Wednesday, November 20, 2013


Heading overseas with no internet for next 3 weeks. BRB.

Friday, November 15, 2013

Let's Play (Part 8) Venomancer


Every week or two I generally try to play a new hero to discover what it’s like playing it, and in this series of blogs I will make a post on my experiences with it.

Official Starcraft 2 Hydralisk art.


Venomancer (Lesale Deathbringer) was based off the Hydralisk from Starcraft. The Hydralisk is a Zerg creature that evolved to fire volleys of poisonous armour piercing spines with tremendous force, posing a deadly threat to infantry, tanks and aircraft alike. Swarms of Hydralisks can be found at the forefront of any Zerg invasion.

In DOTA2, the Venomancer is a support hero with incredible poison damage output. All his abilities are based on poison damage over time coupled with movement impairing effects. Even receiving a single attack from him will leave a lingering poison that will do severe damage - and if he manages to strike a hero with all his skills - certain death is sure to follow. A lane with a Venomancer is highly favoured to strike first blood. All this offensive power comes at a cost, however - the Venomancer has nearly no way to defend himself, and if he gets caught out of position he can be killed very quickly.


His potent slow and high damage allows him to dominate his lane, with a proper lane partner.

Plague Wards provide essential vision during all phases of the game, protecting him and his allies from ganks when placed into the jungle and key passage points, as well as serving as a powerful pushing / anti-push spell.


His poor range (450) and slow movement speed make him awkward to lane. While he is good at dominating his lane from a position of strength, he cannot salvage his lane from a position of weakness due to his lack of defensive and escape abilities.

With no good way to mitigate enemy damage done, and no mobility or protective skills, getting caught out alone means almost certain death. This can pose difficulties for finding farm in the mid-game. Once BKBs come online for key opponents they can ignore him completely.

Venomous Gale

All of Venomancer's skills have really bizarre scaling characteristics and this one is no exception. This skill applies a potent slow (50% move-speed reduction, scaling down to 0% over 15 seconds) making it a popular choice in a tri-lane seeking first blood, however at level 1 this skill does practically no damage. The damage scaling on Gale goes from 25-200-375-550 - so you definitely want at least level 2 of this as fast as possible, and the damage at level 4 is possibly the highest non-ultimate nuke damage in the game. It's a skill-shot that is reasonably difficult to land - it's got a 800 range and radius of 125, and travels at 1200 speed (which means it takes 2/3 of a second to reach its max distance) and has a longer than usual cooldown (22 seconds).

Poison Sting

This passive skill adds an additional poison damage over time and minor slow to your target. Like Gale, this skill starts out weak as well and then grows to ridiculous power by the end - the damage scaling goes from 30-90-180-300.

In mid-game team clashes it's important to at least try to hit every enemy hero with this ability once - combined with the large damage over time from Gale and Poison Nova, this can ensure kills on fleeing enemy heroes.

Plague Wards

This is the skill that makes up for Venomancer's lack of defensive abilities. The Plague Wards are an early-warning system of sorts - he can set them up on cliffs and in the forest, warning him and his allies of approaching ganks. You should always have several Plague Wards watching the river and jungle next to your lane at all times, particularly when trying to push their tower.

The physical damage that the Plague Wards do to heroes is nearly nonexistent - they do piercing damage, so their tooltip damage is cut in half against heroes. The main reason to upgrade them is their hit points - it scales from 75-200-325-400. At level 1 it survives only two auto-attacks from enemy heroes, but at level 2 it takes about four auto-attacks. In particular, at level 1, if the enemy attacks your ward once you can immediately deny it to stop them getting the gold for killing it. The level 2 upgrade is the most crucial in this case, same as Gale. Most of the DPS from the Plague Wards actually come from leveling the Poison Sting passive - the wards apply a toned down version of Poison Sting that does 50% of the original damage.

Skill Selection

It's not very obvious what order the skills should be taken in.

A single shot from a Level 1 Ward with Level 4 Sting does 155 total damage, while a single shot from a level 4 Ward with Level 1 Sting does 34 damage.

If you were comparing continuous damage, however - where the enemy stood there and let the ward attack them repeatedly until he died - the Level 1 Ward with Level 4 Sting does 13 damage per second, while the Level 4 Ward with Level 1 Sting does 15 damage per second. This would apply to team-fights where the enemies died before being able to run away, but the difference is really minor anyway.

In terms of maximizing damage against heroes, maxing Poison Sting first is always the best option, especially because it also boosts your auto-attacks. However, the HP bonus to bringing the wards to Level 2 is very significant. One might also argue that it doesn't matter, since the ward really only needs to fire once before it dies, but having at least Level 2 of it might be enough of a deterrent to the enemy that they back off rather than kill it.

The first rank of Poison Nova is also rather underwhelming. 432 damage in a wide area (830 radius) sounds good, but compared to a maxed Gale (550 damage) or maxed Poison Sting (300 damage) it starts to sound underwhelming, especially considering its heavy mana cost. One might argue at lower levels being able to cast Gale twice is better than being able to cast Gale + Poison Nova, but I'm really undecided on this as the Gale cooldown is really long anyway (22 seconds). At higher levels Poison Nova gets a lot better, going to 812 and then 1215 damage.

Another fact to note is that Poison Nova damage is non-lethal, so you will need another lethal source of DOT to back it up if you anticipate the enemy running away. Higher levels of Poison Sting will work as that extends the duration all the way out to 15 seconds, as well as landing a Gale on them (also 15 seconds). Realistically you can't always land Gale on the targets you want to, which is why maxing Poison Sting first is useful to ensure the lethal damage on targets with Poison Nova. Urn of Shadows damage (over 8 seconds) can be lethal as well.

At level 11 he can do potentially 812 damage from Nova, 550 damage from Gale and 300 damage from Poison Sting, hitting for 1662 damage on multiple targets.


Venomancer is best run in "kill lanes" with other powerful early game heroes to suppress the enemy and even score first blood.

The large amount of poison damage DOT applied is particularly crippling to heroes like Templar Assassin or Treant Protector which have charge based shields.


Venomancer is vulnerable to being ganked, so heroes like Bounty Hunter or Stealth Assassin can make life very difficult for him.

Heroes with large HP pools or healers can mitigate much of the poison effect.

Item build

Veil of Discord deserves special mention here, as Venomancer is primarily a magic damage based hero and can make good use of it.

Most of his items will focus on keeping him (and his teammates) alive, as you need to stay alive long enough to make sure the enemy team is affected by your spells and slowed down so they can't escape, so Mekansm and / or Force Staff make an appearance. Urn of Shadows is ideal for the small amount of mana regeneration and hp boost it provides, as well as the additional lethal DOT you can apply to fleeing enemies who have Poison Nova on them but not been hit by your other lethal DOTs.

Arcane Boots will probably be your boots of choice if you want to use your spells more frequently but Phase Boots are good as well, the choice is situational.

Wednesday, November 13, 2013

Basic weapon DPS theory


This is a "Playing Better" segment focusing on weapon choices.

Basic Math

The basic theme running through all games is that you need a balanced set of stats to maximize values that are determined multiplicatively. As an extreme example, assume damage multiplied by attack speed = damage per second, so A x B = C, and assume buying A is as costly as buying B. You get an optimal result (maximum C) when A = B, for example, if you could buy 6 units of "weapon power" you would get the best result when A =3 and B =3 so A x B = 9. An imbalanced scenario, for example A = 2 and B =4 would yield C =8 and even worse when A = 1 and B = 5 where C = 5.

Example: Mirana

Of course, reality isn't quite so simple, but the principle still holds in general. Let's pick Mirana, for example, at level 20 she has 106 damage and 1.07 attacks per second for a DPS of 113. Ignore her 16% attack speed buff from leap. Her basic set of early game items would maybe consist of Phase Boots, Drums of Endurance and Ring of Aquila, bringing her to a DPS of 186. By level 20, assume she wants to maximize her damage and she could buy two items. The four main items of note are Daedalus or MKB (attack damage) and Mjollnir or Butterfly (attack speed). Common wisdom says you need to pick one from each group. I've crunched some numbers and put them in a table below.

Daedalus MKB Mjollnir Butterfly
Daedalus 612
MKB 607 550
Mjollnir 653 601 585
Butterfly 620 554 565 525

Mirana DPS output of different weapon combinations.

I think the first thing to note is that the variance in DPS output between the choices isn't particularly great. This is because even without advanced items, her core and base stats give her a respectable amount of damage and attack speed already. This means that you should be picking items based on the utility they offer, rather than the raw potential DPS they can output.

Mjollnir and Daedalus seem to be the best combination - not only is it the highest DPS but also offers good AOE from chain lightning and even more DPS from static shield. Building its other counterpart - Butterfly and MKB - offers 20% less damage done and no AOE, but gives you mini-stun, true-strike and evasion. The necessity for utility like this has to be judged on a match to match basis. If you're facing an opponent with a Butterfly then the MKB is immediately better than the Daedalus.

Even some of the weirder combinations that sound terrible aren't really that bad when you think about it, because their damage penalty isn't that great.

Double Mjollnir seems to be in the category of "joke" builds with its 44% chance to proc chain lightning per attack, but you might use it against a team with a lot of summons or multiple Linkens holders (the chain lightning attack breaks the Spellblock, even on secondary bounces). It's not just the increased proc chance that's at work here, it's also the increased attack speed, so there's a synergistic effect - a single Mjollnir has a 61% chance of proccing a chain lightning in a 2 second attack window, while a double Mjollnir has a 91% of proccing a chain lightning in a 2 second attack window. Also note that in the last patch Mjollnir chain lightning got buffed to a 900 search radius, so you're going to be hitting everyone in a teamfight when it procs, dealing significant damage.

Double Daedalus is actually quite commonly seen on Kunkka, because of its interaction with Tidebringer (he only cares about raw damage, not attack speed, and the proc on MKB does not help him) and the damage output is pretty competitive with the other options, which I found surprising.

There is pretty much no reason to ever go double Butterfly or double MKB unless you're trolling the enemy. Maybe double MKB with its 58% chance to stun on hit against a Pugna who has an Aghanims. Double Butterfly is nice with the 51% dodge but don't be surprised if your enemies buy an MKB immediately to put an end to your trolling...

Other notable mentions

Special mention for Desolator and Cuirass if you felt like it would help your team. Taking an opponent's armor from 12 to 0 increases DPS by 71%. Mirana's own DPS with this combination however only sits at around 560, which is on the low side but still competitive with the rest of the item choices in the table. The big advantage of this combination is the fact that it affects towers and buildings making sieges a lot faster and safer, and of course the fact that all other heroes on your team will benefit from this, potentially making this far more powerful than the other options as long as you have at least one other physical DPS oriented ally.

In reality, maximizing DPS isn't the primary concern for heroes. With the amount of disable and escape effects in DOTA2 and the already high lethality of most heroes, most of the time you're better off purchasing utility items that allow you to shrug off disables, lock down opponents or get the jump on them. This means items like BKB, Force Staff / Blink Dagger, Scythe of Vyse or Orchid, etc. You will find that Ursa for example more commonly builds a Scythe of Vyse rather than a Daedalus. But it's good to know the theory anyway.

Appendix (mechanics)

Mjollnir overrides all other unique attack modifiers (UAMs) only when it procs, so it stacks diminishingly with itself.

Daedalus crits also overrides other weaker or equivalent crits when it procs, so it stacks diminishingly with itself.

MKB bash is not an UAM to begin with, so it has independent proc chance (fully stacks) and can double proc (applies two separate instances of +100 damage in a single attack). This 100 damage is magical in nature and thus does not interact with Daedalus. Magic immune targets ignore the stun and bonus damage.

Butterfly evasion stacks diminishingly with itself as of the latest patch.

Tuesday, October 29, 2013

Hearthstone vs MTG


Hearthstone is the name of a much anticipated free to play online collectible card game being developed by Blizzard, and the name is genius in itself - in World of Warcraft, the hearthstone is a magical object that will instantly transport you home, no matter how far away you've traveled. Summoning a Murloc and hearing it go "Mrrraggglhlhghghlgh!" instantly took me back to my memories of leveling my first Night Elf in Darkshore, 8 years ago.

Fill the board with 1/1 Whelps? Sounds just like the Onyxia fight =)

Let me get this out of the way first: if Hearthstone succeeds, it will actually be on the basis of its rich sound and art design - the act of playing cards and attacking with them rewards the player with very satisfying sounds and animations. I can't help but contrast it to MTG online, which is a relatively spartan experience. In Hearthstone, your cards actually leap up and smack the opponent in the face, with more visual effect and force representing higher damage (and the crowd, presumably around your table, going "oooh!" if you get a particularly brutal hit), fireballs and arcane missiles fly through the air, minion cards all have different dialogues for being summoned, attacking and dying, and it's a visual and aural treat to see it all happen at once - there is such a sense of accomplishment in giving all the orders and letting them animate themselves. Forget the game design for a minute: the mechanical act of playing the game is simply joyful, and it sounds counter-intuitive but even if the game design itself was somehow lacking it will still achieve a measure of success.

What is Hearthstone?

Hearthstone is an online card game that drew most of its inspiration from MTG, but with a much more streamlined and simplified rule-set. You pick a hero and duel another player by summoning minions and unleashing spells at each other. While the rules differ slightly, all the common themes of card game mechanics come into play - aggro, control, removal, card advantage, tempo, etc. MTG players will feel right at home. Cinematic trailer here.

Upon completing the tutorial and playing some games with the practice AI, you are awarded a full set of basic cards for all 9 classes (corresponding to the 9 classes in WoW). Further cards can be obtained through opening packs or playing in the arena, for a gold cost - you gain gold by playing games and completing daily quests. There are two types of games - you can play Constructed, where you build a deck from the cards you've collected and challenge other players of similar ranking, and there is Arena play, where you build a deck from a completely randomized set of cards (including ones you haven't obtained yet), and then challenge others in an elimination series that only ends once you've either lost 3 matches or won 9. Arena play requires a gold entry fee, but will reward you with progressively greater prizes the more matches you win. Both these modes feed off each other: you will play some Constructed to earn gold, which you  can elect to spend on the Arena entry fee which then rewards you with more cards that you can use in Constructed.

See how simple the card design and text is. Also, bad memories of Gnomeregan...

You can spend money to buy in-game gold to accelerate your acquisition of cards. But thankfully many of the basic cards are extremely powerful anyway: there are some basic-only decks which players have taken to the highest bracket of play. Also, there is no option to trade or buy and sell cards - after Blizzard's disastrous foray into real money trading in Diablo 3, they have wisely decided to let each player create their own journey of card acquisition. Extra cards are "disenchanted" and turned into arcane dust which can be used  to craft cards you don't own. To put the card acquisition speed into perspective: if you do the daily quest every day, roughly taking half an hour on average, you gain about 40-50 gold. A pack of 5 cards costs 100 gold or $1.50.

The game is still in closed beta but the hype surrounding it has been feverish. Closed beta keys are being doled out by Blizzard for testing but they're not coming fast enough - fans have been reselling keys illegally online for $50 each.

Game Design versus MTG

As this is a game design blog I would like to compare Hearthstone with the giant in CCGs - Magic the Gathering. MTG is basically the WoW of the CCG world - many companies have attempted to overthrow them over the last decade, but none have come even remotely close. Their game design and artwork are completely unmatched. But by definition no one can imagine the king being unseated until it actually happens. A similar situation happened with DOTA and LOL - a former DOTA developer created a more casual friendly game, and now LOL has many more players than DOTA2.

MTG's art direction is simply gorgeous

World Building

MTG had a daunting task ahead of them - they needed to create an immersive world that draws the player in, using nothing but art and the flavour text on their cards. It's safe to say they've done an excellent job - they're set a nearly impossibly high bar for other CCG makers to aspire to.

MTG text is a little on the small side, especially for playing online

Hearthstone on the other hand, doesn't have to. The lore of the world and characters are well known to the target audience (former Warcraft and WoW players). The sound design and art design are all reused for nostalgia reasons. This allows them to skip the flavour text: it also allows them to skip a lot of explanation of card mechanics (since it's online only, the computer can handle the complicated interactions for you). Overall, this provides a much cleaner, simpler card design - much more ideal for online play since you can actually read what each card does without zooming in.

Mechanics: Mana Curve, Deck Size

In MTG, you need to play land cards (one per turn) and tap them for mana. It's possible to draw too many (or too few) lands in your opening hand, making this a complicating variable in deciding whether to mulligan. The second player to go gets compensated by having an extra card.
Everyone has bad memories of being mana screwed in MTG

In Hearthstone, you have a linear mana ramp - you automatically get one mana crystal a turn which you can drain for mana. This means your deck of 30 cards are all spells - compared to MTG's typical composition of about 35 spells and 25 lands. Your starting hand size is 3 or 4, compared to MTG's 7 or 8. The second player to go not only gets an extra card like in MTG, but also gets a single use spell that gives him one mana. Also, both players get to mulligan as many individual cards they like out of the starting hand in the hopes of drawing a more favourable starting hand - this makes it more viable to run less early game drops, because you have a cost-free way to improve your chances at getting them. Most critically, this helps the late game go smoother, because you're always drawing answers from your deck, rather than top-decking a useless land card.

Mechanics: Turn Order

In MTG, the attacker would declare his intention to attack, and the defender would assign blockers. Both players are active on either turn - even if it's not your turn, you can play instants and creature abilities, and you need to assign blockers for when the enemy attacks. In fact, due to the spell stack, every time you play an ability or card, you need to give the opponent a chance to respond. This really kills the fluidity of the game in my opinion. (I play a card. Ok? I use this ability. Ok? I attack. Ok? Damage is assigned. Ok? I use some ability. Ok? And so on.)

In MTG instants can be played during the opponent's turn, while sorceries can't

In Hearthstone, it's the opposite: the attacker gets to choose what he wants to attack. Things are a lot simpler - there's nothing to do when the enemy is running his turn, so you can sit back and relax. There are no instant speed abilities to run when it's the opponent's turn, and thus there's also no spell stack. This clear delineation of active / passive turn allows the game to resolve very quickly - if you know what you're doing, you can give all the orders in a few seconds and then pass the turn back to the opponent. The downside, of course, is that this removes a lot of player interaction and strategic depth from the game.

Mechanics: Persistent Damage

In MTG, damage is wiped at the end of every turn - if a creature didn't take lethal damage, it gets healed fully in the end step. This is because the mechanics of tracking damage counters across multiple creatures would be too difficult.

Again, the card design is so much cleaner and screen-friendly

In Hearthstone, damage is persistent, so it's possible to slowly ping a large creature to death across several turns. Having the computer calculate and keep track of things makes mechanics such as random damage or copying cards easier to implement.

Final thoughts

I like the new streamlined design of Hearthstone compared to MTG. The games play out a lot quicker. In all the areas that matter, I think the loss of strategy and depth has been outweighed by the gains in simplicity and convenience. Their version of drafting - Arena - doesn't have anywhere near the depth of strategy that MTG drafting does, but has the incredible convenience of letting the player build their deck and play their games out any time they want - the matchmaking system simply matches players with the same number of wins / losses (if you've gone 1-2 so far in the Arena, you'll probably get matched up with an opponent who also went 1-2). You could play one 10 minute game every two days if you wanted to, while in MTG you needed to commit a solid 3-4 hours to play a draft.

The ranking system in Constructed is also a great way to separate the player tiers - although this game is "pay to win" in some sense - some Legendary cards are particularly strong - it's also made irrelevant by the matchmaking system. If you decide to play a really basic deck with a low power level, you can have a completely good time playing in Bronze or Silver. In fact, some would argue it's an even better play experience there, as you will see a much larger diversity of classes and deck types. In short, paying money for Constructed cards is completely optional: if you decide not to, it's unlikely you'll be matched against opponents who did anyway.

Never thought I'd hear the sound of a Shield Slam again

There's of course some ongoing problems with the game, but I don't think it's anything unusual for something that's just getting started. The card pool is still really shallow - only 400 cards total - thus limiting the possible strategies for Arena and Constructed, but nothing that can't be fixed with the ongoing release of new card sets. Class balance and power level is an ever contentious issue. Still, even in its current form in closed beta I'm having a great time playing it and I think it's definitely already ready for release - having a game a little rough around the edges adds a little charm to it in my opinion.  

Monday, October 21, 2013

Design (Part 8) Asymmetry, and transitioning from LOL to DOTA2


One of the most common questions I get from LOL players just starting DOTA2 is “so… what’s the team strategy here?” LOL has a standard laning strategy that even low level unranked players are expected to know and follow. All players know the roles that are required for the team and what items to buy - solo top bruiser, solo mid AP Carry, duo bot lane with support and AD carry, and a jungler. Furthermore, the composition is mirrored across the lanes because the map is symmetrical - the carries face off with each other, the bruisers face off with each other, etc. DOTA2 can seem very chaotic in comparison, with a seemingly random number of heroes in a lane (between 0 and 3) and an arbitrary amount of junglers, with no set rules on team composition (anything from 0 to 4 carries) and an asymmetrical map where the side lane matchups are always unfair for one side or the other, so you're never on even footing with your opposition. Pick your strategy wrong, and you can get absolutely crushed. In comparison, you can't pick the wrong strategy in LOL, because there's only 1 strategy, so every match is a close match (relatively, at least, compared to DOTA2)

This is actually an interesting question from a game design perspective - is fairness anti-fun? Can it be possible for a game to be "too fair" and "too balanced?" Is forcing some players to play at a disadvantage - for example, in DOTA2's safelane vs hardlane mechanic - a good thing or a bad thing? There are some clear disadvantages to DOTA2's chaotic lane strategies - with the lack of communication in pubs, everyone has at some point ended up playing a melee hard carry forced to go into the hard-lane solo against 3 enemy heroes and gotten absolutely crushed through no fault of their own - such a thing would never happen in LOL. This aspect of gameplay has both been proclaimed as its greatest strength and also greatest weakness - in DOTA2 when you get crushed, you really get crushed, and many losses at low level play stem from your team composition, not the individual skill of the players. While in LOL, people complain that every match more or less plays out the same way, with little variety in strategy and team composition making it stale.

Laning in LOL

LOL has a stable strategy because many of their gameplay elements are symmetrical.

Ability Power (AP) and Attack Damage (AD) carries both scale into the late game. Given you have two carries in the game, you would always get one of each type to force the enemy team to get an inefficient mix of magic resist and armor stats. Mid lane is given to the AP carry, because in general AP carries are easier to gank - mid lane is the shortest lane and thus the safest for them. Also, the blue jungle buff is relatively close by. Typical mid lane champions are Ahri, Annie, Brand, Lux, Orianna.

Where does the AD carry go? The Support champion will accompany the other carry, given a choice between top or bot lane, we find the bot lane is the superior choice because the Dragon pit is there. In the early to mid-game, most of the game will focus around the Dragon, since Baron is not an option yet, so it is better for your duo lane to be there. Typical AD carries are Ezreal, Vayne, Corki. Typical supports are Thresh, Taric, Sona.

Top lane will thus be 1v1, and generally the Bruiser or Tank type champions will go there. They need to be tanky, because they need to be able to hold the lane on their own – they are a long way away from any help. Typical bruisers are Renekton, Nassus, Vladmir, Tryndamere, Shen, Jarvan, Zac, Elise.

And finally, there is always a jungler, because it’s difficult to translate a 2v1 advantage into a large enough XP and Gold advantage to offset the disadvantage of losing a jungler. Generally, a team which sends 2 heroes top or mid will “win” those lanes, but still end up behind the other team who is running a jungler when comparing total gold and exp. Typical junglers are Jarvan, Elise, Lee Sin, Cho Gath.

Because the lanes are symmetrical, each team’s composition will exactly mirror the opponents they are facing in the lane.

So in a nutshell, this is why the champion composition and laning strategy is the way it is: it’s achieved a local maximum, any change to this composition makes it worse off. Virtually every game you play in LOL will follow this formula, unless you are top tier team experimenting with unusual strategies to throw your opponent off their game, or you get a rebel in your team wanting to try something new and isn’t afraid to be scolded by their teammates for breaking the meta.

Why is the DOTA2 metagame so unstable?

DOTA2’s map and game mechanics have a number of asymmetrical attributes that allow “either/or” decisions by the teams. Usually game designers aim for symmetry because it’s the easiest way to ensure the game is balanced, but asymmetrical gameplay elements can create an unstable metagame and allow a variety of viable strategies to exist at the same time.

1) Asymmetrical map – in LOL, the map is perfectly symmetrical - the terrain features don’t favour either side during the laning phase. However, in DOTA, the map has what is called “rotational symmetry”. Each side lane has terrain that favours one side over the other, so the top lane is the Dire easy lane and the Radiant hard lane, while the bottom lane is the Dire hard lane and Radiant easy lane. This creates an unfair matchup that teams need special strategies to handle – teams can either allocate more resources / heroes to the difficult lane to ensure they win it, or they can allocate less to the lane to ensure the impact of losing it is smaller, or run balanced lanes which can adapt to the enemy’s strategy. It’s very rare for carries to face off against each other - the carries typically go to their respective safelanes to get farm, which are on opposite sides of the map.

2) Asymmetrical scaling –  In LOL both AP and AD champions scale with items so there is no advantage to running an all AP team or all AD team - you mostly end up using one of each. In DOTA2, the hero scaling is asymmetrical - Intelligence heroes get no spell damage scaling and tend to peak earlier in the game, but Agility heroes tend to peak later and harder with items that increase physical damage. Teams can choose to either run teams with more early game heroes (to win early) or more late game heroes (to win late) or to build a balanced team which can adapt to the enemy’s strategy.

3) Runes on mid – In DOTA the 2 minute runes form an important part of the strategy, offering powerful temporary boosts and regeneration (from refilling the bottle). Mid lane can either be run LOL style by placing a hero with poor mobility there for safety but giving up the runes, or it can be run as an aggressive mid by placing a hero with strong mobility to contest the runes and use them to gank the side lanes.

4) Amplifying asymmetry - Deny mechanics and creep pulling / jungling– In DOTA2, if the enemy lane presence is weak enough, players can use these mechanics to further dominate the lane they are in. Denying creeps reduces the experience and gold the enemy obtains from the lane, and creep pulling further boosts your lane gold while completely stopping enemy experience gain for a time. This is why junglers are optional in DOTA2 – it forces one of your lanes to play at a disadvantage, and this exp and gold disadvantage can be larger than the gold and exp your jungler is generating while jungling. There can be anywhere from 0 to 3 junglers in a DOTA2 game (defensive jungle, offensive jungle, and ancient stacking). Creep stacking also allows supports to sacrifice their farm to create jungle stacks for carries, amplifying their farm, allowing teams to run anywhere from 1 to 3 carries - such a redistribution of income is not possible in LOL, where typically your bruiser, mid and ADC can all obtain a high amount of farm every single game.

5) Limited Vision and Teleports - in DOTA2 there is a limited amount of observer wards available - typically only 2 up on the map at any given time - and the Smoke of Deceit consumable allows heroes to pass under wards undetected anyway. This means that it’s much harder to protect a hero against ganks if the opponent really wants to commit to it, preventing stalemate scenarios where the map is well warded enough that neither side can disrupt the early game farming phase of the other. It also raises the stakes when one side manages to deward even one or two observers - it means they know the enemy has literally no vision on the map for several minutes, giving them temporary map control. Overall this means is that if the opponent wants to disrupt your strategy, they usually can if they devote enough resources to it. At the same time, the existence of teleport scrolls allow any hero to teleport to another tower within 3 seconds - which means that if one of your lanes is getting dived, your team can always decide to send help - it's just a matter of how much resources you want to devote to it - yet another either / or scenario to choose from. Even if you successfully stop the gank, the total gold and exp you lose by leaving the lane could leave your team worse off.

As a cool final fact - the positioning of Roshan (the equivalent of Baron - his name, Nashor, is a homage to the original Roshan in DOTA) heavily favours the Dire team, and this plays out in competitive statistics - of all Roshan kills at pro level gameplay, 66% go to the Dire team. However, the Radiant / Dire winrate is nearly equal, at 50% plus minus 1%. This means the asymmetries in the rest of the map - of which there are many - must be offsetting it almost exactly. For example, Radiant mid can be pulled and Radiant ancients can be jungled at level 1.

Every one of these factors force a team decision that could go either way – what sort of heroes to pick, how many heroes to allocate to which lane, what type of mid to use, whether to utilize a jungler or not – leading to a multitude of possible lineups.

General laning strategy in DOTA2

Very broadly, each of the 3 lanes yields different levels of experience and gold income and exposes your hero to varying levels of danger, and this affects the suitability of different heroes in each lane.

Mid Lane Solo
(Gets high experience, medium gold, and has moderate danger of ganks)

Mid specialists like Puck or Queen of Pain scale a lot better with experience than with gold - hence they are typically put in the Mid lane. Being a short and safe lane it’s unlikely they can be harassed out of experience range even by 2 or 3 heroes, guaranteeing them solo experience gain and allowing them to be the highest level on the map. Mid heroes generally want AOE skills to clear waves, mobility skills to contest runes, and high base damage to out-last hit their opponents. The sequence of play revolves around clearing the creeps to push the lane just before the 2 minute rune respawns so you can safely pick up the rune (and recharge your bottle) without wasting exp and gold. This leaves the enemy in a catch-22 situation, where if he tries to contest the rune he loses a full wave of experience and gold, yet if he doesn't you'll eventually force him to leave the lane due to the extra regeneration you're getting by bottling the runes. There are also heroes which can’t contest runes and generally “lose” mid but go there anyway - for example Nightstalker, or Drow who don’t have AOE clears or mobility skills. They still lane mid anyway and take the loss because they need a safe lane to gain quick experience for their powerful ultimate they will use to dominate the enemy later. These heroes rely on being higher level than their opponents - they can't "play from behind". There's also a third class of heroes who can go mid, like Outworld Devourer or Bloodseeker who have such superior last-hitting power that it doesn't matter that they can't contest the runes, they win by denying the enemy gold and experience.

Safe Lane Carry 
(Gets  medium experience, high gold, and low danger of ganks)

Hard carries like Faceless Void and Anti-Mage scale a lot better with gold than with experience - hence they are typically put in the Safe Lane Carry position. Their supports will defend them against enemy harassment, allowing them to get a lot of last hits, however the presence of supports in the lane reduces their experience gain. The supports will typically be also engaged in pulling the camps (to gain exp / deny exp to the enemy) and warding the forest to prevent ganks and stop their pull camp from being blocked. They typically buy “farm accelerators” like Midas or Battlefury as soon as possible, and then spend some more time farming one or two major items before they’re ready to teamfight at about 35 minutes in. There are also Semi-carries like Juggernaut or Razor who are drafted into teams who want to win with an earlier timing - they may directly buy a teamfight item like an Aghanims or BKB and take the fight to the enemy. If the enemy team had a hard carry who bought a farm accelerator instead, they will be in a weaker position for awhile until they catch up.

Off Lane Solo
(Gets medium experience, low gold, and high danger of ganks)

Off lane specialists like Bounty Hunter, Mirana or Windrunner have abilities which allow them to solo the dangerous off-lane even against multiple enemies. Like mid specialists, they also scale a lot better with experience but have even less reliance on gold. In the off-lane, they can usually stay within experience range of the creep wave to gain relatively quick levels but it’s likely they won’t be able to get much in the way of last hits. They typically need to ward the enemy pull camp and jungle area to stop the pulls and to protect themselves against ganks, and they may bring sentry wards themselves to counterward. Heroes like Nature’s Prophet and Lone Druid can also go to the offlane, because their summons can manipulate the enemy creep wave (dragging them to your tower through the forest so you can farm them safely) and also block the enemy pull camp. Because of the asymmetry of the map, it’s a foregone conclusion that you will “lose” the lane, it’s just a question of “by how much”. Against skilled opponents it’s considered a victory if you can block their pull camp (making sure their supports stay underlevelled) and you manage to eke out some levels in the off-lane before moving into the midgame. The offlane solo lane may not be viable depending on the number of disables on the enemy team - it’s not uncommon for the offlane solo to abandon their lane and find farm in the jungle or try to be useful elsewhere.

2x Supports (Babysitter / Initiator / Jungler / Roamer)
(Get low experience, low gold, and are the ones usually doing the ganking)

Supports are a “catch all” term for the other heroes in the game without a defined role and lane. There are hard supports like Crystal Maiden or Shadow Demon who can perform well even when behind on levels and farms relative to their opponents, and will typically be tasked with ganking, warding and pulling / stacking the jungle. Stacking deserves a mention here - neutral creeps respawn every 1 minute if their spawn box is empty. By attacking and dragging the creeps away from their spawn box at the minute marks, you can create larger and larger creep camps. These can be used to deny entire waves of your own creeps by pulling them to your friendly creeps, or you can simply grow the stack size for your carry to farm later if he has an AOE skill. On the other hand, pulling creeps into an unstacked camp will not kill your creeps, but only delay them while you kill the neutral camp - this creep wave will now join up with the next upcoming wave, so now you have two waves of creeps marching together towards the enemy tower at once. This is an effective way of making a strong push on the enemy tower. Support heroes are very active in the early game moving around the map and exert a great deal of influence on how safely their carries can farm, how safely their team pushes, and how well your team is protected in ganks (supports are typically the ones to teleport in to stop a gank). There are also initiator supports like Tidehunter or Sand King who need early farm to get a blink dagger for initiation. There are also jungling supports like Enchantress or Chen, who primarily gain gold in the jungle but can exert pressure on the adjacent lanes by ganking it repeatedly. And there are Roamer supports like Vengeful or Spiritbreaker, but they have currently fallen out of favour in 6.78 (though they may be making a comeback in 6.79).

Complete Summary

So this is a handy chart I've made - personal opinion only and based on what I've seen work in competitive games - of which heroes tend to play which role. Click through to see the full version - I've added colored sections to indicate heroes which can play multiple roles. (direct link) Inevitably the placement of some heroes can be controversial, for example I've labelled Sven and Kunkka as a semi-carry when they are able to one-shot the entire enemy team late-game with sufficient farm which is more than most hard carries can achieve... and Slark is in the wrong place, he's a Semi-carry, not a hard carry....I'll probably update the chart as time goes by.

Typical laning compositions

Within the constraints of the mechanics described above several popular laning setups have come to be in popular use. (notation is x-y-z from Radiant side corresponding to hard lane - mid lane - easy land).

2-1-2 (Balanced)

This splits the farm up most equally, and your team plans to contest all 3 lanes. This is the most common “non-strategy” used in the majority of lower skilled public games. The safelane support can be a jungler, turning it into a 2-1-1 plus a jungler.

1-1-3 (Defensive Trilane)

2 supports and a carry on your safelane virtually guarantee your carry gets every single last hit, however, the opponent’s carry in their safely will likely be farming unopposed as well. Your two supports can leave via your jungle to gank mid once you’ve forced the enemy offlaner away, also putting the enemy mid in constant danger. This is the most common “non-strategy” used in the majority of high skilled public games. This setup is very flexible in terms of junglers - a single safelane support can also be a jungler, and the offlaner can also abandon their lane if it’s too dangerous and transitions into a jungler as well, making it a 0-1-2 setup with two junglers. A 1-1-3 Defensive Trilane can transition into a 1-1-1 Tricore lineup if the enemy abandons their offlane, freeing up the two supports to jungle (to boost their exp and gold gain) or roam at will.

3-1-1 (Offensive Trilane)

2 supports and a semi-carry head to your offlane to challenge the enemy carry and deny their farm. This is an aggressive and early game oriented composition that is an asymmetrical counter to the Defensive Trilane - the enemy will be picking a “safe” hero with typically poor scaling to go in their offlane, which you will beat 1v1 with a greedier semicarry solo like Clinkz or Weaver, and your offensive Trilane setup should beat their Defensive Trilane as your semi-carry in the hardlane is active from the start while their carry only comes online 20 minutes in. Also, if the enemy does not expect this, they may mistakenly ward your jungle (to stop the pulls and hinder your jungler) when in fact most of your team is not going to be there anyway. The major downside to this strategy is that your trilane will usually end up underlevelled - all 3 of you will just be relying on lane creeps and hero kills in your hard lane, while if you were in your safelane you could pull and jungle more safely to supplement your income. On the other hand, ganking mid is much easier from the offlane due to more favourable terrain. This strategy usually results in shorter games, and is seen in very high skill pub games where one team ends up with a draft that cannot compete with the other in the lategame, so they try to force an early game win like this.

2-2-1 (Dual Mid)

This used to be the default pub line up many years ago but has fallen out of favour. Dual mid ensures that you shut down the enemy mid hero entirely and dominate the runes, while your safe lane is run solo as it has the advantage of terrain. Some hero drafts don’t have a natural midlaner and this might be the optimal laning choice for your team. The additional pressure your team puts on the enemy mid can hinder his gold and exp gain to the point he becomes irrelevant to the game. Some hero compositions force the team to run this role, for example when you need to gain levels on a critical support hero (like Wisp) and a normal trilane type composition will not make sense.