Open AI Championship

Round 1

Project

Sandbox

Round 1

Round 2

Finals

Recent comments

19 february 23:14: karloid wrote comment on post The closing of the Russian AI Cup 2018
19 february 13:38: tyamgin wrote comment on post The closing of the Russian AI Cup 2018
17 february 14:58: tsapelman wrote comment on post The closing of the Russian AI Cup 2018
13 february 09:15: Hal wrote comment on post The closing of the Russian AI Cup 2018
12 february 10:48: AlexKol wrote comment on post The closing of the Russian AI Cup 2018
12 february 10:48: AlexKol wrote comment on post The closing of the Russian AI Cup 2018
12 february 10:12: bearf wrote comment on post The closing of the Russian AI Cup 2018
11 february 17:54: tsapelman wrote comment on post The closing of the Russian AI Cup 2018
10 february 11:36: Scarlet wrote comment on post Finals: Results

VK Group

Telegram chat

 

Round 1 will be held starting on January 11, 2018 at 21:00 UTC and will finish on January 12, 2018 at 21:00 UTC. Only 1080 best strategies will be participating in it. The format of the games is 2x2. This stage will consist of 2 parts, and there will be a break between them (Sandbox will be available during it), that will allow participants to improve their strategies. Each part will last 12 hours, and a break — 24 hours.

In each part of Round 1 only last correct strategy, submitted before the corresponding part, will be used for the games. Games will appear in waves. In each wave each strategy will take part in exactly one game. Number of waves will be not less than 10.

Each game participants will be chosen randomly and equiprobably (in contrast to the Sandbox, where it is much more likely to meet an equal oppopent). The final results table of Round 1 will be formed by using sums of scores, gained throughout all the games, and 300 best strategies (in case of equal score, participant that submitted his strategy earlier is in priority) will pass to Round 2. The initial rating value of each participant is equal to 0, and each game he will be given a specified number of points:

  • 1 place — 2 points,
  • 2 place — 0 points.


dbdr

dbdr

Budapest

I think the random selection of opponents is creating quite a lot of unfairness in the round results. With 42 games at half round (so I expect 84 in total), playing a few more or fewer times against top players affects a lot the final ranking. This could be very problematic for Round 2 too. (The finals are fine, since everyone plays against everyone, there is no randomness involved.)

Do others agree this can be a problem, and what are ideas to improve the situation?

It seems to me that the system used in Sanbox is more fair in this regard: play more often against people around your rank. This makes an underrated strategy naturally climb up, and and overrated one drop down, stabilizing closer to their real rank. Would it be possible to use this for Round 2?

Jan 12, 2019 2:38:15 PM
amurushkin

amurushkin

ХНУРЭ (бывш. ХИРЭ, ХТУРЭ)

dbdr

dbdr

Budapest

I think the random selection of opponents is creating quite a lot of unfairness in the round results. With 42 games at half round (so I expect 84 in total), playing a few more or fewer times against top players affects a lot the final ranking. This could be very problematic for Round 2 too. (The finals are fine, since everyone plays against everyone, there is no randomness involved.) …

does not they pair games with similar round 1 ranks? it should be fair enough i think. btw they run this system for many years already and usually everything was fine with that

Jan 12, 2019 3:07:21 PM
jMind

jMind

Moscow

it’s not a problem since 300 of 670 will pass to Round 2.

Jan 12, 2019 3:10:11 PM
sergio-dna

sergio-dna

Moscow

dbdr

dbdr

Budapest

I think the random selection of opponents is creating quite a lot of unfairness in the round results. With 42 games at half round (so I expect 84 in total), playing a few more or fewer times against top players affects a lot the final ranking. This could be very problematic for Round 2 too. (The finals are fine, since everyone plays against everyone, there is no randomness involved.) …

Absolutely agree. The fact that there are 300 participants selected does not alleviate the issue in any way: the ones in the Sandbox top-100 are relatively safe, but the ones at 200-300 ranks are very vulnerable to bad luck. The same is going to be true for 30-50 ranks during round 2.

Jan 12, 2019 3:38:47 PM
mixei4

mixei4

КГУ

sergio-dna

sergio-dna

Moscow

Absolutely agree. The fact that there are 300 participants selected does not alleviate the issue in any way: the ones in the Sandbox top-100 are relatively safe, but the ones at 200-300 ranks are very vulnerable to bad luck. The same is going to be true for 30-50 ranks during round 2.

Just don’t be out of top-100 before round 1 and out of top-30 before round 2.

Jan 12, 2019 9:02:55 PM
Poligraph

Poligraph

Ufa

mixei4

mixei4

КГУ

Just don’t be out of top-100 before round 1 and out of top-30 before round 2.

Thank you sir!

Jan 13, 2019 8:13:22 AM
amurushkin

amurushkin

ХНУРЭ (бывш. ХИРЭ, ХТУРЭ)

By the way quite good idea in rounds to run games according to the Swiss system. I think this will give a more correct location than in the randomness. May be worth it to try such a system may not this, so in the next competition.

Jan 13, 2019 3:40:18 PM
azt-yur

azt-yur

ОГУ

amurushkin

amurushkin

ХНУРЭ (бывш. ХИРЭ, ХТУРЭ)

By the way quite good idea in rounds to run games according to the Swiss system. I think this will give a more correct location than in the randomness. May be worth it to try such a system may not this, so in the next competition.

but do not forget that this is a game. The game should be the probability of accidents. I think all now is who really is able to win the competition that just released, and for all other excitement should be kept at least at the level of randomization.

Jan 13, 2019 6:35:58 PM
dbdr

dbdr

Budapest

azt-yur

azt-yur

ОГУ

but do not forget that this is a game. The game should be the probability of accidents. I think all now is who really is able to win the competition that just released, and for all other excitement should be kept at least at the level of randomization.

True, but the game in itself has quite a lot of randomness (just try playing a bot against itself or against a bot of similar level). Adding more randomness from the luck of playing weaker opponents is not needed.

Jan 13, 2019 6:58:08 PM
alkozel

alkozel

Калужский филиал ГУ СДА при Спецстрое России

azt-yur

azt-yur

ОГУ

but do not forget that this is a game. The game should be the probability of accidents. I think all now is who really is able to win the competition that just released, and for all other excitement should be kept at least at the level of randomization.

Of course you are right, but for 2 rounds of Swiss would be good. In the end 1-a circle with Rand, 2-Swiss, 3-all. All fun and fair.

Jan 13, 2019 7:00:47 PM
azt-yur

azt-yur

ОГУ

alkozel

alkozel

Калужский филиал ГУ СДА при Спецстрое России

Of course you are right, but for 2 rounds of Swiss would be good. In the end 1-a circle with Rand, 2-Swiss, 3-all. All fun and fair.

That’s how I participated tops always got to the finals. They will receive prizes honored in any system, and for other may remains a shred of hope. That’s why I like the system RAICа, and transoms after a round forces you to make some improvements and try to get on. In General, I for the current rule )))

Jan 13, 2019 7:19:58 PM