Friday, January 10, 2014

Necrophos - Change the name, Change the game!

Nearly 3 months after the 6.79 patch was released, the competitive metagame is still very open. As most players and casters note, this is a very positive sign and suggests either a very healthy version or at least one which is very difficult to figure out. It's an exciting time to think about the game because whoever has the right ideas first will have a significant lead over those that only later come to discover how best to play this version. Or perhaps this will be a version which never settles on any particular playstyle. Different heroes are first-picks from one week to the next at the moment and the current patch is challenging all teams to be able to adapt quickly and efficiently during any given draft. Never before has knowledge of the current metagame been such a weak determinant of how to out-draft an opponent. Some have even begun to criticize teams like Alliance for their inability to adapt - a criticism which I feel is unduly harsh, but does perhaps explain their unexpected last place at the D2L LAN finals this past week.

That said, there are at least some trends that are emerging, and some patterns worth looking at. Many of these are very obvious. Heroes that have begun to pop up as heavily favoured by various competitive teams include Doombringer, Alchemist, Luna, Dragon Knight and Pugna. That said, these are just a handful and different teams are showing different preferences which themselves have changed over these last 3 months. When the patch initially came out, it was clear that the draft was all about the supports. 4 top supports quickly emerged in the form of Crystal Maiden, Visage, Venomancer and Lich. However, the rise of Lich was short-lived, probably due to the ease with which teams can push in this version, thus countering both Sacrifice and Chain Frost and rendering the hero very ineffective. In Lich's place, Alchemist has subsequently cemented its spot - although Alchemist is not always played as a support, it is played often enough as one to be included in the 'big 4'.

But which heroes have been winning the most? I decided to take a look at the data on datdota and see which heroes had the highest win percentages in 6.79 so far. What I found was rather interesting: 

So, of the top 10 highest win rate heroes, 4 of them are Wisp and her friends. Okay, cool, Wisp wasn't all about Tether. Relocate is an excellent spell. Perhaps interesting is that despite Wisp's increased win rate [from 56% in 6.78 to 67% so far in 6.79], the hero is being picked substantially less than before. This might be because teams have been slow to realize that the hero is still very effective or perhaps it's because the hero doesn't lane as well any more and is thus just harder to fit into a draft. Either way, it remains outside of the 'big four' supports in terms of popularity, despite its having a far superior win rate to any of those 4 heroes.

Meanwhile, Tusk and Warlock are in the top 10 due to the insanely small sample size, having only won 1 more game than they've lost, respectively.

Elder Titan holding his ground near the top suggests that nerfing anything except Natural Order will not be able to reduce the dramatic effect the hero has. For me this means the nerfs have been vindicated since the hero shouldn't be able to slaughter lanes if it's good enough to pick without having that ability.

Beastmaster is a hero that seems to come and go and is perhaps at the moment coming rather than going. It's hard to say why its win rate is around 60% at the moment when its overall usual win rate sits around 50%. Is it becoming a thing again? I would need to think a lot more about this to make any meaningful comment here. Honestly, I don't know.

Perhaps one of the biggest surprises is that Treant Protector, despite a substantial nerf to Living Armour, has more or less retained the same high win rate it had before. The 6.79 sample is quite small, but I'm inclined to believe it's representative because there is a good qualitative explanation for it. While Living Armour got nerfed, Leech Seed got massively buffed. The hero is just played differently now but has managed to stay very effective. 

Necrophos is actually the original inspiration for me writing this piece as my investigation into 6.79 hero trends was ultimately prompted by an observation I made about this hero. I did not, however, expect it to have the highest 6.79 win percentage. That said, the sample size here is very small and thus potentially misleading. Indeed, what I observed suggests that the hero might only be being drafted into positions where the very fact of its having been drafted near guarantees it to win!

A few professional teams [though mostly LGD] have begun to run this hero as a standard response to certain types of picks. But it's not the oldschool 'pick Necrolyte to counter a tank' idea. Put simply, if a team realizes that their opponent's first few picks in the draft are not capable of dealing with early 5-man pushes, they can start building towards a strategy that will ultimately centre around Necrophos. Normally the Necrophos itself is picked last but this, I think, is mostly to stretch the advantage as far as possible - not a necessity. And as teams begin to realize this and start to ban the hero in the second or last ban phase, it might just start getting picked a bit earlier in drafts. To demonstrate the kind of draft it fits into, here is an example. 

This is how the draft looked in game 1 of iG vs DK in the grand finals of WPC ACE. Despite DK starting off exceptionally well with a 5-2 lead before creeps spawned, they were really nowhere in this game. Burning got his farm, and the lane stage looked okay for DK, their 5-2 lead becoming an 8-4 one. But Hao's Necrophos got good farm too and, shortly after finishing his Mek, it became very clear that only one team could win this game. Storm Spirit, which was the 3rd pick for DK in this game, probably sealed the deal for iG in terms of building towards a Necrophos pick. The hero is only marginally more useful than a Lich at defending against a 5-man push early on. Storm Spirit thrives on its mobility, which allows it to pick heroes off or abusively kite in skirmishes. However, if the entire enemy team is holding hands from early on, you aren't going to be making any pickoffs without becoming one yourself. Meanwhile, Lifestealer needs lots of items to be able to handle that type of approach - and even if the game had lasted that long, Burning would still have had to deal with the threat of heavy disable, always at risk of being chain-stunned and burst down if caught by anything.

Initially I thought iG had just figured DK out well during this draft and out-picked them excellently. After a while I realized just how important the Necrophos was to their strategy. It's essentially the fuel that keeps a 5-man push strategy chugging along. Notice the only direct 'pushing skills' iG had here are DK's Dragon Form and maybe Visage's Familiars. Their ability to pressure buildings is more about being too threatening to approach than it is about actually killing the buildings really fast. And that's what Necrophos enables. In a way, this is reminiscent of the role Nightstalker used to play in push strategies - although what he did was more chase enemies away from defending than prevent them from trying to defend just by being there. In the case of Nightstalker, a fight could still go badly if the other team initiated well. In this case, it's hard to see how that would happen. So yeah, this game provides a clear lesson that leaving your draft open to a Necrophos 5-man assault will lose you a game very easily.

Sadly, Fnatic's Fly did not make this observation and we saw an almost exact replica of the above draft in game 1 of LGD vs Fnatic at D2L. 

In this case, Fnatic also picked Storm and Naix in 3rd and 4th picks [though the reverse order to DK] but what they also did was pick Slark before either of the two. Slark is definitely less useful against a push than the Clockwerk that DK had above and Shadow Demon is probably mildly less useful than Rubick here. Meanwhile, in place of iG's Nyx, LGD ran a Venomancer. This makes them slightly weaker in terms of disable but substantially stronger in terms of their ability to force down buildings without giving Fnatic any option to defend them. This game was even more one-sided than the iG vs DK game, perhaps because Fnatic did not pull off any pre-creep heroics in this game.

Presumably, most professional teams will have observed at least one of these cases and will be a bit more ready for this sort of thing now. That said, it remains to be seen whether or not the Necrophos needs to be picked 5th in order to make its strategies work. Being picked last, in the correct draft, it seems very near to unstoppable. Perhaps if it is picked a bit earlier, but also in the correct draft, it will merely be, I don't know, extremely good? 

I'd like to stop and make a point about the sample size here. Hopefully having done substantial qualitative analysis will already work against concerns about a small sample size. That said, I want to point out that while Necrophos's insanely high win rate of 78% in 6.79 is only based on 23 appearances in 3 months, the hero's win percentage in 6.78 was a mere 28%. Meanwhile, looking at the overall stats of the hero on datdota, it's clear that this is its first spell of consistent success in a long time.

This means that Necrophos was 9-26 before 6.79 and is 18-5 in 6.79. While both samples remain reasonably small, the comparison between them is undeniable. The hero almost never won and now it almost never loses. How deep this trend will run remains to be seen.

* * *

One final thing that I would feel intellectually dishonest about leaving out is that while writing this piece I discovered that, in fact, Rubick is the 3rd most picked support in 6.79, ahead of both Venomancer and Alchemist. That said, I still feel disinclined to include it in 'the big 4' simply because I don't think it's being prioritized in the same way as those 4 are. Which is to say, Rubick slots into any draft nicely, but most drafts seem to want a Crystal Maiden, Venomancer, Alchemist or Visage on support. 

Saturday, January 4, 2014

draft is hard

This evening, I came across an interview with 7ckingmad on the subject of drafting. It struck me as interesting that despite the fact that reasonable questions were asked and reasonable answers were given, I didn't feel like I came out of reading the interview any better equipped to draft well. I also don't think this is because I'm a genius at drafting or that it's because everything said was obvious. Rather, what this interview demonstrates is that when it comes to drafting, there's never a simple or straightforward answer. While it may seem evasive to say things like 'it depends' when asked if something is a good idea or not, the reality is that drafting is extremely complex and completely situational and thus this answer is probably the most appropriate answer that can be given to any question about drafting.

For the same reasons that the interview wasn't ultimately that instructive, this post will probably fail to be as well. However, what I plan to do here is attempt to outline a handful of things that 'it' can 'depend on' and thus hopefully illuminate some of these complexities, while emphasizing just how complicated drafting really is.

1. Having a Plan

As 7ckingmad suggests, it's generally not going to happen that you are able to have a complete 5-hero plan going into a draft and be able to stick to it entirely. In fact, in most cases it would just be a mistake to try force a strategy and mostly a matter of luck whether it works out or not in a particular game. That said, it is critical that you do have some plan. It's even possible to only develop this plan once you have all 5 heroes. How? Well, what I mean by a 'plan' is an idea of how your team is going to win the game with the heroes you have against the heroes they have. It's very helpful to try and construct a narrative in your head as to how you'd like the game to run. Of course here you need to be as objective and impartial as possible - it won't be helpful to assume all your lanes will win just because it makes it easy to imagine how the game will progress in your favour if they do. If a lane match-up looks like it's going to win you can treat it as such and vice versa.

Now, while I've noted above that it's possible to develop a plan post-drafting, for obvious reasons it's a lot easier to design your draft around a plan than to design a plan around your draft. To some extent this will always be a reflective equilibrium as your picks inform your plan and your plan informs your picks - in fact, during the game, it's very likely your plan might have to change if the situation changes. That said, in all likelihood, to end up with a solid draft you're going to want to start designing a plan at some point during the draft, whether it's before the first pick or before the last pick. Having a plan not only helps to ensure that your draft remains coherent despite having to adapt to the enemy's draft, but it also helps each member of your team to understand what their role might be at any given point in time during a game. If you want to control the game via split-pushing, then the split-pushing heroes will know that this is what they need to be doing. If you planned to get an early advantage via good rotations from supports, then the supports will know what to do. In contrast, if your plan is to play passively and only counter-gang, this gives the supports equally useful, yet opposite, guidance. 

So having a plan can be very useful and is generally important when it comes to drafting. Here's the thing though - because of the importance of having a plan, at any given point during a draft, a particular pick or ban might be better [or worse] for you than it usually is, where all else is equal. Your plan creates a context within which picks and bans need to be evaluated. Thus, it's nearly impossible to evaluate somebody else's draft properly without having insights into what their plan is at least until the entire draft is over - and often even then you need to see how the draft plays out to understand whether or not the picks were the best ones they could be. What's more is that sometimes poor play can skew even this kind of evaluation - warping our analysis of a draft because of how we saw it work out. 

2. Counter-picking

You know who else will have a plan? The team you're drafting against! What does this mean? Well, both drafters will be constructing narratives in their minds - stories of how their draft leads them to victory - during the draft. As a result, both drafters will need to try imagine what the other drafter is imagining and incorporate it into their preferred narrative. So we were planning to 5-man push early on. It looks like they might try to hold us outside their base while they split-push. Well this means we need one of three things - a way to force an initiation at their base, pushing power so strong that we cannot be held back, or a way to control their split-pushing either before we push or during our push. Naturally, your opponent will try to anticipate this thought process as well and again be one step ahead. This thought process can be recursive and, to be honest, this can be very damaging to a draft. If a drafter spends too much time during the draft thinking the kind of 'they know that we know that they know that we know but they don't know that we know that they know that we know that they know that we know' thoughts this can result in the draft becoming far too abstracted at points and also far too dependant on assumptions about the other team's plan which may or may not conform to your expectations. 

Instead of trying to get into the enemy drafter's head in terms of their plan, you might think that counter-picking at the level of heroes is more appropriate. So lets not try to fool ourselves into thinking we know exactly what they're trying to do here but we do know that they have hero X and we know that hero P is good against hero X so perhaps we should consider picking hero P. Unfortunately, this level of counter-picking, although less abstract, is not without its own risks. It's not enough that a hero can theoretically interact well with one enemy hero. That is not a good enough argument for picking it. How well does this hero interact with the other enemy heroes? How well does is synergize with your other heroes? Does it fit into the plan? Are you making assumptions about how the enemy hero you are trying to counter will be played? Imagine the enemy team first picks a Mirana and you decide to counter-pick against it. Perhaps your concern is that Mirana is a very difficult hero to kill because of her leap and thus presents a unique threat when farmed. But what if the other team decides to use her as a support that game? Will your counter-picking her on that basis make sense then? Did you just pick a Storm Spirit because it's good at killing one of their supports?

So what am I trying to say here? What I'm not doing is arguing against counter-picking. The point is just that counter-picking can be very dangerous. Your drafting should be sensitive to the enemy team's draft, both in terms of what their plan is and in terms of individual heroes picked. However, it also shouldn't be too concerned with either of these things, much the same way it's dangerous to force a particular plan going into a draft. What we have here is further context for a draft. Not only is the value of a pick or ban relative to what your plan is, it's also relative to what the opposing team's plan is, what their picks are, and what your picks are so far. 

3. Knowing your players

This one I expect is a bit more controversial as a lot of people think that it's a weakness of a team if their players aren't capable of playing all heroes at the highest level. That said, the reality is that there is not a single team in the world for whom this is the case. So, while we're in the world we're in, lets talk about hero-to-player allocations. The brute fact is that each player is better at some heroes than others. Moreover, players will tend to do better in particular play-styles than others and might excel more in some situations than others. So if your plan involves making the guy who doesn't cope well without farm handle a very difficult lane, then your plan is probably going to fail. Now this doesn't mean that a drafter must be completely at the mercy of the preferences of his or her teammates nor does it mean he or she should only draft them what they're best at. There is always room to grow and improve and the only way to do this is with practice. Obviously, this is a two-way motion. Drafts outside of a player's comfort zone are the only thing that will gradually make them more comfortable with more things but drafts within their comfort zones will, in the short term, usually produce better results. So there is a trade-off. 

That said, even when looking at the best teams in the world, it's evident that different players are better at different things and thus different teams are better at different things. If it's not obvious, what this means is that depending on who you are, the value of banning or picking certain heroes changes. Take a minute and process this. This is one of the big reasons that teams that manage to define metagames usually succeed more than those who fall into them. If the metagame becomes a certain way because a team is doing really well with certain types of drafts, unless your team is also really good at using those kinds of drafts, it's in your interest to try your best to shift the metagame. Obviously, sometimes the dominant metagame is based on how the game is, and not just what's being preferred at the moment - in which case your team must adapt. But this is certainly not always the case. 

4. Timing

Depending on what stage of the draft you are at, the same pick or ban might be excellent or awful. Normally more versatile heroes are good early on because they don't give away much information about your draft. Meanwhile, heroes which can have a dramatic effect on the game provided the other team hasn't planned to deal with them can often be very powerful picks late in the draft. Enigma, for example, is usually picked late in a draft because the hero operates in a very specific way. It's going to probably jungle. It's going to push things. It's going to try cast Black Holes. Okay, well this is a pretty transparent pick. What this means is that it's easier to counter-pick and thus much more valuable to pick at a point where it's too late for the other team to do that.

Timing also has to do with roles. The later you can go into the draft without the other team knowing which heroes will be performing which roles, the more advantage you're going to have over them. Alternatively, first-picking a hero that normally signifies a particular thing, you can dictate a lot of the opposing team's draft, or at least put a lot of pressure on them to deal with said thing - even if you've come up with a different way to use that hero. 

The most obvious timing-related consideration is that heroes that you expect both teams to be going for are usually better picks earlier on for the simple reason that later on they might not be there anymore. I've left this until last because I think people over-obsess about this point. Very often you hear casters asking why a team has picked a certain hero so early in a draft because they feel the hero would have still been available later on. While it is true that one reason to pick a hero early on is that you are worried that a hero you want to get will be snatched by the other team, this is not the only relevant factor. In fact, everything discussed above is an argument against that. So yeah, absolutely, if you know both teams really want a hero and you've decided you must have it, by all means pick it early. But other considerations can trump this. It might be that your team is extremely confident when executing strategies that use Vengeful Spirit. Maybe you have figured out a way to use the hero that other teams have not yet. This could very well justify first-picking the hero, despite the fact that it's very rarely picked early in a draft. After all, heroes that are described as 'first picks' weren't always described that way. Someone has to be the first to recognize the strength of a hero - or the strength of that hero for their team. 

Returning to the point - it is important to think about the timing of your picks. The same five picks chosen in various different orders can result in drastically different responses from your opponents or might affect your ability to even get those five picks. 

* * *

So I've discussed 4 factors that are relevant to drafting which make drafting very difficult by creating a context against which all picks and bans must be evaluated. Sadly, though, the context within which any pick or ban takes place in any game of Dota is several times more complex than just these 4 factors. It's not possible for me to discuss every factor that could be potentially relevant at any point in time in any draft and, to be honest, it wouldn't be that helpful even if it was. Part of this is just the harsh reality that, as with Dota in general, there is simply no substitute for practice. If you want to get better at drafting you need to draft a lot. Some very good drafters might struggle to string a sentence together justifying their picks and the most flawlessly advocated picks might, ultimately, not be a good idea for reasons our Dota lexicon cannot even express yet.

Drafting is really, really hard.