Daily Oracle
Блог

Meet Lea: the statsman invited to TI10 Panel, who created Spectral.gg

There are two masterminds of in-depth statistics in Dota – Noxville and Leamare. Noxville is well-known for his database website DatDota and keeps us informed of competitive records on his twitter.

Leamare, on the other hand, has only just started getting public attention. Their personal website — Spectral.gg — is a place where you can find a lot of data and reports about tournaments and games at the Immortal rank. They were working with PuckChamp last season and it was recently announced that Lea will go to TI10 as a part of the Broadcast Talent crew!

What motivates them to work with statistics for free? What’s inside the mind of the greatest stat geek in CIS Dota? What makes South America so good and how a misjudgment of Yaroslav “NS” Kuznetsov dragged him into the world of data?

Roman Galeev spoke with Lea to get answers to all those questions.

* * *

– Congratulations on your invitation on The International 10! It was a pleasant surprise for me. And you?

– It was kind of a funny story. I didn’t pay attention to the mail at first since didn’t really expect to get anything like that. I was on vacation at the moment so I didn’t really pay attention and removed it.

– So, how did you finally get informed about your invite?

– Another person reached me out directly in discord and asked: “Are you ignoring your mail? Can you check it?”

– What would be your responsibilities at TI?

– You won’t see me on the screen. I will be providing stats and other historical data to the broadcast. You might notice that I have posted many tweets about previous TIs. 

– How much time did it take you to collect all the data from TI1-9?

– It took about a week to collect data from TI1, just because there weren’t many games played to begin with. And I was working on all this even before the invite anyway. 

To be fair, it was hard enough to find all the data from TI1 and TI2. For example, at TI1 players didn’t even use their own accounts – it seems like they used special ones given to them by Valve. And there was no recorded information about drafting order at TI1 and TI2, thus I had to edit it manually.

The older TI is, the less data you have about it. I think at the moment it’s safe to say that I am the only person in the world with the complete draft stats for TI1 and TI2.

– Do you have an explanation why a Russian statistician was invited to the English-speaking studio instead of Russian?

– No, I have not! (laugh). I guess many people from CIS wondered why I was invited. I’ve started making various types of content – translation of patches, guides for new heroes – as far as 6 years ago, but it just so happens that the Russian-speaking community is not as aware of my work as the English-speaking one.

Do you want a crazy story? Once upon a time I wrote an article in my VK blog [Sports: a Russian social network]. Two days later I translated it into English and posted it on Reddit. The following day a Russian esports media took it from Reddit, translated and made a title: “A redditor made…”. It looked like nobody was even considering the possibility of a Russian-speaking person doing something like this.

Unfoutunately, we don’t have a content hub similar to Reddit is here in CIS. 

I tried to be a caster at some point, a panel-like analyst, a coach for teams, but none of it was working for me or the audience..

– If you had a choice between going to TI with the English-speaking crew or to be with the Russian crew, what would you choose?

– I don’t think there’s a demand for such content in the Russian-speaking community, at least not yet… On the other hand, I feel more comfortable with the English-speaking community and people, and it’s kind of strange to me! So, I would not change anything at all. The only problem is the need to cooperate for a flight, preparing documents and so on –  it becomes harder when you don’t really have anybody you know to ask about this kind of stuff.

Lea-statsman. And software developer for teams

– Did you have any experience with gathering data for esports before Dota?

– No, I hadn’t. Before Dota I only played Sonic the Hedgehog titles and competitive Unreal Tournament. When I got into Dota I found out about Dotabuff and was so excited that it hooked me.

– In Dota I only know about 2 full statisticians – you and Noxville. Any other stat lovers coming to your mind?

–  In CIS there are Dota Science, StatsByLeo and a couple of well-known people too. And in the international community there were, obviously, Nahaz and Scant. But in CIS specifically I haven’t really seen many people working with stats in Dota.

– How many Dota teams have their own statistician?

– It depends on the region, I guess. There are many teams without a budget so the work of a statistician is usually done by a coach or even a manager. It is how PuckChamp functioned before. I wrote to them: “I have a deal. I will provide data for you in exchange for your feedback”. I saw a lot of potential in those players and decided to help them out.

There are teams with some custom software for stats. When I worked for Winstrike, it was extremely difficult to get in touch with the players. I’m not even sure they are aware of my existence! So, it was easier for me to make software to work with the players through and integrate my expertise into it. It became easier to work in this regard when the latest Winstrike squad was signed (they were Modus Unity before). This is how I’ve met the PuckChamp guys, and I’m still helping out Winstrike as a team sometimes, as a friend.

Many European teams went this route as well, but instead of developing their own solution, they often buy it from somebody else (or rent it). 

As far as I know, Liquid uses SAP, some teams prefer Shadow. In particular, some teams from South America approached either me directly or Winstrike for the software I was working on. So it is free to say that I was the analyst for many teams, but indirectly.

CIS players don’t appreciate data that much, but in other regions teams are know just how valuable stats can be.

– Esports stats software is kind of a mystery. Everybody knows it exists, but nobody knows the details.

– I can’t really show exactly what it looks like, but in one way or another features from it appeared on my personal website or were used in my blog posts.

What would it be? First, a short summary about every team: best counter-picks, their signature heroes and most successful drafts. How they move across the map, where and when they use smokes. Second, some more in-depth data like:

  • What is the resource distribution between five players?
  • What type of money (from lane creeps, from kills etc) is prevalent for a certain player?

This kind of software helps to sum it up in one place.

And last, but not least: the software can tell you which heroes were picked by opponents in matchmaking. Of course, nearly all players have their accounts closed, so we need to collect all the matches played by the potential rivals. It’s somewhat of a dirty trick since it pretty much ignores the privacy of the players, but in a way it’s an arms race. If your opponents might use something like that against you, regardless if your stats are public or not, then you need to be able to use it as well.

– It is time for a stupid question from non-tech guy. Is it really hard to write this kind of software you have just described?

– It is not so hard, I guess, but the devil is in the details. You need to know where you can find the data of matches and how to gather it. There are numerous approaches to it, but it is more about side-questions. If you are clever enough to write this kind of app, you will resolve these minor problems, which simply requires time and effort.

Lea as esports analyst. What’s wrong with some coaches in CIS and why South America – an ideal?

– Would you share what you’re doing for PuckChamp?

– My main role is to analyze data of opponents and some tier one teams (to improve PC’s playstyle), but I must underline that I don’t work for Puckchamp full-time and I don’t get anything out of it. It is more like “If you need something – pm me” and sometimes they ask to help them out. There were cases when PuckChamp requested to analyze their own playstyle.

– What player is reaching out for you to request the analysis?

– Usually it is Zhanna (laughing) [Sports.ru: Zhanna – manager of PuckChamp]. Players are not so communicative at all, but Zhanna is rather close with them and knows exactly what they need. But I also know that Dukalis has been keen on warding ever since Winstrike: he told me back then what additional data I should add.

– Is it possible for them to come to you and request advice on a draft?

– They don’t come to me with that, but, in a way, I come to them. When the players get my data, they see my “advice” on draft, but it is based on math and stats. It is up to them to use  it or not.

– Is there a type of information that seems very important for us, viewers, and doesn’t make much sense for a real analysis?

– It is hard to answer this kind of question from the get-go. I think that the draft order and the opponent’s draft are overestimated. It mainly comes from players’ preferences and habits. But then again, it is a really hard question.

– Take KDA as an example.

– KDA, damage… We look at things differently. I don’t think KDA and damage mean as much to professional players.

There is an interesting concept that I’m interested in, but can’t check right now if it’s viable or not. I think average values for heroes, including KDA or damage numbers,  can be used for calculating a chance of victory. For example, Stratz is using a neural network based on drafts data, and based on draft as an input it generates the answer. But you need to repeat the learning process every patch because of that.

I hope using averages to weigh the drafts and determine the winner without looking at the heroes. But it is just an abstract idea right now and I need to check it.

– Have you ever faced a problem when one of your teammates didn’t understand why he needed your numbers?

– A lot! There was a coach who’d opened my summary on opponents just 5 minutes before the game! Of course, they didn’t take data into consideration and the team was eliminated from the tournament almost immediately after that. One could simply open the summary and see the exact same heroes the opponents picked and banned. They simply just went for their most comfortable draft.  

One coach said to me: “We are not playing dota for numbers”. It got to me, but not as much as it backfired his own team in the end. Some players are very conservative and it’s tough to explain to them why these “numbers” are important.

Imagine a professional player. He comes to the team and says: “Shadow Fiend is soooo broken!” Then you check stats: SF has 30% winrate in the current patch, while the team has 0% winrate with SF. But it doesn’t matter to the player, he whispers to the captain during the draft: «Pick SF or I’ll break my items». Finally, he gets his SF and is like: “What a hero! BOOM – 3 razes in a row, WOW – what a highlight I have just made! Oh, we lost? It is because of my team”. I am exaggerating, of course, but you’ve got the idea.

South America is in a way a perfect region. Its players are a mix of CIS-recklessness and responsible attitude to data. I have experience with some of SA teams and they were the most engaged players I have ever worked with. It’s sad to know many of SA teams can’t really go for a statistician though..

– It is really a hot take!

– I know, stereotypically you think about South America and imagine carnivals, one leg-plastic chairs in an internet cafe and bad dota players, but it’s all wrong. If South American players had more resources at their disposal, it would be the strongest region in Dota because of its attitude to the game and its way of practicing for tournaments.

– How do you communicate with PuckChamp?

– There was a time when I gave all my data to Artstyle, but it turned out that it’s better to give it to Zhanna. Sometimes I watch their practice games and look out for them on Discord. I am like a spectre behind them. And I know that Zhanna asks for some data to use in her manager’s stuff.

– Since Artstyle left for NAVI, the place of PuckChamp’ coach is empty. Have you ever thought of being a coach of PuckChamp?

– I tried to be a coach earlier and discovered two important things:

  1. I have never wanted to be a professional player. You need to practice a lot and be in the optimal mental and physical form. Sometimes I get in a bad mood… Don’t get me wrong, I am not talking about tilt. But my performance during these kinds of days is much worse, because I don’t think about the game as much and can’t focus.
  2. It is quite challenging for me to communicate with people at all, while communication is the most essential part of being a coach. You have to set players up and be an authority to them. I am not your life and soul of the party. I can act as  a personal coach. But then again, you need to spend a lot of time to be able to coach someone. My number of games per week has been decreasing for 2 years.

At best, I could aim for being an analyst at a panel.

– So, if PuckChamp players will offer you to be their coach tomorrow, will you decline it?

– Yes. I think I will do more harm than good.

– Apart from PuckChamp and Winstrike, you’ve worked for 2 months for Fnatic. Could you tell me about that experience?

– It’s quite simple. Fnatic’s analyst at the time reached out to me just before Kuala Lumpur Major: “Hi, are you interested in helping us out with stats?” I agreed and started working on my own software for collecting data.

It didn’t really last that long though. Not because they weren’t satisfied, as far as I’m aware.It’s just hard to work with a person from a different region, who can’t really join you on a bootcamp even. Aside from that, I was working with the analyst most of the time, and he left the team for personal reasons. Once he was gone, I was gone too.

Lea and Spectral.gg: started because of a talent’s mistake. Makes almost no money from Spectral.gg

– Let’s talk about your site, Spectral.gg. How much time per week do you spend on it?

– It required plenty of time at first, but now I have made some of the processes automated, so I can spend more time on research. I interfere only when something unusual happens, such as cases when a tournament’s ticket is changed or when there are two divisions in one tournament, like it was in EPIC League.

Nowadays I work with some new ideas. I would sit down, look at a wall and think: “Hmm, it would be great if you could see to what extent winrate of a hero changes with a player’s experience on it!” I cannot work on anything else until that idea gets out of my head.

I started to collect stats because of TI6. Before that I would post some stuff in my VK blog and that’s about it. And at the time I was doing my own “live text cast” of the event while watching the Russian broadcast, when Yaroslav “NS” Kuznetsov said some random numbers, comparing Void safelane and Void offlane. His words were misleading, so I got mad and started to collect stats to prove him wrong. In the end my urge was satisfied and I found joy in working with data, so I decided to continue.

– Spectral.gg has zero advertisements and the only way to make money from it is donations. How much money have you got from it?

– It is public info you can check on the “Support” page. [Sports.ru: according to Spectral.gg, Leamare received $111 for the last 30 days].

– Do you have any ideas on how to monetize it?

– I have an idea, but don’t really want to talk about it since I am working on it right now. But generally speaking I have a principle: all data must be open and available. It seems wrong for me to hide some data behind a paywall. I can only think of hiding some specific filters that would only be used by a few people anyway.

I placed ads on the site once, just to check it out. I’ve earned like 1 USD from it for 2 months. It made it clear that this kind of stuff is not worth it, since it will make the experience worse for others, while also distracting people and not bringing much to the table to justify it.

And I don’t really want to deal with bookmakers.

– Did bookmakers come to you?

– There were some strange people, but no straight offers, really. Well, except one legitimate one, but I kindly declined.

Lea-player: a huge fan of Spectre (What a surprise!), that trust in new Venomancer

– How many games do you play Dota per week?

– It was 2-4 pubs per day 2 years ago. During OMEGA League, when I was working with Cap, I played around 1 game per day. I’ve been working almost all the time for the past 6 months, so it got to 3-4 matches per week. And recently I found enjoyment in playing Turbo to release some steam while not having time and energy for a “real” game. My peak MMR was sometime in 2017 – it was around 6800. Since then I didn’t really have any reason to focus on ranking, so these days I would rather play some full-stack party games instead of a solo queue. If there’s no party, I would rather work more on the site’s features.

– Imagine you suddenly feel an urge to play SF. The hero has a 30% win rate and you haven’t practiced with it for 3 years – it’s a straight way to lose. Will you pick it?

– Nowadays I don’t need to prove something to someone, so I don’t care about MMR at all. I don’t play a certain role, but instead I play certain heroes. Check my profile – it is full of Spectre games. My mind tells me: “The hero is trash”, but my body says: “Pick the waifu, play Spectre one more time, even as a support”. Just because I want to.

– All of my friends who like Spectre claim that the hero is dead. Can you explain why?

– There was a patch, when Blade Mail was reworked. New BM fitted Spectre perfectly, but it was a bit too broken. The nerfs must have been done, but Valve nerfed not only BM, but Spectre too. So, BM went from the state “It is good to buy after PT” to the state “You either rush it before boots or lose because of it”. It weakened Spectre in terms of possibilities and varieties.

Sure, there is an Echo Sabre trend created recently, but Desolate and Spectre’s talents were also changed, not in a good way for the hero. So, the hero is kind of weak right now. There is an option to start with Falcon Blade or Mage Slayer, but the hero itself would not become better. 

Moreover, there are too many tempo heroes in the meta, that have their power spikes much earlier, and the new Silver Edge is pretty strong too. Even experienced players of Spectre have only about 50% winrate. Inexperienced players – 39%. On average, winrate of heroes played by experienced players is nearly 54%. It just shows how hard it is to make Spectre work.

* * *

EDIT: After the original version in Russian was published, the new trend on Spectre with Aghanim has appeared. So I asked Lea to evaluate it.

– Yes, people started going for Aghs and Orchid. The trend isn’t really new, high MMR Spectre spammers already experimented with it about half a year ago. However, this new trend seems to have problems of its own. There is a reason why Spectre spammers stopped playing with it after all.

To begin with, Sabre was nerfed and it became more timing dependent – and it was a big reason why this trend found its way into the meta. The build was revived by Chinese pro players and later on by EternalEnVy, but as pros (and TI10 attendees especially) experimented with it, they stumbled across other issues.

Orchid doesn’t really synergise with the hero that well and feels a bit excessive. Coupled with Sabre being not as good anymore, they eventually started dropping this item from the build (unless the game is a stomp).

The other major problem is more complex: Manta + Spectre and Skadi + Spectre are two of the strongest item-hero combos in dota right now (and they were in the top-10 strongest until recently). This shows just how strong of a powerspike Manta is for Spectre, and the Aghs build forces you to delay Manta (especially since Aghs is quite timing dependent itself).

So here comes the general trend I noticed: a lot of players, after a while, converged towards a Falcon Blade build. It starts with Falcon Blade (instead of Sabre), then goes either Manta followed by Aghs, or vice versa. Then comes Skadi (which might be purchased before Aghs depending on the game, but never before Manta), and on rare occasions you might see Orchid. This trend continues to evolve, and you can see how it goes on Dota2ProTracker.

However, this build doesn’t solve any of Spectre’s problems. Before the trend Spectre had 47-48% winrate at Immortal, and after the trend she had 44-46% winrate with a similar pickrate. All the tempo problems are still there, as well as Silver Edge and high DPS burst. Winrate of experienced players didn’t get better either. Some people got overexcited with the new trend, saying «we will see Spectre dominating at TI», but I doubt it will be the case. It takes more than just a shift of a build and a cheap mana regen to raise her winrate.

* * *

– Is Tinker broken or trash?

– I think people don’t really understand how to play the hero at the moment. I am not a Tinker player and don’t give a clue how it should be. But I think it will take a while until a new playstyle emerges.

Many people started playing Tinker as a support, but I am really frustrated with their builds. They ignored Defense Matrix, even though it gives you HP after resists. So, if you cast it on a hero with high armor, it will absorb more damage. And you can buy Solar and cast it on allies! I saw 2B practice this build and it seemed very strong..

EDIT: Since the interview was recorded, people sure found a new way to play Tinker: Overwhelming Blink and Shiva’s Guard made the hero too powerful.

– Who is the hidden imba of 7.30c?

– Well, there are plenty of strong heroes, but none of them are “hidden”, really. I wonder if Venomancer can be that strong. I think it is possible to create something interesting with Veno, but I didn’t see enough Veno players to come to conclusions.

– What do you love more: Dota in statistics or statistics in Dota?

– The answer is, obviously, yes..

– What must happen in your career for you to say: “I have achieved in Dota what I wanted”? 

– First of all, to finish the project I always wanted to make. Then, to convince the wide audience that stats in Dota can be exciting. I wanted to go to TI one day, but I don’t have to worry about this one anymore. I guess only switching to Dota full-time remains!

– Do you have any other job than your job in Dota?

– Sure, otherwise I wouldn’t have any money for a living. By now, donations in Spectral.gg are the only way of making money in Dota for me. 

 

 

 

 

 

 

 

 

 

Комментарии

Возможно, ваш комментарий – оскорбительный. Будьте вежливы и соблюдайте правила
  • По дате
  • Лучшие
  • Актуальные