Pick This, Not This (6.66 Stats) *UPDATED Feb 4th*

[FBB.Fatal]

*Updated Feb 4th w/ Bans and Pick Rates and 7 more games!


Quick Observations:

• Slith/Tide/VS are permabanned or permapicked
• Storm is now 8-22, pls buff him. (26% win rate)
• QoP no bans, but promising stats. (10-4)
• BM didn't appear in 9 of 109 games at all, I guess some team's don't have good BM players.
• KotL is now 9-0 rofl..

-----------------------ORIGINAL POST BELOW--------------------------


Hey everyone I just compiled some stats on 109 competitive 6.66/6.66b replays from gosugamers. The results are a bit surprising. Take a look for yourself below.












Short Explanation for math nubs =D

• We first put in the raw data (wins/losses) then calculate the total games and win rate.
• Then we calculate the Z-values for the data.
• Then we normalize the Z-values (using a Z-chart or excel's normalization function)
• Then we construct a conversion table for overall strength.
• Finally we can determine the strength of heroes, as long as N>15. Strictly speaking we should use N>30 though, but N>15 gives us some good conclusions already.

Short Example using Potm

• Take PotM as an example, 33 wins, 25 losses, 58 games, 56% win rate
• Next we calculate the Z value for Potm, (0.56-.5) / Sqrt((.5*.5)/(58) = 1.05
• We then can normalize the z-value, which gives us 0.83
• A normalized z-value of 0.83 correlates to 'overpowered' using our chart.

Some quick observations:

• Tidehunter and Beastmaster have a 70% win rate
• ES/Admiral/Sven/Sandking are a bit below 50%
• PotM and VS are almost 60%
• SF is overrated
• Morph is a good carry if you need a balanced one
• CM is shit, Lock is shit, Storm is super shit
• Lion is ok, puck is ok
• Engima/Pugna win about 60% of the time
• Voljin has a 80% win rate (WTF?)
• Kotl has a 100% win rate (DOUBLE WTF? 8-0)

Some conclusions:

• Pick Tidehunter over ES/Admiral/Sandking
• BM is firstpick worthy
• Pick Morph/Sniper/Razor over SF if you wanna win.
• Pick PotM/VS before you pick ES/Admiral/Sven/Sandking
• Pick Lion over CM/Warlock
• For the love of God stop playing Storm pls...
• Engima/Punga are solid picks if you go for a push strategy
• Puck/Rhasta/DP/QoP/Lich all nice picks to finish off the draft
• Voljin and Kotl are imba lol....?

[Nyx Nyx Nyx]

Sample size too small for certain heroes, but still... KotL 8-0 hahahahaha nice!

Oh and you might wanna highlight QoP and Phantom Lancer's performance

[Meur]

yeah, the 3 surprises for me were QoP, WD, and KotL.

[xTimmTomx]

Originally Posted by A∴A∴

yeah, the 3 surprises for me were QoP, WD, and KotL.

played 14 times, 11 times and 8 times respectively. The sample sizes are just too small to infer anything from them unfortunately. Maybe some of the heroes with larger samples could be tested to see whether its statistically significant or not, dunno.


Also this doesnt take into account anything about what game mode was being played and what the opponents picks were. You cant build a team on the basis of "these heroes statistically are the best", what if the majority of the time Kotl was picked as a counter to a specific type of lineup? Could mean that great, he works as a good counter to that... but doesnt mean he has the capacity to be picked in all lineups.

[vvjacobo]

Originally Posted by FBB.Fatal

PotM and VS are almost 60%

Blasphemy, If you didn't read the forums last week, 70% of players said she is played for fun, not because she is imba.

Anyway too small sample, in need to be 20 times bigger to have a 80% of trust

[Baconnaise]

If these were taken from -cd games (like in some tourny's) the results are not that significant. You should probably only compile -cm games since -cd the hero pool is MUCH more situational... Tidehunter and BM imba. KotL has such a high win rate because he is almost always paired with phantom lancer btw...

[callme]

this is a good idea, bravo. One thing to note though even if it has already been said, you can't tell much from a small sample size for some of the observations, you would need at least 30+ games to even think of spotting a correlation in the data, in the case of witch doctor and kotl tou just don't know without more games.

[SirDumpling]

Give the kid some credit, all of you are crying for more games. It's still an impressive list!
Thanks for the time you put into this. Very cool!

[PeGBoY]

I am not a fan of win rates and nonsense like that as it only looks at a single hero in isolation. Also, your conclusions are hardly meaningful with such a small amount of data.

[FBB.Fatal]

Originally Posted by PeGBoY

I am not a fan of win rates and nonsense like that as it only looks at a single hero in isolation. Also, your conclusions are hardly meaningful with such a small amount of data.

Quite a few of the win rates above are statistically significant to alpha values of 90 or 95%. I just can't be bothered to calculate those yet lol..

I know there is a small amount of data, only 102 games, but that's because only 102 -CM games have been uploaded to gosugamers.

[EternityPala]

Well, you took your time off and did this.
Still, it was real sad to see Lanaya being played only once, on par with TB, and Techies.

[vvjacobo]

Originally Posted by FBB.Fatal

Quite a few of the win rates above are statistically significant to alpha values of 90 or 95%. I just can't be bothered to calculate those yet lol..

I know there is a small amount of data, only 102 games, but that's because only 102 -CM games have been uploaded to gosugamers.

You are right you can already have some conclusions in a 50% chance to win game with 25 or more games played.

[XaTrIxX]

I think 8-0 Kotl means something, even if not much. At least Kotl is a hero to consider in high level play. For many versions I prefer Kotl over Dazzle btw. Kotl is underrated in my opinion.

[Shadowplay]

You might want to look here if you want a more comprehensive stastic over the heroes: DL Hero Statistics - Single Instant Games - Dota-League Forum

There is more than this, but here is a snippet

Originally Posted by insanelf','index.php?page=Thread&postID=70374#post 70374

Most played heroes:
1 Priestess of the Moon: 1824
2 Sand King: 1208
3 Shadow Shaman: 1117
4 Crystal Maiden: 1098
5 Axe: 1052
6 Earthshaker: 1046
7 Admiral: 962
8 Clockwerk Goblin: 941
9 Lord of Olympia: 914
10 Demon Witch: 912

Winner Heroes:
1 Vengeful Spirit: 61.4%
2 Spectre: 57.1%
3 Lich: 56.5%
4 Tauren Chieftain: 56.4%
5 Venomancer: 56.2%
6 Dark Seer: 56.1%
7 Warlock: 55.0%
8 Holy Knight: 55.0%
9 Alchemist: 54.8%
10 Necrolyte: 54.6%

Looser Heroes:
1 Sacred Warrior: 37.3%
2 Butcher: 38.8%
3 Centaur Warchief: 41.2%
4 Windrunner: 41.2%
5 Queen of Pain: 41.6%
6 Storm Spirit: 41.7%
7 Chaos Knight: 41.9%
8 Stone Giant: 42.2%
9 Slayer: 42.9%
10 Lifestealer: 43.1%

Hero Hunters (AVG Kills):
1 Bloodseeker: 10.0
2 Netherdrake: 9.8
3 Morphling: 9.0
4 Enchantress: 8.7
5 Spectre: 8.7
6 Lightning Revenant: 8.6
7 Clockwerk Goblin: 8.5
8 Dwarven Sniper: 8.4
9 Tinker: 8.4
10 Juggernaut: 8.1

Master Killers (AVG longest Killing spree):
1 Morphling: 4.8
2 Dwarven Sniper: 4.7
3 Bloodseeker: 4.6
4 Netherdrake: 4.6
5 Enchantress: 4.5
6 Anti-Mage: 4.4
7 Spectre: 4.3
8 Tinker: 4.2
9 Lightning Revenant: 4.1
10 Clockwerk Goblin: 4.1

Supporters (AVG Assist):
1 Spectre: 13.8
2 Alchemist: 13.7
3 Lord of Olympia: 13.6
4 Dark Seer: 13.5
5 Vengeful Spirit: 13.2
6 Earthshaker: 13.1
7 Tauren Chieftain: 13.1
8 Batrider: 13.1
9 Enigma: 12.9
10 Faerie Dragon: 12.8

Farmers (AVG creep Kills):
1 Shadow Fiend: 191
2 Gorgon: 186
3 Morphling: 174
4 Pit Lord: 167
5 Dwarven Sniper: 165
6 Tinker: 164
7 Lightning Revenant: 156
8 Admiral: 155
9 Prophet: 149
10 Storm Spirit: 143

Wooders (AVG neutral Kills):
1 Enigma: 68
2 Axe: 68
3 Centaur Warchief: 66
4 Prophet: 66
5 Holy Knight: 62
6 Lifestealer: 59
7 Dark Seer: 47
8 Anti-Mage: 34
9 Juggernaut: 32
10 Bloodseeker: 31

Tower Rapers (AVG Tower Kills):
1 Shadow Fiend: 1.37
2 Drow Ranger: 1.36
3 Dwarven Sniper: 1.35
4 Death Prophet: 1.33
5 Morphling: 1.29
6 Shadow Shaman: 1.25
7 Oblivion: 1.22
8 Gorgon: 1.08
9 Netherdrake: 1.03
10 Anti-Mage: 1.03

Ward Bitches (AVG Wards Placed):
1 Crystal Maiden: 9.3
2 Keeper of the Light: 8.2
3 Warlock: 7.5
4 Shadow Priest: 7.0
5 Demon Witch: 7.0
6 Holy Knight: 6.1
7 Lich: 6.0
8 Twin Head Dragon: 6.0
9 Slayer: 4.7
10 Vengeful Spirit: 4.4

[FBB.Fatal]

Originally Posted by Shadowplay

You might want to look here if you want a more comprehensive stastic over the heroes: DL Hero Statistics - Single Instant Games - Dota-League Forum

There is more than this, but here is a snippet

Yep those are nice stats for sure, but they only cover public/ih games. -CM games are quite different than public/ih games, and 6.66 is quite diff than 6.64. Thanks for the link though, it's interesting.

[Lycan]

Originally Posted by XaTrIxX

I think 8-0 Kotl means something, even if not much. At least Kotl is a hero to consider in high level play. For many versions I prefer Kotl over Dazzle btw. Kotl is underrated in my opinion.

Kotl and Dazzle Serve a different Semirole

Kotl fits in a Gank/Nuker line up and Dazzle fits better in a Physical line-up.
But they both push insane and support awsome.

[Imverysorry]

The "times played" column made me laugh like hell...I mean, 60 times is the highest number --> thats far away from a number relevant for statistics

[FBB.Fatal]

Originally Posted by Imverysorry

The "times played" column made me laugh like hell...I mean, 60 times is the highest number --> thats far away from a number relevant for statistics

40 is enough to conduct statistical significance tests on two populations, in case you fell asleep in stats class buddy.

[MX.2000]

Good work OP

[Gheizen64]

Dirge 70% win rate
Lancer 83% win rate
Leviathan 70% win rate
Ezalor lol (100%)
Lycan 66.7%
Panda 75%

Gogogo turtling strat gogogo