General Discussion

General DiscussionHero Talent Tier (Compared to the alternative talent)

Hero Talent Tier (Compared to the alternative talent) in General Discussion
Murranji

    Dotabuff.com hasn't done a Talent Tier article yet so here's my submission on talents where there should be an obvious choice to the alternative. Since 7.07 is coming out in a little over a week this will need to be updated when the data is available.

    Rather than a subjective list where people argue about "oh but its strong/weak in the context of their skills" etc I've only used dotabuff.com data to develop this. Ie talents where one talent is indicated to be clearly superior to another on a particular hero. Yes yes "winrate doesn't mean anything" (other than how often on average picking a certain talent results in a victory across the millions of games played each month and yes there's some situational exceptions eg wraith king against mana burn heroes but overall the percentages are a good indication of how choosing a certain talent is more likely to lead to victory).

    This list covers only those talents where the winrate is at least 4% or above than the alternative talent so not all heroes are listed.

    Tier 1 - >7% winrate than alternative talent

    Chaos Knight: 12 all stats - 7.6% over +120 gpm.
    Invoker: 1 forge spirit summoned - 8.7% winrate over 30%xp.
    Wisp: 10 mana regen - 8.1% over 10 strength and 20 health regen over 120gpm.
    Kotl: 300 chakra magic mana - 7.6% over 20%xp.
    Lich: Attacks apply 30% ms and as slow - 8.5% over 35 frost armour for buildings.
    Skywrath Mage: 14 mana regen - 7.1% over -4 second ancient seal cooldown.

    Tier 2 - >5% winrate than alternative talent

    Antimage: both 10 to all stats - 6.6% over 15% evasion and 25 agility - 6.2% over -50 mana void cooldown.
    Arc Warden: 30 damage - 5.6% over 200 health.
    Axe: 3 mana regen - 5.2% over 6 strength.
    Bloodseeker: 10 all stats - 5% over 14% rupture damage.
    Chen: +1000 min creep health - 5.6% over 120 gpm.
    Clinkz: 10 all stats - 5.3% over 20% evasion.
    Lina: 35/3% fiery soul stacks - 5.3% over -4 Dragon Slave cooldown.
    Monkey King: 40 damage - 5.3% over 20% magic resistance.
    Morphling: 400 waveform range - 6.4% over 50% extra replicate damage.
    Omniknight: 6 mana regen - 6.4% over 100 right click damage.
    Storm Spirit: 40 attack speed - 5.5% over 8 armour.
    Terrorblade: 15 all stats - 5.8% over -30 second sunder cooldown.
    Tidehunter: 6 mana - 6% over 15 strength.
    Underlord: 50 health regen - 6.2% over +0.4 pit of malice root.
    Vengeful Spirit: 20% damage aura - 5.9% over magic missle pierces magic immunity.
    Visage: 120 familiar movespeed - 5.1% over 20% spell amp.

    Tier 3 - >4% winrate than alternative talent

    Abaddon: 300 aphotic shield health - 4.4% over 25 strength.
    Ancient Apparition: +30 health regen - 4.7% over -1 ice vortex cooldown.
    Arc Warden: 10% cooldown reduction - 4.6% over 100 attack range.
    Beastmaster: 1 extra boar summoned - 4.1% over 12 strength.
    Brewmaster: 140 attack speed - 4.1% over 2000 extra brewling hp.
    Broodmother: 25% xp gain - 4.2% over 60 spawn spiderling spell damage.
    Clinkz: 30 searing arrows damage - 4.9% over 15 strength.
    Doombringer: 40 health regen - 4.2% over 2% extra infernal blade damage.
    Dragon Knight: 40 damage - 4.6% over 40% exp.
    Drow Ranger: 25 marksmanship agility - 4.3% over 400 extra gust knockback distance.
    Faceless Void: 600 timewalk range - 4.4% over 20% evasion.
    Kunkka: 40 torrent damage - 4.5% over 30 right click damage.
    Lifestealer: 25 movespeed - 4.6% over 15% evasion.
    Luna: both 25 attack speed - 4.4% over 15% magic resistance and 15 all stats - 4.2% over -2 second lucent beam cooldown.
    Puck: 75% extra illusory orb speed - 4.8% over 420gpm.
    Queen of Pain: 60% spell lifesteal - 4.9% over 550 aoe shadow strike.
    Rikimaru: 10 agility - 4.2% over 30%xp.
    Sand King: 50 health regen - 4.7% over 4 extra epicentre pulses.
    Shadow Shaman: 4 extra wards summoned - 4.2% over 3 second extra shackles.
    Slark: 15 agility - 4.4% over 15 strength.
    Techies: 2 mana regen - 4.2% over 20 movespeed.
    Templar Assassin: 40 damage - 4.9% over 275 health.
    Troll Warlord: both 10 agility - 4.2% over 7 strength and 40 damage - 4.7% over 350 health.
    Viper: 16 agility - 4.1% over 15 strength.
    Visage: 30%exp - 4.1% over 90gpm.
    Weaver: 15 agility - 4.3% over 200 health.
    Winter Wyvern: 15% arctic burn slow - 4.7% over -3 splinter blast cooldown.
    Witch Doctor: 20 voodoo restoration healh - 4.6 over 175 additional death ward attack range.
    Wraith King: Reincarnation casts wraithfire blast - 4.2% over no reincarnation mana cost.

    Special Tier
    Meepo has a talent tree which should only be built in one way: 15 damage instead of 4 armour, 15% lifesteal instead of 25 movespeed, 25 attack speed instead of 10% evasion, and 400 health instead of -3 second poof cooldown.

    Tinker has a talent tree which should only be built in one way: 8 intelligence instead of 6 armour, 4% spell amp instead of 225 health, 75 extra cast range instead of 15% magic resist, and 100 laser damage instead of 20% spell lifesteal.

    Riguma Borusu

      The problem with this is that you cannot blindly use the statistical data like this without correlating things further.

      For example, an invoker is probably not more likely to win if he opts for the 2nd forge spirit, but you can be sure as hell that if he feels he is getting enough exp, he is going to go greedy and get the 2nd spirit instead of a faster lvl 25 (% xp talent). Likewise, if he feels really pressured to perform well despite being rekt in the laning stage, he's going to go for the exp talent to try to catch up.

      GPM talents are often catchup talents, as well. Every sort of a 'catchup' or 'defensive' talent is more likely to have a lower winrate because you always take them when you feel like you are likely to die or lose the game, and sometimes when you feel it's a good idea (like taking the HP talent as bs against necro, for example).

      Therefore, I'd say you should just completely disregarded the winrate data and try to analyze this in a more meaningful way - what does it mean for which talent to have a higher winrate, why it does so, not just that it does.

      Chi.

        ^ I agree, you can also see it in Wraith King Level 25. You only take the no reincarnation mana cost if they have a lot of mana burners already who are already likely to beat you. Where AOE fireblast you can take in every other scenario.

        Xignum

          since most talents are situational anyway

          CORE BABY HOT

            You misunderstand causal relationship between talents and winrates.Many of those “superior” talents have higher winrates only because they are useful in the situation when you have advantage.Let me explain:when you rape enemy team as lifestealer(thats the cause) you will pick up ms talent to continue chasing and beating shit out of them(thats the effect) . However, in hard, even in equal games evasion is preferable.So in many cases wining the game influences the talent choice.That’s why talent winrate is meaningless

            Na`Vi.Miracle-

              ^Some of them really are just straight up better though

              Riguma Borusu

                ^yes but we are not going to derive that information from winrates alone

                Dire Wolf

                  I think many of them you can tell just eyeballing and the win rates back it up. Like most gpm ones are crap, most exp ones are crap.

                  Like CK 12 stats is amazing vs 120 gpm. Obviously you're not going to carry around a casual ult orb but let's just pretend that's like the 12 stats equivalent, it's worth like 2100g. It'll take 17.5 minutes to get 2100g from 120 gpm. And it's his level 20 talent. Consider the highest xp average besides meepo is AM at 614, it'll take almost 28 minutes to get to level 20 on average in a decent game. 17 more minutes is 45 mins. The game should be well over by then. That's a super long game.

                  Other stuff like stats/dmg over evasion is pretty obvious too cus evasion can just get countered later and becomes a waste of a pick.

                  ETd

                    Dotabuff had an article about this, or at least mentioned these multiple times before. The talents are skewed because of situation. For example, you are more likely to get a GPM talent when you are behind. Meaning you are getting this talent usually from a losing position. So the GPM talent isn't bad, but it's statistically made to seem so because people get them in losing positions, meaning its more likely to lose regardless. Whereas the alternative talent is gotten in assured victories, in comparison. One is "win more", the other is "lose regardless". Same goes for the other talents with minus winrate, not just GPM/XP gain

                    Dire Wolf

                      No it's still awful even if behind. Takes a long time to equal the amount of stats you get in my ck example. Other heroes are different but on ck and dk it's shitty.

                      We just released an updated hero and facet statistics page featuring more data and advanced filters. Oh, and it supports Turbo!