- Dota 1 Hero Models 2017
- How To Change Dota 1 Hero Models
- Dota 1 Hero Models Download
- Dota 1 Hero Models Full
This chapter is all about Dota basics which are needed to understand Support Decisions better. 2.1 Position Model. Someone developed a model to describe how the farm is divided onto the played heroes in a game. Since there are 5 heroes per team the following 5 positions came up: As the tiers progress, the more farm and levels they get. Not all heroes use an ability model. Where applicable, these can be previewed in the Dota 2 loadout by selecting the ability model type next to 'Hero'. Details about the gameplay use of specific ability models can be viewed by selecting the relevant heroes on the Dota 2 Heroes page. I generally like the DotA 2 versions better. Except the DotA 1 models for Slardar and AM are still much cooler imo. Classic DotA Wisp is clearly superior to the new one. Once we perfect building Mechs, I expect there to be a factory in Mexico pumping out a brand call Mechsican. This will give you Dota 2 sound and theme for Warcraft 3 5. Now to add models and icons for DotA map download Models Packs Pack #1 Pack #2 Pack #3 Pack #4 Pack #5 Pack #6 Pack #7 Pack #8 Extract all Models Packs to Warcraft 3 game directory (C: Program Files Warcraft III ) Now you only need modified DotA map. No-AI version (6.74c) and AI verison.
Warcraft 3: Frozen Throne - DoTA 6.69
AI for DoTA 6.69. Updated so that a lot of the AI bugs and exploits have been fixed. This is probably the last updated and final version of DoTA.
New Heroes
New agility hero (Murloc Nightcrawler)
Range: 128
Movement Speed: 305
Hero Type: Agility
Agility: 21 + 2.0
Strength: 21 + 1.0
Intelligence: 16 + 1.6
Dark Pact:
Slark calls upon deities of the swamps to seal a dark pact. Sacrificing some of his lifeblood, Slark is redeemed of all negative effects. After 2 seconds, a dark wave of destruction emits from the Nightcrawler, dealing damage to himself as well as all nearby enemies.
Damage: 75/150/225/300
AoE: 325
Manacost: 55/50/45/40
Cooldown: 10
Deals half the damage to you immediately as part of the cost.
Pounce:
Surges directly forward and latches on to the first enemy he hits, binding it with chains of darkness. The enemy is unable to move more than 275 distance away from the point of impact for 3.5 seconds.
Pounce Range: 400/500/600/700
Pounce Damage: 50/75/100/125
Leash Range: 275
Leash Duration: 3.5
Cooldown: 16
Manacost: 75
Note: Moves forward in the direction it is looking, this is not a target position cast.
Essence Shift:
Passive.
Reciting ancient Nightcrawler incantations, Slark enchants his blade to steal the power and courage from his enemies. Every time Slark hits an enemy hero, he steals 1 point from all attributes and 1 additional from the main attribute of that enemy hero while Slark gains 4 points of Agility.
Duration: 15/30/45/60
Shadow Dance:
Has an active and a passive portion.
Covers himself with a dark smoke, he becomes invisible and does not reveal himself while attacking or casting spells for 5.5 seconds.
Passively increases Nightcrawler's movement speed and regeneration when he is out of sight of hostile units. Has a 0.7 delay when activated or deactivated.
Castle story mac download. Movement Speed: 30/35/40%
Health Regeneration: 2/3/4%
Active duration: 5.5 seconds
Cooldown: 25
Manacost: 120
New intelligence hero (Ancient Apparition)
Range: 600
Movement Speed: 295
Hero Type: Intelligence
Agility: 20 + 2.2
Strength: 18 + 1.4
Intelligence: 25 + 2.6
Custom hero model by PrMosquito
Cold Feet:
The Apparition curses his foes with The Frozen Mark which will freeze enemies to the ground. Unless target enemy moves at least 740 away from the mark within 4 seconds he is frozen still.
Damage per second: 37.5/50/62.5/75
Stun duration if triggered: 1.25/2/2.75/3.5
Cooldown: 15/13/11/9
Manacost: 90
Single Target Spell
Note: Target takes damage per second until he is stunned or moves out of range.
Ice Vortex:
Gathers ancient frost energy into one area to create a small, immobile ice vortex at a location. Whenever enemies are near it, their movement speed and magical resistance are reduced.
Movement Reduction: 18/22/26/30
Magic Resistance: -10/15/20/25
AOE: 275
Duration: 12
Cooldown: 6
Manacost: 80/90/100/110
Chilling Touch:
Releases a gust of cold wind at a target area, enchanting allied heroes with bonus magical damage on their attacks. Lasts for a limited number of attacks up to a maximum of 40 seconds. While the enchanted heroes have this chilling touch, their attack speed is reduced by 15%.
Maximum number of attacks: 2/3/4/5
Bonus Damage: 40/50/60/70
Attack Speed Reduction: -15%
Cooldown: 50/46/42/38
Manacost: 140
Note: Damage only triggers on enemy heroes, not creeps.
Ice Blast:
Ancient Apparition projects his presence out to make way for a powerful ripple of hailstone and dark magic. Once cast, it can release the spell to strike targets within 200 area of where his presence was upon release. It will deal damage to enemy units and apply a frostbite curse on them for 7 seconds where their health regeneration will be frozen and will shatter if they fall close enough to death. This spell has global range depending on how long you wait for the projections to move. Frostbitten units will take minor damage over time.
Impact Damage: 250/350/450
Shatter Threshold: 10/11/12%
Cooldown: 44/32/20
Manacost: 100/125/150
Note: You can charge (non-channeling, time based) this as long as you want if you want it to travel further outwards. Has an initial aoe of 200, but increases by a small amount the further it travels.
Hero Changes
Alchemist
- Reworked Unstable Concoction
Razzil brews up an unstable concoction that he can throw at an enemy hero, to stun and deal damage. Once Razzil starts brewing, he can continue for up to 5 seconds. Each second of brewing will increase the potency of the concoction and make Razzil more red. After the 5 seconds, there is a small buffer window. If Razzil does not throw the concoction soon after it will blow up and Razzil himself will suffer as if he was hit by the concoction. Concoction can blow up mid air if fired too late.
Charge Duration: up to 5 seconds
Buffer Duration: 1.5
Damage: Up to 150/200/250/300 (Physical damage type)
Stun: Up to 1/2/3/4 seconds
Cooldown: 16
Manacost: 90
The longer it charges up, the more damage and stun it deals. Listed values are the maximum.
Note: This is hero only target spell. You cannot release it if there isn't a visible target nearby (if you don't release it on time, it explodes on you instead, damage and stunning you). After casting, it builds up for 5 seconds. There is a small buffer duration after that.
Note 2: Alchemist has a red counter that counts down from 5 ontop of his head. Everyone can see this. It will be clear to the enemy when you are charging this.
Anti-Mage
- Mana Void cooldown reduced from 120/100/80 to 70
- Mana Break scaling increased from 16/32/48/64 to 28/40/52/64
Bone Fletcher
- Searing Arrows cooldown from 2/2/0/0 to 2/0/0/0
Bounty Hunter
- Reworked Jinada
Old Jinada:
Passive.
Crit: 1.25/1.5/1.75/2.0x
Crit Chance: 15%
Evasion: 5/10/15/20%
Evasion Chance: 15%
New Jinada:
Dota 1 Hero Models 2017
Passive. Adds a critical strike and small maim to your next attack. Has a cooldown.
Critical Strike: 1.4/1.6/1.8/2.0x
Maim: 25% MS/AS for 3 seconds
Cooldown: 22/17/12/7
Blood Seeker
- Bloodrage Rebalanced from 6/9/12/15 duration and 10 cd to 6/7/8/9 duration and 12/10/8/6 cd. HP loss at a constant 20 hp per sec.
Broodmother
- Spiderlings gold bounty increased from 6-8 to 11-13 and XP bounty increased from 12 to 20
Chaos Knight
- Cast Point Improved
- Reworked his Blink Strike (7225)
Reality Rift:
Target an enemy unit. Teleports you, any images you have and the enemy unit to a random point along the line between the two of you. Gives you bonus attack damage for one attack.
Cast Range: 550/600/650/700
Bonus Attack Damage: 25/50/75/100 (for one attack)
Cooldown: 24/18/12/6
Manacost: 50
Note: Chance is weighted, more likely to appear near center +/- 200 than on the extremes.
Dark Seer
- Vacuum cast range decreased from 800 to 550
- Vacuum damage decreased from 60/120/180/240 to 40/80/120/160
Death Prophet
- Witchcraft's movement bonus has been restored (3/6/9/12%)
Furion
- Force of Nature's Treants now have 33% spell resistance
Lina
- Fiery Soul buff duration increased from 5 to 7 seconds
Leshrac
- Movement speed increased from 300 to 310
Medusa
- Stone Gaze cooldown increased to 100
N'aix
- Movement speed increased from 300 to 315 Mods for the escapists.
Night Stalker
- Void manacost decreased from 90/112/130/145 to 80/90/100/110
Panda
- Lowered Fire Panda armor by 2 points, hp from 900/1350/1800/2200 to 1200 and reworked damage from 52/105/214/234 to 70/112/140/180
- Primal Split cooldown from 180/150/120/90 to 180/160/140/120
Priestess of the Moon
- Starfall AoE increased by 50 to fix issues with it not hitting units on the edge of your attack
Pudge
- Meat Hook scaling rebalanced. Cast range from 400/600/800/1000 to 550/700/850/1000, cd from 14 to 14/13/12/11 and manacost from 140 to 110/120/130/140
Pugna
- Cast Point improved
- Decrepify can now be cast on Nether Ward
Queen of Pain
- Base Intelligence increased by 5
Razor
- Agi growth decreased from 2.5 to 2.0
- Strength grown decreased from 1.9 to 1.7
- Eye of the Storm mana cost increased from 100 to 100/150/200
Shadow Priest
- Shallow Grave cooldown increased from 54/40/26/12 to 60/45/30/15
Silencer
- Base damage modified a little from 39-61 to 43-57
Sniper
- Shrapnel duration decreased from 10 to 8 seconds
- Shrapnel AoE decreased from 450 to 325
Spectre
- Dispersion now spreads the reflected damage over the number of nearby heroes. So it is the same in a 1v1 case, but greatly reduced in team fights.
Twin Head Dragon
- Ice Path stun rescaled from 0.8/1.2/1.6/2.0 to 1.00/1.33/1.66/2.0
- Ice Path cooldown scales from 16 to 16/15/14/13
- Macropyre spawn speed improved
Ursa
- Overpower is no longer removed when BKB is activated
- Overpower will not interrupt your movement when cast
Warlock
- Upheaval cooldown increased from 20 to 65 seconds
Visage
- Reworked hero (124153)
Range: 600
Movement Speed: 295
Hero Type: Intelligence
Agility: 11 + 1.3
Strength: 20 + 2.2
Intelligence: 21 + 2.5
Grave Chill:
Visage drains the life from a target, slowing its attack and move speed by 32%. He then gets bonuses equal to those drained.
Stolen MS & AS: 32
Duration: 3/4/5/6 (for both you and the target)
Manacost: 70/80/90/100
Cooldown 10
Soul Assumption:
Visage sends out a surge of agony towards an enemy unit, echoing the suffering that nearby heroes (foe or allied) have taken. For each 130 damage taken around him, his soul counter increases by one for up to 6 seconds.
Deals 20 damage + 60 for each charge.
Max Soul Counters: 3/4/5/6
Cooldown: 7/6/5/4
Manacost: 100/120/140/160
Note: Does not count damage taken from creeps. Charges when either allied heroes or enemy heroes take damage. Damage only has a 6 second window.
Gravekeeper's Cloak:
Passive.
Visage surrounds himself with the magical energy of the dead, granting extra physical armor and magic damage resistance. However, the energy degenerates with every instance of damage taken and will require time to recover back to full strength. Has 4 layers.
Armor per layer: 2/3/4/5
Magic Resistance per layer: 4/8/12/16
Cloak Recovery: 14/12/10/8 seconds
Instances: 4
Note: This is only affected by player based damage.
Summon Familiars:
Visage conjures up two blind Familiars from the realm of death. Familiars have high attack power but each attack, up to 6, drains their power and temporarily weakens them.
Familiars can transform into stone, stunning nearby enemy units upon landing and regenerating their health and attack power. Familiars give a large bounty (150 each) if killed.
Level 1: Up to 72 attack damage, 300 HP, 1 second stun.
Level 2: Up to 90 attack damage 400 HP, 1.25 second stun.
Level 3: Up to 108 attack damage 500 HP, 1.5 second stun.
Familiars take reduced damage from spells.
Stoneform Cooldown: 30
Summon Cooldown: 160
Summon Manacost: 150
Note: Can only have 2 Familiars up at any point.
Note 2: Stoned familiars are temporarily immune to damage, but they revert to normal form after 7 seconds
Note 3: Familiars will not autoattack enemies on their own, as to not consume their limited attack energy
Windrunner
- Windrunner cooldown improved from 30/25/20/15 to 15
Witch Doctor
- Paralyzing Casks now do 50 damage to heroes on each bounce
Item Changes
- Added a new recipe item (Urn of Shadows)
Requries:
- Sobi Mask (325)
- 2x Gauntlets of Strength (300)
- Recipe (250)
Total: 875
Provides:
+ 50% Mana Regeneration
+ 6 Strength
+ Active
Starts with 0 charges. Each time an enemy hero dies near you, it gains a charge. The first charges increments by 2, so it goes 0/2/3/4/5/etc
Each usage of the item decreases the charge by 1.
Each hero death can only charge up one of these items per death, so more than one hero having this will not increase the charges for both of the items.
Can be cast on an ally to heal, dispels on non-creep damage. Can alternatively be cast on an enemy hero to do half the heal (200) as damage over 8 seconds.
Heal: 400 HP over 8 seconds
Cooldown: 10
Scroll of Town Portal
- Reworked mechanics
- The casting time system for scroll teleports are now dynamic.
Whenever a hero teleports to an area, it causes any other teleports to that area to be delayed with extra time based on the number of heroes that teleported or are teleporting there. Each teleport leaves has memory lasts about 20 seconds for that location.
This does not affect teleports to the fountain area (it accounts for rough areas around shop area as well).
Example: First hero teleports to his tower, he takes the usual 3 seconds casting time. Then lets say the 3 other heroes teleport right after. The second one that telports after the original has a 5 second teleport time. Then 6 seconds, 7, etc.
- Teleport location effect no longer lies to the enemies about the landing spot
- Enemies now see the colors of the teleporting heroes
- Increases max range on where you can choose teleport near a building from 350 to 525
- The color of the teleport effect changes if you are in delayed teleport mode. It goes to normal once you have the standard 3 seconds left. So both sides can tell if its a quick or a delayed teleport and time things accoringly.
Aghanim's Scepter
- Added support for Axe. Cooldown decreased from 75/65/55 to 20
Boots of Speed
- Now available in the side shops
Blademail
- Duration reduced from 5 to 3.5 seconds
- Cooldown increased from 15 to 25
- Blademail can no longer reflect blademail sourced damage
Force Staff
- Push distance increased from 500 to 600
Linken's Sphere
- Properly blocks Unstable Concoction and Death Coil
Maelstrom
- Maelstrom attack speed bonus increased from 15 to 25
Planeswalker's Cloak
- Price increased by 100
: Radiance
- AoE damage effect can now be toggled on and off
Vanguard
- Damage block reduced from 40 to 20 when equipped on ranged heroes (50% less block like stout and Poor Man's Shield)
Gameplay & Cosmetics
* Readjusted the -bonus gold/xp system that is given in an aoe around dying heroes from being static unsplit to a dynamic split variation (numbers adjusted as well)
* Rebalanced the creep stats for your creeps when you kill an enemy rax
Old Melee:
XP Bounty: 41
Gold Bounty: 21-41
New Melee:
XP Bounty: 25
Gold Bounty: 18-26
Old Range:
XP Bounty: 25
Gold Bounty: 21-41
New Range:
XP Bounty: 25
Gold Bounty: 18-26
* Added a new small neutral camp (102473)
* Increase the buffer time in -CM from 45 to 60 seconds (This means 15 seconds total for each player, not 15 seconds for each turn)
* Added the new hero death timers to observer scoreboard title
* Reworked Recipe tooltips (128284)
* A text message and minimap indictator is now displayed when an ally purchases a Courier
* Observer board now shows ultimate cooldowns (120509)
* Added a new physical model for Gem of Truesight (124547)
* New and reworked heroes are temporarily unavailable in -CM
* Added a new visual effect when you have the Haste powerup (83240)
* A special death animation is now played when Pudge suicides (17542)
* Improved spawning location code for couriers, to help against minimap icon overlap
* Decreased cooldown on Courier Morph and added a mini techies courier (123947)
* Fixed Inner Beast buff icon
* Removed extra spikes that appear on some (except Leshrac) characters with Aghanims Scepter (11289)
* Added new sound effect for Scream of Pain
* Improved quality of Panda's Thunderclap icon
* Removed the gorge in the terrain near the bottom sentinel lane
* Added visual effect for Huskar's Scepter (20814)
* Some Nighstalker visual improvements (5227)
* Rotated the direction the Frozen Throne is facing (82193)
* Changed Scourge consumable shop model from Graveyard to Tomb of Relics (114591)
* Added hero pick/ban data to replays for -CD mode (same data format as -CM)
* Added hero pool information to replays for -CD mode (Format: Pool+index,id)
* Fixed hero death timer overwriting player name color for leavers
How To Change Dota 1 Hero Models
* Added a camera control command (-center) that locks the camera to your hero. -centeroff disables it (Not recommend for regular games)
* Clarified Ward tooltips
* Fixed lots of typos
* Replay Data: host playerid is 0 if the game was created using HCL automode
* Adjusted where the kill timer shows on the open scoreboard. Names are always visible now.
* Some minor code optimizations when heroes are picked
* Invoker: Forged Spirit units are now added to your control group when created
* Invoker: Added an Ally/Observer only visual effect that plays before sun strike lands
* New icon for Essence Aura
* New icon for Darkness
* New icon for Desolate
* New icon for Void (126145)
* New icon for Enchant (126145)
* New icon for Impetus (126145)
* New icon for Nature's Attendants (126145)
* New icon for Backtrack (126145)
* New icon for SA's Blink Strike (126145)
* New icon for Crystal Nova (118362)
* New icon for Death Ward (126145)
* New icon for Divided We Stand (126145)
* New icon for Dual Breath (126145)
* New icon for EMP (126145)
* New icon for Exorcism (126145)
* New icon for Geostrike (126145)
* New icon for Impale NA (126145)
* New icon for Juxtapose (108662)
* New icon for Phantom Edge (108662)
* New icon for Malefice (126145)
* New icon for Reaper's Scythe (126145)
* New icon for Timelock (126145)
* New icon for Timewalk (126145)
* New icon for Invoke Tornado (126145)
* New icon for Last Word (126145)
* New icon for Amplify Damage (126145)
* New icon for Jinada (105435)
* New icon for Track (105435)
Bugs Fixes
* Fixed bug with swapping with Kunkka and getting Quelling Blade
* Fixed Kraken Shell not getting disabled when Doomed
* Fixed an area where you could TP and become stuck
* Fixed a bug with Forged Spirits dying when getting an allied War Cry buff
* Fixed a recent issue with Frost Arrows cast range
* Fixed Eul's movement speed bonus not functioning correctly sometimes (if your hero has a unit morphing ability)
* Fixed a minor bug with Ursa's Enrage + Fury Swipes combo against Vanguard
* Fixed some Ion Shell stacking bugs
In the last decade, machine learning applications have proven their capabilities and potential in various applications. Especially in the past few years, they have gained rapid prominence in the gaming industry and now there are countless projects creating an endless array of models interacting with different games.
As a machine learning engineer and a person who very much enjoys being a part of the gaming community, I have always wanted to dip my toes into this pool and to give it a shot myself. I have created a project using our platform at maiot called the Core Engine. Today, I would like to show you how I created a simple, yet powerful, end-to-end ML pipeline which aims to predict the winning team in a game of DotA 2.
Before we dive in, I would like to take a step back and briefly talk about the game for those who are not that familiar with DotA 2.
DotA 2
DotA 2, or also known as Defense of the Ancients 2, is an online strategy game developed by the company Valve. The game puts two teams of five players against each other, where the main objective of each team is to destroy the enemy team’s main building before the other.
At the start of each game, players get to select a character to play from a selection of more than 110 heroes. Each hero comes equipped with its own unique talents and skill set. While you progress throughout the game (which usually takes 30 to 60 minutes), you have to battle the enemy heroes, manage your economy, and coordinate with your team to destroy the enemy’s base.
The game has a significantly large player base. As I am writing this blog post, there are 600,000+ active players in the game. Additionally, with the emergence of e-sports in the last decade, it has also become one of the leading titles in the industry with tournaments for 30+ million dollars.
Last year’s The International, the main event organized by Valve every year, had a prize pool of $34.3 million and reached almost 2 million online viewers during the finals.
From a creative perspective, an online game with a large player base where each match is unique due to the dynamics of hero selections provides a perfect ground and potential for any machine learning practitioner to build a cool project. Perhaps one of the coolest and most popular examples can be found right here.
Now, let’s get back to the project!
The Dataset
The DotA community is amazingly data driven and open for sharing.Within the scope of this project, I have used a dataset that includes all the parsed data from more than 3.5 million public matches (from January 2015 to December 2015), which can be found here. Full credits of data collection and publication must go to the great people over at OpenDota (formerly YASP).
For this project, I have put the dataset on a table in Google BigQuery, which allows for easier exploration of the data, and is a database the maiot Core Engine can automatically plug into. Please let me know if you would like access to this BigQuery table as well - I am working on making it more accessible to the community.Each entry in the table (which corresponds to a single match) includes a large number of features ranging from first blood timings, gold advantages, barracks status to hero picks.
The Goal
Each hero in DotA 2 has its own strengths and weaknesses. As a result, some heroes are naturally better or worse when they are played against certain opponents. Selecting the right hero to play can easily be the distinguishing factor between a victory and a defeat. That is why drafting plays such an important role in the game.
The goal of this project is to put this idea to test and understand whether we can make a sensible prediction about the outcome of the game based solely on the selection of heroes.
The Model
Before I talk about the architecture of the model, I would like to put an emphasis on two important challenges:
- Each hero in DotA 2 has its own skills and role within the game. Unfortunately, it is impossible to put this information into a format that is directly interpretable by a machine learning model.
- DotA 2 is a team game and each team has 5 heroes. What makes up a good team is not just about selecting 5 powerful heroes but building a joint composition of heroes with good compatibility with each other while maintaining a powerful opposition to the enemy team.
I need to figure out a way to represent each hero separately, and moreover, I have to represent the conceptof a team. In order to overcome these problems, I got a little bit of inspiration from natural language processing (NLP) models.
I have treated heroes in the selection pool like words in a dictionary, i.e., the first layer in the architecture is an embedding layer, which learns how to create a representation for each individual hero. Moreover, as words come together and form sentences, in this case, the heroes come together and form teams. Since the order of heroes also does not play a role, I have used an average pooling layer, which pools up the 5 heroes within the same team. The output of this pooling layer is then fed to a fully connected network which ultimately makes the prediction.
In order to put everything into a single end-to-end pipeline with the Core Engine, all I have written was a single YAML file. In this section, I will briefly mention some of the critical blocks in this file that are relevant to our topic. However, if you would like to learn about further details, you can check our docs here.
We start off with the feature selection. Each key under the main key
features
will denote the name of a data column that is utilized during the training. In our dataset, each column holds an integer value ranging from 0 to 112. This value represents the id of a selected hero for that player in that match. For instance, p_2_hero_id
holds the index of the hero Player 2
has selected. Furthermore, Player 0
to Player 4
form the first team, whereas Player 5
to Player 9
form the second.As for the label, the configuration is quite trivial. The problem can be interpreted as a binary classification problem (thus the binary cross-entropy loss function) and I have selected the data column
radiant_win
, which holds the value of either 0
or 1
, depending on the outcome of the game. Ultimately, the model is evaluated on its accuracy.The
split
key is used to configure how the pipeline reads the data from the source and splits it into a training and an evaluation dataset. The first part of the block is quite self-explanatory - it is an 80-20 split. However, the second part needs some explanation.Amongst the 3.5 million matches, we have only used a match if:
- the mode of the game was either
all pick
orall draft
- the game was played by 10 human players (so no bots)
- the game was won by either one of the teams
- the game was either a
normal
game or aranked
game - the game was played during either patch 6.84 or patch 6.85 (getting rid of the infamous ‘ho ho ha ha’ patch)
In the end, we had a total of 2.835.720 matches remaining.
Small note: If you would like to learn what the constant values stand for, you can check here.
Ultimately, there is the
trainer
configuration. I will not go over the details, but in a short explanation, it creates a simple feedforward neural network dealing with a binary classification problem. As explained before, the input goes through an embedding layer first, followed by an average pooling layer. The output is fed into fully connected layers, which ultimately handle the prediction.By using different variations of the aforementioned config file, I have created two pipelines with different embedding dimensions, respectively 16 and 3.
In total, it took roughly 50 minutes to completely execute the first pipeline, whereas the second pipeline took only 21. The drastic difference between the execution time of both runs is due to caching. Whilst working on a project on the Core Engine, you can choose to save the outputs of your intermediate steps within your pipeline. Because once you execute a similar pipeline, this might come in handy.
For instance, in our case, I have initially ran the first pipeline. Once it was finished, I ran the second pipeline, where every step from the data ingestion up until the embedding layer was identical. So, the second pipeline opted tosafely skip these steps and use the intermediate outputs from the first pipeline, saving both time and computation cost.
Additionally, almost every step along the way such as data ingestion or pre-processing was handled in a distributed manner except for the training (which is a feature coming soon).
Dota 1 Hero Models Download
As for the actual results, both models achieved a similar accuracy around 59-60% on the validation dataset. Considering the small set of features and the limited number of matches, they were both deemed to be successful, bearing in mind that the predictions are practically made right after the drafting phase and before the actual gameplay starts.
Dota 1 Hero Models Full
DotA 2 is an ever-changing game. Some heroes can be extremely good or terribly bad depending on the current version of the game. But one thing which always stays the same is the fact that drafting is an integral part of any game in DotA2 and this project shows that it plays a critical role in the outcome of the game.
It is also important to note that this was achieved by a model, which relied solely on the hero picks without any additional information. Taking all the other available features into consideration, it barely scratched the surface of what is possible when it comes to analyzing different aspects of a DotA 2 match. I am already building up to another follow-up project (possibly with a follow-up blog post) in my head with a deeper look into the output of the embedding layer with different heroes.
On a more general level, it is also important to note that this whole experiment was conducted using an end-to-end pipeline which required just a yaml configuration file to execute. Since serving is one of the features thatwe are planning to publish soon, this yaml file could very well be the only thing standing between a developer and a deployed model in an application.
Finally, if you are interested in the Core Engine, you can find more details on our website right here. Additionally, if you have any questions or any other cool ideas which can be applied to this dataset, feel free to reach me at [email protected].