Tournament No.2 - construction thread

[MagicalHacker]

I read your discussion on IRC; too bad I couldn't be there.

Note that my tool makes some assumptions specific to the ruleset I designed: specifically, that the top TWO players of the second round get through. Now, if you wish to add the absolute rule that no two players ever face each other twice in the preliminaries, we can devise the following:

In each second-round game, each player in the game must have played in a different game in the first round.

Which means:

The number of players in a game must be smaller or equal to the number of games.

Or, mathematically:

Definitions:

P == number of players

G == number of games per round

F == number of players in the finals

(P / G) <= G

P <= G^2

round_up(sqrt(P)) <= G

The top TWO players of the second round get through. Therefore, 2 finalists per game.

2 * G = F

Therefore,

round_up(sqrt(P)) <= 0.5 * F

F >= 2 * round_up(sqrt(P))

The number of finalists is larger than or equal to twice the square root of the number of players, rounded up. So yes, it scales poorly to the lower regions. If you want to use it with less players, you either have to loosen up the rule that no two players should battle twice, or play small games in which only ONE player gets though.

[Cisz]

Thats not too big a problem imo. We can use a different method in case of a small tournament.

How about either playing 3 rounds of prelims and 1 final or a 4 round single elimim w/o prelims (2nd concept bad imo)?

What interests me most is this:

The method suggested by jolin012 seems to succesfully seed the 2nd round of the tournament w/o the organizers intervention. The best players of the event will either all be winners in the first round and get into a different game on the second, or they will happen to be in the same game initially, which leads to them beeing in different games on the second round too.

I doubt this will work all the time, because if it would work, it would be a very good system and I can't see a reason why I have never heard of it so far. It's too good to be true.

So far I can't see why nobody uses this, let's give it a try.

If you (jolin012 and MagicalHacker) are ok with it, we should post an understandable version of this to a public section of the forums, and proceed. If it really is defunct, maybe some player will point out then, or, at worst, we will know after a few tournaments.

- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -

How about this worst case scenario:

Since the best two players advance from the second round, problems will occur if 3 or more players that should be in the final end up in the same game of round 2. For this to happen, assuming they all do as well as they should in round 1, they would have to come from three different games. And for this to be possible, they would have to be seeded like this:

Game one - 1 strong player

Game two - 2 strong players

Game three - 3 strong players

And then it could happen that the second round would have a game like 1st of game one, 2nd of game two, 3rd of game three, with all three players beeing strong. Still we can assume that the best of them are more likely to proceed to the finals, we would just loose good players in the prelims.

- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -

And what about seeding for the finals? Am I right that we are clueless how to seed them? We could use the "combined ranking of the players they have beaten" as some sort of ranking maybe?

Like this maybe (22 players), listed are the placements in each game:

Round 1 

Game 1 
Player  8       5
Player 19       1
Player  5       2
Player  2       4
Player  7       3

Game 2 
Player  9       1
Player 22       3
Player 16       2
Player 20       4
Player  4       5

Game 3 
Player  3       3
Player 15       2
Player 11       4
Player 12       1

Game 4 
Player  6       1
Player 14       4
Player 17       3
Player 18       2

Game 5 
Player  1       3
Player 10       4
Player 21       2
Player 13       1





Round 2
                (Last
Game 1          round)
Player 19       (1)             1       Advance
Player 16       (2)             2       Advance
Player  3       (3)             4
Player 14       (4)             3

Game 2                  
Player  9       (1)             2       Advance
Player 15       (2)             1       Advance
Player 17       (3)             4
Player 10       (4)             5
Player  8       (5)             3

Game 3                  
Player 12       (1)             1       Advance
Player 18       (2)             2       Advance
Player  1       (3)             5
Player  2       (4)             3
Player  4       (5)             3

Game 4                  
Player  6       (1)             1       Advance
Player 21       (2)             2       Advance
Player  7       (3)             4
Player 20       (4)             3

Game 5                  
Player 13       (1)             2       Advance
Player  5       (2)             1       Advance
Player 22       (3)             3
Player 11       (4)             4

I assumed that players tend to repeat their result in the second round, if they are strong players. Player 5 and 15 happened to end up in a game with other strong players in the first round, and do better on the second. And the lower ranks are more likely to have random results, as they probably are dependant on what elements are rolled, and other stuff.

The results after this example preliminaries are:

Player          Places  Rank

Player  6       1/1     1       Advance
Player 12       1/1     1       Advance
Player 19       1/1     1       Advance

Player  5       2/1     4       Advance
Player  9       1/2     4       Advance
Player 13       1/2     4       Advance
Player 15       2/1     4       Advance

Player 16       2/2     8       Advance
Player 18       2/2     8       Advance
Player 21       2/2     8       Advance

Player 22       3/3     11

Player  2       4/3     12
Player  3       3/4     12
Player  7       3/4     12
Player 14       4/3     12
Player 20       4/3     12
Player 17       3/4     12

Player  1       3/5     18
Player  4       5/3     18
Player  8       5/3     18
Player 11       4/4     18

Now how do we seed the 10 players into two semifinal games? If i sum up the resulting ranks for the players they have beaten so far, I get this:

(work in progress - to be continued)

[MagicalHacker]

I doubt this will work all the time, because if it would work, it would be a very good system and I can't see a reason why I have never heard of it so far. It's too good to be true.

This may be due to the fact that eletd has a weird property: it allows both for one-on-one games, and for games with more players. Most standard tournament systems are designed for either of those game types, not for the combination.

Also, never assume something doesn't work because the rest of the world never thought about it. I found that to be a rather poor assumption

If you (jolin012 and MagicalHacker) are ok with it, we should post an understandable version of this to a public section of the forums, and proceed. If it really is defunct, maybe some player will point out then, or, at worst, we will know after a few tournaments.

Of course.

How about this worst case scenario:

Since the best two players advance from the second round, problems will occur if 3 or more players that should be in the final end up in the same game of round 2. For this to happen, assuming they all do as well as they should in round 1, they would have to come from three different games. And for this to be possible, they would have to be seeded like this:

Game one - 1 strong player

Game two - 2 strong players

Game three - 3 strong players

And then it could happen that the second round would have a game like 1st of game one, 2nd of game two, 3rd of game three, with all three players beeing strong. Still we can assume that the best of them are more likely to proceed to the finals, we would just loose good players in the prelims.

Well, my system isn't perfect. It's designed to be a good balance between fairness and player understandability. A system like yours may be somewhat fairer, but:

[*]Harder to explain to players - "You play round 1, we do the math, based on the math you end up in some round 2 game, we do more math, and you may get in the finals."[/*:m:31lczman]

[*]It scales poorly with a small number of rounds. Your system greatly depends on the number of preliminary rounds for fairness; it works wonders with 5 preliminary rounds, but poorly with only two. Mine gets slightly better with more rounds as well, but two rounds is good enough too (and arguably, so is one round, if you were desperate).[/*:m:31lczman]

Now how do we seed the 10 players into two semifinal games?

Random? Well, not completely random; my tool tries to keep the number of 2nd-round-1st-place players and 2nd-round-2nd-place players in a semifinal game equal.

[Cisz]

(continued)

Here are the advancing players of my example again, complete with the players they faced, if they won or lost against them, and what rank those players had after the first two rounds (the rank is given in ()):

Name            Places  Rank    Has beaten player no.                                   Has lost to

Player  6       1/1     1       14(12), 17(12), 18(8) and 7(12), 20(12), 21(8)          -
Player 12       1/1     1       3(12), 11(18), 15(4) and 1(18), 2(12), 4(21), 18(8)     -
Player 19       1/1     1       2(12), 5(4), 7(12), 8(18) and 3(12), 14(12), 16(8)      -

Player  5       2/1     4       2(12), 7(12), 8(18) and 11(18), 13(4), 22(11)           19(1)
Player  9       1/2     4       4(21), 16(8), 20(12), 22(11) and 8(18), 10(21), 17(12)  15(4)
Player 13       1/2     4       1(18), 10(21), 21(8) and 11(18), 22(11)                 5(4)
Player 15       2/1     4       3(12), 11(18) and 8(18), 9(4), 10(21), 17(12)           12(1)

Player 16       2/2     8       4(21), 20(12), 22(11) and 3(12), 14(12)                 9(4) and 19(1)
Player 18       2/2     8       14(12), 17(12) and 1(18), 2(12), 4(21)                  6(1) and 12(1)
Player 21       2/2     8       1(18), 10(21) and 3(12), 4(21)                          13(4) and 6(1)

And here I added up the ranks of the players they faced:

Name            Places  Rank    Combined rank of        Combined rank of players 
                                beaten players          he has lost to

Player  6       1/1     1       64 (6 Players)          -
Player 12       1/1     1       93 (7 Players)          -
Player 19       1/1     1       78 (7 Players)          -

Player  5       2/1     4       75 (6 Players)          1 (1 Player)
Player  9       1/2     4       103 (7 Players)         4 (1 Player)
Player 13       1/2     4       76 (5 Players)          4 (1 Player)
Player 15       2/1     4       85 (6 Players)          1 (1 Player)

Player 16       2/2     8       68 (5 Players)          5 (2 Players)
Player 18       2/2     8       75 (5 Players)          2 (2 Players)
Player 21       2/2     8       72 (4 Players)          5 (2 Players)

Apparently this has to take into account how many players they have beaten/lost to.

(to be continued)

[jolin012]

1st of all, you've added together the rankings of the players. more players beaten gives higher total rank beaten, aswell as more unskilled players beaten, that just don't stack.

I think it's rather obvios that the ranking you get from the 2nd round is higher priority than the 1st round. so even a player who did 4/1 should be ranked higher than someone who did 1/2.

Why?

Because the ranking of round 2 is the only one for trust, the players are randomized in round one, not seeded. round 1 result is like 50% skill 50% luck. maybe we could use round 1 result as a second priority(because it's better than randomization which is 100% luck) for a player who played 2/1 and one who did 1/1 to say 1/1 is better, if we really need to define that. however, i feel "ranking of total players beaten" is far out of range if we wanted to keep this rather simple, and it wont add so much accuracy to the seeding of the finals. if MagicalHacker proves "only counting 2nd round result" effective and simple enough i agree to him, "only count the second round result, if there are any players who did the same result, randomize who will get into which one of their slots."

randomize who will get into which slot means something like this, example of a 8 finalists semi:

Semi 1

someone who did 1st place in round 2

someone who did 2nd place in round 2

someone who did 1st place in round 2

someone who did 2nd place in round 2

Semi 2

someone who did 1st place in round 2

someone who did 2nd place in round 2

someone who did 1st place in round 2

someone who did 2nd place in roud 2

hmm in magicalhackers way we would randomize the winners of round 2 to (like the list shows) put 2 of them into each group and 2 of the 2nd placers in each group.

in cisz way, if he is convinced that round 2 ranking is most important, we would find out additional info( not with full accuracy) and try to rank all the players into different ranks, to find out which of the 1st and the 2nd are the best, not to put the 2 betst 1st and the 2 best 2nds into the same group.

now if that runs smooth i might agree to cisz, i still don't know what side i am on

maybe i shall start my own side which would be somewhere in the middle, using a) 2nd round rank and 1st round rank and c) randomize if there is something left(that probably is rather unimportant but has to be decided)

[Cisz]

1st of all, you've added together the rankings of the players. more players beaten gives higher total rank beaten, aswell as more unskilled players beaten, that just don't stack.

I know, working on that.

I think it's rather obvios that the ranking you get from the 2nd round is higher priority than the 1st round. so even a player who did 4/1 should be ranked higher than someone who did 1/2.

Why?

Because the ranking of round 2 is the only one for trust, the players are randomized in round one, not seeded. round 1 result is like 50% skill 50% luck.

If I beat a very good player in round one, that's the same effort as if I beat him in the second round.

[jolin012]

how do we know it is a very good player? if this is one of the first priorities in ranking? it could be a very bad player aswell.

in the 2nd round however there´should be one very strong one strong one weak and one very weak player, and all groups should be fair, a 1st place in round 2 should be about the ssame hard to get, nomatter what group you get into. however a 1sr place can be easy in one of teh groups in round 1 while it can be very hard in another group of round one.

[Cisz]

Here is a tournament invite list:

hyperprimate

good

cyberseer

jolin012

13est

MagicalHacker

Cisz

Kingzor

Those are people that are working on the organising or are likely to win or both. So we should try to fix a date that suits many of those. Did I forget someone?

[ocmassreo]

yes me ^^

[Karawasa]

Let me know when you have decided a date, I could post something about it on the site.

[jolin012]

Sure thing, and thanks for willing to post on site, as soon as something new happens this thread is probably where it's first posted.

[Karawasa]

This going to happen still?

[Cisz]

Yes. Semester is nearly over, planning restarts.

[Guest ma23]

12 games done in 4 rounds. bottom half play 2 games, top half plays 4 bye player plays 3

shouldn't be much complaint about ranking of top 8 players

round 1

r1g1 ----- r1g2 ---- r1g3 ---- r1g4

1,2,3,4 - 5,6,7,8 - 9,a,b,c - d,e,f,g

(this is also order of place. so 1,5,9,d win. 4,8,c,g last)

round 2 (just so everyone can play at least 2 games)

purely random scramble of all players except 4,8,c,g who all play in 1 game

r2g1----- r2g2 ----- r2g3 ---- r2g4

1,6,b,d - 3,7,a,f - 2,9,5,e - 4,8,c,g (for instance)

round 3

best of rd1 and 2 gets bye (this example 1 gets a bye)

1st and 2nd of r2g1,r2g3,r2g4 advance. 1st of r2g4 advance

--- r3g1---- r3g2

1 - 6,3,4 - 7,2,9

round 4

losers of rd3 play to determine ranking

2,4,3,9

winners of rd3 advance to final championship

1,7,6

end of tournament ranking: 1,7,6,2,4,3,9

sigh looks like ive been sucked into this thread