General Discussion

General DiscussionHero Match-Up Win Rate Percentages Do Not Add Up

Hero Match-Up Win Rate Percentages Do Not Add Up in General Discussion
Aaron the Intern

    Hi all,

    Perhaps someone could enlighten me on this topic. I noticed the match-up win rate percentages between any two pair of heroes do not add up to 100%. Why is this and why does this happen?

    Pandamonium(You Died)

      Why would wirnate between any two pairs of hero add up to 100? They are not connected in any way.

      Aaron the Intern

        Sorry for being unclear.

        This is the match up win rate between two heroes; not the general win rate of each hero. Take alchemist and ancient apparition:

        alchemist match up win rate against ancient apparition = 48.83%
        ancient apparition match up win rate against alchemist = 49.40%

        I would think the sum of these two values would equal 100% (for this example, the sum is 98.23%)?

        Este comentario fue editado
        6_din_49

          Maybe they aren't calculated all at once?

          Probably is the same thing as it's with friend games: it shows you have X games with a friend, and when you look at friend's profile, you see he had a different number of games with you.

          Este comentario fue editado
          Twinkle

            Might be that unrecorded games aren't included? So 1.77% of the games were no stats recorded games.

            Or, something horribly wrong is going on...

            Este comentario fue editado
            Villain

              Perhaps it's because of the opposing teams, for example, let's say they are against Drow Ranger.

              Alchemist win rate against AA (Drow is with AA in this game) = 48.83%
              AA win rate against Alch (Drow is with Alch in this game) = 49.40%

              Drow win rate against both AA and Alch = over 50%

              The way to test this would be to see if any hero match ups have OVER 100%.

              Este comentario fue editado
              Weeb

                dont take too much with dotabuff stats, its completely unreliable, the last 3 days, my friend got 2 more wins, where someone abandoned early, and everyone leaves, one stays, and him and his team r made the winners on dotabuff, some many data here would be wrong if u looking for something accurate such as that

                6_din_49

                  Stats aren't 100% accurate, but are good enough to make an idea what hero is good against what.

                  Relentless

                    You really want to look at the differential advantage number anyway. They are not going to add up to 100% because the difference will depend on the relative performance of the hero compared to other heroes you could have picked.

                    The small +/- % number shows the advantage impact.

                    Let's consider as an example Broodmother vs Alchemist. For the past month alchemist has an overall win rate of 45.59% on 1.023 Million games. While Brood has 38.92% win on 0.546 Million games.

                    Alchemist played brood in 23,573 games, While Brood played Alchemist in 23,572 games at the time I opened the pages for these stats. So the games clearly do line up.

                    When Alchemist plays against Brood his odds improve to 56.19%, but the Advantage is -0.4%
                    When Broodmother plays against Alch his odds improve to 41.85%, but the Advantage is -1.69%
                    About 400 games did not count due to early abandon.

                    This shows that while Broods win rate is higher against Alchemist then a simple average of all games, Brood still is a worse choice against Alchemist than the baseline "other hero" you might have picked instead.

                    Also, while Alchemist is far more likely to win against brood than a simple average of all games...Alchemist is really slightly worse against brood than the average "other hero" you might pick.

                    Este comentario fue editado