Page 1 of 1

Arena logic with win probability

Posted: 22 August 2022, 23:57
by NoelfromNY
It seems like arena only takes into account player rating and not actual historic matchup performance when calculating win percentage. If I've played somebody 40 times and only won 30% of the matches, why is my probability to win 66% because the person doesn't play much and has a lower rating? Doesn't seem very logical.

Re: Arena logic with win probability

Posted: 23 August 2022, 09:49
by Tisaac
NoelfromNY wrote: 22 August 2022, 23:57 It seems like arena only takes into account player rating and not actual historic matchup performance when calculating win percentage. If I've played somebody 40 times and only won 30% of the matches, why is my probability to win 66% because the person doesn't play much and has a lower rating? Doesn't seem very logical.
Why ? Sounds very logical to me to use ratings that should represent a player level over anything else. Maybe you just add luck in these 40 matches against him before but he is stronger than that in reality. History against someone else dont weight much, unless you really played a lot of games against him, but in this case this means that these matches would actually change the other ELO enough so that it's taken into account.

Re: Arena logic with win probability

Posted: 29 August 2022, 17:41
by NoelfromNY
30 cases are considered a significant level of case study for probability analysis. Luck is a significant factor in the short run of dice forge. But the more matches between two people the less of a factor that is. Euro games have much less luck, and ELO is calculated the same way.

Re: Arena logic with win probability

Posted: 04 September 2023, 06:48
by Ceaseless
NoelfromNY wrote: 22 August 2022, 23:57 It seems like arena only takes into account player rating and not actual historic matchup performance when calculating win percentage. If I've played somebody 40 times and only won 30% of the matches, why is my probability to win 66% because the person doesn't play much and has a lower rating? Doesn't seem very logical.
Why should past games with a specific player matter? That could have been back when one player was still learning, and has since substantially improved, for instance. If another player's rating is only lower because they played less, than that implies they are underrated, and should be gaining points as they play. The rating system is just still calculating the proper rating as it gets more information.