bramboy wrote:

zoom wrote:

Show hidden quotes

n0el wrote:

zoom wrote:Please link me the seeding system rules.

Top 16 from last tourney. Rest are filled by ELO.

Next event we will improve it.

Thanks. Are the rules published anywhere?

viewtopic.php?f=399&t=12517#p264225

TWC wasn't weighted because it was a minor tournament. We only use results from major seasonal tournaments to determine preseeds. I don't think this will change anytime soon since minor tournaments often have different rules. In this case, I wouldn't say winning the TWC tournament takes the same amount of skill as winning a seasonal tournament, and awarding half seeds will get very messy and obscure.

I know it was. That's why I suggested counting it as a lighter weight. What's messy about just counting the winner of TWC as runner-up or semi-finalist?

In the end seeds from previous tournament fuck the actual seeding process really.

Jerom wrote:In the end seeds from previous tournament fuck the actual seeding process really.

The are the only reasonable measure for seeding. You cannot go purely by ElO because old accounts that still have 2700 ELO would always get the top seeds over active players. And you cannot just go by feeling because everyone will have different opinions and there would be much drama by players who feel underestimated.
We should simply start to incorporate all major ESOC Tournies of the past and just the most recent one.

aligator92 wrote:

Jerom wrote:In the end seeds from previous tournament fuck the actual seeding process really.

The are the only reasonable measure for seeding. You cannot go purely by ElO because old accounts that still have 2700 ELO would always get the top seeds over active players. And you cannot just go by feeling because everyone will have different opinions and there would be much drama by players who feel underestimated.
We should simply start to incorporate all major ESOC Tournies of the past and just the most recent one.

The elo system kinda ensures though that the top elo players will be in the region of the top seeds I believe.

About the seeding, I think we need to use a formula weighing the last tourney results (for example the 3 last tourney results, with bigger coefficients for the more recent tourneys) as well as the ELO. It's important that the formula is continuous, because as we can see with the current seeding system, there is a problem between seeds 16 and 17, as the first 16 and the last 16 seeds are not determined by the same method (seed 17 was usually a much better played than seed 16, which is not logical). In other words, imo, it needs to be a unique formula to seed everyone. For example:

Rating = (ELO - 2000)/5 + 100/R1 + 50/R2 + 20/R3

Where R1, R2 and R3 are the rank of the player (so 1 if he was first, 2 if he was second, and so on) in the three previous tourneys, R1 being the most recent and R3 the least. I substracted 2000 to ELO because I considered that all players that would deserve a seed would have more than 2000 ELO, and only the part over 2000 is relevant to compare.
Of course the numbers can (and probably should) be tweaked, depending on how important we want the ELO to be for example. There's still a big problem in this formula: what if the player didn't play in the previous tourney(s)? I didn't really find an answer to that, but I think there must be a way to give a fair compensation in the rating to someone who wouldn't play in a past tourney.

Kaiserklein wrote:About the seeding, I think we need to use a formula weighing the last tourney results (for example the 3 last tourney results, with bigger coefficients for the more recent tourneys) as well as the ELO. It's important that the formula is continuous, because as we can see with the current seeding system, there is a problem between seeds 16 and 17, as the first 16 and the last 16 seeds are not determined by the same method (seed 17 was usually a much better played than seed 16, which is not logical). In other words, imo, it needs to be a unique formula to seed everyone. For example:

Rating = (ELO - 2000)/5 + 100/R1 + 50/R2 + 20/R3

Where R1, R2 and R3 are the rank of the player (so 1 if he was first, 2 if he was second, and so on) in the three previous tourneys, R1 being the most recent and R3 the least. I substracted 2000 to ELO because I considered that all players that would deserve a seed would have more than 2000 ELO, and only the part over 2000 is relevant to compare.
Of course the numbers can (and probably should) be tweaked, depending on how important we want the ELO to be for example. There's still a big problem in this formula: what if the player didn't play in the previous tourney(s)? I didn't really find an answer to that, but I think there must be a way to give a fair compensation in the rating to someone who wouldn't play in a past tourney.

r u a math wizard?

surprised anyone had bothered to read it tbh

region #7 easiest region

Hazza54321 wrote:surprised anyone had bothered to read it tbh

Sorry that I'm actually posting some content, maybe I should post only 1 sentence everytime like you

Zutazuta wrote:

Kaiserklein wrote:About the seeding, I think we need to use a formula weighing the last tourney results (for example the 3 last tourney results, with bigger coefficients for the more recent tourneys) as well as the ELO. It's important that the formula is continuous, because as we can see with the current seeding system, there is a problem between seeds 16 and 17, as the first 16 and the last 16 seeds are not determined by the same method (seed 17 was usually a much better played than seed 16, which is not logical). In other words, imo, it needs to be a unique formula to seed everyone. For example:

Rating = (ELO - 2000)/5 + 100/R1 + 50/R2 + 20/R3

Where R1, R2 and R3 are the rank of the player (so 1 if he was first, 2 if he was second, and so on) in the three previous tourneys, R1 being the most recent and R3 the least. I substracted 2000 to ELO because I considered that all players that would deserve a seed would have more than 2000 ELO, and only the part over 2000 is relevant to compare.
Of course the numbers can (and probably should) be tweaked, depending on how important we want the ELO to be for example. There's still a big problem in this formula: what if the player didn't play in the previous tourney(s)? I didn't really find an answer to that, but I think there must be a way to give a fair compensation in the rating to someone who wouldn't play in a past tourney.

r u a math wizard?

no i just big nob sry

For example with this formula, I calculated the rating of a few players who played in the 3 last tourneys:

- Lordraphael: Rating = (2625 - 2000)/5 + 100/2 + 50/1 + 20/2 = 235
- Kaiserklein: Rating = (2524 - 2000)/5 + 100/3 + 50/12 + 20/24 = 143.13
- Mitoe: Rating = (2466 - 2000)/5 + 100/4 + 50/12 + 20/6 = 125.7
- mongo10: Rating = (2354 - 2000)/5 + 100/12 + 50/3 + 20/6 = 99.13
- SomppuKunkku: Rating = (2354 - 2000)/5 + 100/12 + 50/6 + 20/48 = 87.88
- GoodSpeed: Rating = (2348 - 2000)/5 + 100/12 + 50/24 + 20/16 = 81.27
- dicktator_: Rating = (2351 - 2000)/5 + 100/24 + 50/24 + 20/48 = 76.87

I gave to the coefficient the following values: 6 for RO8, 12 for RO16, 24 for RO32, 48 for RO64, as an average score for being in the said rounds, since there's no way to have an accurate rank except for the first 4 players.
Now in the case of H2O_: he didn't play in last tourney, so I think we can calculate in this way:

H2O_: Rating = (2747 - 2000)/5 + 80/2 + 40/1 + 15/2 = 236.9

I used the 3 most recent tourneys in which he played, and applied the coefficients 80, 40 and 15 instead of 100, 50 and 20, as a penalty because of those tourneys being longer ago. Again, the numbers can of course be tweaked.
In the case of BlackStar_OP: he played in the last tourney and in an old tourney, so we can apply this kind of coefficient:

BlackStar_OP: Rating = (2769 - 2000)/5 + 100/1 + 15/1 = 268.8


I think calculating in this kind of way would be more accurate than blindly taking ELO or last tourney rank (or even worse, a mix of both).

seems accurate

Kaiserklein wrote:

Hazza54321 wrote:surprised anyone had bothered to read it tbh

Sorry that I'm actually posting some content, maybe I should post only 1 sentence everytime like you

It wasnt a shot at you, i didnt think youd be french about a joke

@Kaiserklein the formula looks good, and the decay for the rating of older tournaments seems fine.

Kaiserklein wrote:For example with this formula, I calculated the rating of a few players who played in the 3 last tourneys:

- Lordraphael: Rating = (2625 - 2000)/5 + 100/2 + 50/1 + 20/2 = 235
- Kaiserklein: Rating = (2524 - 2000)/5 + 100/3 + 50/12 + 20/24 = 143.13
- Mitoe: Rating = (2466 - 2000)/5 + 100/4 + 50/12 + 20/6 = 125.7
- mongo10: Rating = (2354 - 2000)/5 + 100/12 + 50/3 + 20/6 = 99.13
- SomppuKunkku: Rating = (2354 - 2000)/5 + 100/12 + 50/6 + 20/48 = 87.88
- GoodSpeed: Rating = (2348 - 2000)/5 + 100/12 + 50/24 + 20/16 = 81.27
- dicktator_: Rating = (2351 - 2000)/5 + 100/24 + 50/24 + 20/48 = 76.87

I gave to the coefficient the following values: 6 for RO8, 12 for RO16, 24 for RO32, 48 for RO64, as an average score for being in the said rounds, since there's no way to have an accurate rank except for the first 4 players.
Now in the case of H2O_: he didn't play in last tourney, so I think we can calculate in this way:

H2O_: Rating = (2747 - 2000)/5 + 80/2 + 40/1 + 15/2 = 236.9

I used the 3 most recent tourneys in which he played, and applied the coefficients 80, 40 and 15 instead of 100, 50 and 20, as a penalty because of those tourneys being longer ago. Again, the numbers can of course be tweaked.
In the case of BlackStar_OP: he played in the last tourney and in an old tourney, so we can apply this kind of coefficient:

BlackStar_OP: Rating = (2769 - 2000)/5 + 100/1 + 15/1 = 268.8


I think calculating in this kind of way would be more accurate than blindly taking ELO or last tourney rank (or even worse, a mix of both).

Why do you calculate some players according to their max elo and some by current one

Zutazuta wrote:

Kaiserklein wrote:About the seeding, I think we need to use a formula weighing the last tourney results (for example the 3 last tourney results, with bigger coefficients for the more recent tourneys) as well as the ELO. It's important that the formula is continuous, because as we can see with the current seeding system, there is a problem between seeds 16 and 17, as the first 16 and the last 16 seeds are not determined by the same method (seed 17 was usually a much better played than seed 16, which is not logical). In other words, imo, it needs to be a unique formula to seed everyone. For example:

Rating = (ELO - 2000)/5 + 100/R1 + 50/R2 + 20/R3

Where R1, R2 and R3 are the rank of the player (so 1 if he was first, 2 if he was second, and so on) in the three previous tourneys, R1 being the most recent and R3 the least. I substracted 2000 to ELO because I considered that all players that would deserve a seed would have more than 2000 ELO, and only the part over 2000 is relevant to compare.
Of course the numbers can (and probably should) be tweaked, depending on how important we want the ELO to be for example. There's still a big problem in this formula: what if the player didn't play in the previous tourney(s)? I didn't really find an answer to that, but I think there must be a way to give a fair compensation in the rating to someone who wouldn't play in a past tourney.

r u a math wizard?

No, but I am and I approve of this message. You can tell by Kaiser's use of the word coefficient and his organization that he has some experience with mathematical modeling and/or statistics/data processing.