General Discussion

General DiscussionTalent choice win rates vs overall win rates

Talent choice win rates vs overall win rates in General Discussion
Dire Wolf

    This doesn't make sense to me. I can't get exactly since 7.0 and introduction of talents but troll warlords win rate last 6 months is 53%. That's almost when 7.0 dropped with talents.

    Looking at his talent choice win rates, picking 7 strength is 49.6% win rate. Every single other choice is over his average win rate. Some by a lot, just next level move speed is around 56, armor is around 55. So how can he have an overall win rate that much less than all his talent choices?

    On the one hand you can say well when troll gets to level 25 he should have an average win rate around 65%. That would explain why his overall win rate is only 53%, cus many games where he doesn't get to 25.

    But if you look at his talent choices at level 10, his expected win rate should be the weighted average of the talent choices. How do we find the averages of the averages? Easy, just treat like numbers, 21.5 x 49.6 + 78.5 x 54.8 = 5368.2 / 100 = 53.682% chance to win once you hit level 10. Difference between 53.68 and 53.08 is 0.6%.

    So that would mean to drag his win rate down a bunch of his losses must be without hitting level 10. This to me seems crazy. Games would have to be under 20 minutes for this to happen cus any core is hitting 10 by 20 mins at least.

    Now the math escapes me, I'm not sure how to figure out what percentage of games would have to be losses before level 10 to drag that win rate down 0.6% total or what his winrate is pre level 10. But it seems like it'd be a lot of games to drag it that low.

    Does this seem reasonable or do the win rate numbers flat out not match?

    Win rates on talents always seem much higher than overall win rates, so much so it seems off. Though someone with better math than me could possibly prove it.

    This topic was edited
    cartel

      :thinking: :thinking:

      Jacked

        whats the issue? because talent winrates are higher than 53%? i dont get it

        anyway good to see you're still passionate about troll. i kinda got bored of the hero, and he seems to have fallen off just a little bit

        This comment was edited
        one syllable anglo-saxon

          probably soemthing wrong with their data gathering

          look at that: https://www.dotabuff.com/heroes/troll-warlord/trends

          troll has above his average winrate for every level he ends the match with, and so do all heroes
          no matter the pool they get this data from, u cant have positive avg winrate on all heroes, it's the problem with their algo or smth

          one syllable anglo-saxon

            actually this deserves a separate thread i believe

            주 롄양

              dude talent winrate means:
              if x=talent a winrate

              then its actual winrate is x*(troll warlord overall winrate)

              is dis kind of math even dat hard? :thinking:

              TheMaverick427

                Its also possible that people skip his level 10 talent and skill his abilities instead, so there will be games where he's above level 10 but hasn't got the talents.

                Dire Wolf

                  You guys don't understand what I am saying. If troll by picking either talent at level 10 wins 53.68% of his matches, but his overall win rate in last 6 months (as close to since talents were introduced as we can get), is 53.08% that means that there's some winrate he has when he doesn't hit level 10 that is lower. The thing is it's impossible to tell what it is, because we don't know how many games were played. It could be 0% win rate but with very few games played, or it could be like 50% with tons of games played, enough to drag the overall win rate down 0.6%.

                  That's the math I don't have time to do right now. However it seems questionable because the amount of games where a core hero doesn't hit level 10 has to be an insanely low percentage. I couldn't find a single one on troll in my 100+ matches. only 2%, 50 matches out of my total have been less than 20 mins, and only like 2 since patch 7.0. I think it's safe to assume any match where a core hero doesn't hit level 10 is going to be under 20 mins.

                  However examining the numbers maybe it is possible. Let's just assume he loses every single match where he doesn't hit 10 and it's 2% of all troll matches (which seems high to me but whatever). That would make games past level 10 98% of matches, and we know he wins 53.68% past level 10. Losing all matches under level 10 and having that be 2% would drop his win rate to 52.6%. Math is 53.68 * 98 / 100. So yes that would be enough.

                  But that's a lot of assumptions, that really in 2% of game he doesn't hit level 10 and loses every single one of them. So let's say he only plays 1% of matches where he doesn't hit 10 and he wins 20% of them. Now you get 53.14%. That's pretty close to his actual win rate.

                  So maybe it is explainable.

                  Dire Wolf

                    The thing Alenari posted is obviously flawed data since yeah it's impossible for every level to outpace overall win rate. Also 4% of troll matches end at level 1? 4% end at level 2? That graph can't be right. Unless shit tons of people abandon games early or something.

                    >tfw no gamer girl peeing gf

                      You need to relax, OP :axe_laugh:

                      Luxon

                        Like half of the matches end before Troll hits 15.

                        Dire Wolf

                          That I can believe, I'll bet in tons of normal games and low priority etc people abandon once behind and games end super early. But I have a very hard time believing that 13.75% of troll games end when he is level 3 or lower.