Interview between CIS popular player Jusha and Creative Director Andrey Beletsky. A lot of topics were discussed, such as the matchmaker, artillery, ± 1 tier MM of battles, game servers and more. It’s from 16 April.
The matchmaker remembers the tiers you played in recently, but does not remember the tank or your account in general. And it makes sure that you have the best possible distribution by tier.
As of yesterday morning, the queue structure in tanks was as follows (RU):
• 17% of people in the queue went to tier 10
• 17% of people in the queue went to tier 9
• 27% of people in queue to tier 8
• 12.5% of people in queue to tier 7
• 12.5% of people in queue to tier 6
Then comes a strong fall, and then the matchmaker begins to cope with justice.
• 6% of people in the line went to tier 5
• 3% of people in the line went to tier 4
• 3% of people in line went to tier 3
• 1.5% of people in line went to tier 2
• 1.5% of people in the line went to tier 1
- The day of the week / weather on the street / holidays and all other moments greatly affect the queue. Remember that.
As of yesterday morning, battles at tier 8 (excluding platoons and preferential tanks):
• Three-tier battle (10-9-8), where tier 8 was bottom tier – 19%
• Two-tier battle (9-8), where 8 was bottom tier – 24%
• One-tier battle (8), all tanks tier 8 – 31%
• (9-8-7), where 8 was middle tier – 8%
• (8-7), where 8 was top tier – 12%
• (8-7-6), where 8 was top tier – 6%
This is how it is distributed in reality.
About 66% is a single-tier battle / battle in the middle of the list / battle against tier IXs.
As of yesterday morning, battles at tier 6 (excluding platoons and preferential tanks):
• (8-7-6), where tier 6 was bottom tier – 20%
• (7-6), where tier 6 was bottom tier – 27%
• (6), all tier 6s – 29%
• (7-6-5), where tier 6 was in the middle – 9.5%
• (6-5), where tier 6 was top tier – 11.5%
• (6-5-4), where tier 6 was top tier – 3%
- Bots for tiers 1-5 (especially for tier 1) are the most adequate solution in terms of comfort. This decision will be able to guarantee a situation in which players play in comfort, no matter what happens with the lineup. We already have bots on small clusters in South America and Australia, but we don’t use them anywhere else. For two years, bots have been used there. At the moment, we do not seriously think about other clusters. At the moment, the situation is OK.
- There have been experiments with neural networks, there are no results yet. The problem of the neural network is that it cannot be debugged.
- The main reason is economic: bots are very expensive (they take up a lot of computing power) if we want them to be adequate. And therefore, just because of this, the inclusion of bots at low tiers for our main clusters is not planned.
- In 2019, we changed the hardware of our servers, yes, sometimes you could have problems through our fault.
- One WoT server holds up to 130,000 players; in total, at the capacities that we have now, we can hold 900-950k players. This does not mean that it will work flawlessly – it just means that nothing will fail and crash.
Servers may vary in power. And also, if players begin to break simultaneously into one specific new mode, for example, event mode, we have big problems. And separate server capacities have nothing to do with it.
±1 matchmaking will make the current situation only worse in several respects:
- The variety of battles will become worse: the smaller the volume of tanks that can potentially get into battles, the more repetitive and of the same type they will become. ± 4 in some ideal world would even be funnier.
- A game is a way to learn and a way to generate emotions; there are no positive emotions without negative ones.
- The situation with ± 1 for some tiers will make it many times worse, and not better. And I’m not talking about queue times.
There is currently no reason to believe that ± 1 will be of benefit. But there are many reasons to believe that this will be harmful.
- Nation change for Rudy is not super popular among players. So far it does not look like this is something that we are ready to spend efforts on. Technology is there, so in theory, any tank can do this now.
- We attempt that all 3D styles don’t affect the perception of the collision model. Well, in general, everything that affects your decision where to shoot to score a penetration.
In 11% (and this is a lot) of battles a certain player plays arty in all the battles, that all players play in a day.
- For self-propelled guns:
♦ The most popular tier with artillerymen: 6 (15% of all arties)
♦ Second most popular tier among artillerymen: 7 (14% of all arties)
♦ Third most popular tier among artillerymen: 10 (12% of all arties)
♦ The most unpopular tier among artillerymen: 8 (7% of all arties)
• Three-tier battle (10-9-8), 3 self-propelled guns are found – 15.32%
• Three-tier battle (10-9-8), 2 self-propelled guns are found – 26%
• Three-tier battle (10-9-8), 1 self-propelled guns are found – 35.95%
• Three-tier battle (10-9-8), no self-propelled guns – 22.72%
• One-tier battle (10), 3 self-propelled guns are found – 41%
• One-tier battle (10), 2 self-propelled guns are found – 33.24%
• One-tier battle (10), 1 self-propelled guns are found – 19.24%
• One-tier battle (10), no self-propelled guns – 6%
When players come across against 3 arties, it is a lot more distinct and noticeable than 1 or 2.
- Main job for arty – to disrupt the players.
- To make reserves last for multiple battles have been discussed for a year now, technically, everything is there. But the feature does not get time in Roadmap – situations more important arise at this moment. But the idea itself is good.
- We even tested the pre-implementation of the technology, as in WoT Blitz, for this New Year. But by the way, no one turned their attention to this. There is no data yet, there is no specifics, we need to process the data and understand whether we are going in this direction or not. There are no prerequisites for the appearance.
- There is an automated search for coexistent plans. (?)
Step 1: non-verbal communication, as in Apex Legends. Next week we’ll study a prototype in order to decide whether to put them on the track or not. In production phase.
Step 2: remaking text communication. There are many tasks that we want to do with the chat.
Step 3: redo voice communication. It is very closely connected with step 2 and what service serves us. And this whole story should end with auto-matching: selection of platoons and smart good auto-matching of clans.
Now we are at step 0. This year we very much hoped to start doing steps 1, 2, 3, but due to problems in the negotiations that we had with the provider of such services, nothing was successful.
When firing at tanks: the server does not determine the value of an event at the moment when it occurs. The server calculates a hell of a lot of these values in advance, such as the aiming circle size during firing, shell dispersion inside the circle, damage you will do, indicators for armor penetration, it normalizes all of this. When there is time and opportunity, there are millions of such rolls. And when some system says, “Give me a random value,” the server says, “Hold on.” From this collected lump, he provides one at a time. In a sense, each shot of ours in the battle has already been calculated somewhere, but since the order of occurrence of events when each player does something is unknown, this remains a random variable, despite that there’s already a sequence somewhere.
Since in some sense these events have already occurred, we only determine the order of their occurrence in fact by chance, that is, from the action of a multitude of players at each moment in time.
When such data ends, the server still generates new ones.
If you shot and did not hit, this does NOT mean that next time you should have better dispersion (the projectile should fly to the center). If you are unlucky with this sample, the shell, three, and four times in a row can deviate greatly from the center. There is no pattern.
-There is no advantage in the battle and extra comfort if you have loaded premium ammo or are using (the boost you buy with bonds). Punishing a player for this is simply unprofitable, it will not pay off. There are no punishments in the game, and no one will ever do them. We fight on the server for every nanosecond of CPU time. This is a belief among old experienced players, it is very ancient.
(!!!) The AMX 30 B is not a very comfortable tank, but according to statistics it is good, it has a good audience. The edits that were considered for the AMX 30 B in that form are completely cancelled.
Tech Tree inflation is inevitable, but we try to keep it minimally slow. At that moment in time when certain collector tanks will begin to sag below normal rates and get below a few indicators, at that point in time we will do something with them.
- What is usually a norm indicator?
There is such a thing as a skill dependence curve. This is a kind of construction that says how a particular tank affects your skill (expected percentage of victories). And if there is such a tank that is systematically higher than the norm for all players from your account indicators, then it improves this hidden value. Such tanks are not normal, either an overpowered or an underpowered tank.
Separately, there are still extremely skill-dependent tanks (wheelies) and very stable tanks (many heavies).
The T-62M tank somehow slipped into the tank carousel in one of the WG videos. What was it?
– The T-62M is a very interesting tank. This is a 1983 design, a deep modernization of the T-62, with a ballistic computer, modern fire control system, package of active armor, smoothbore gun; it can shoot rockets, and it has a dual armor package. This is an easter egg. Why added? Find out, maybe sometime later. If such a tank got into WoT, it would be at tier 13, and maybe even 14. The tank is very scary.
Zeroing the account through an application to the CPC: the cancellation was stopped only on the RU-server, in Europe it continues. Why? All this business was processed in the manual mode. In manual mode means two things:
a) long from the point of view of people;
b) errors occur, and these errors led to the emergence of new applications in the CPC with requests to correct these errors, which made it a simply unacceptable luxury. Up to 30 thousand requests were simultaneously in line.
Now we have written specials. a group of people who are developing such internal services for all of our games. Filed in order to make this procedure automatic.
We want and are discussing little by little how this could be used to make it a little more interesting than just dumping everything, losing it and leaving some economic things there. The CPC simply could not cope with the load, so it was closed. At the same time, everything is OK on the EU and NA.
(No clue what they are talking about, but doesn’t seem important. Possibly the tank recovery system)
We are working on a Subscription system, as it is now in WoWS. I don’t want to do it with a dumb procedure.
– The plans for 2020 for collectors in WoT are NOTHING. Although yes, there are many collectors in the game.
– Global Map will be talked about soon. We will continue to develop this thing. Plus, we try to make sure that GM and other events don’t coincide. GM – this is one of the large direction vectors for tanks for the next 3 years. With GM, everything has become quite good in the last 2 years. We try to devote quite a lot of time to improve this feature. Specifically, the problem of this spring is a DDOS attack. For which we have already apologized, and I will apologize again (we are talking about the transfer and about the fact that 5 events in tanks were crumpled at once in one week). There is a long strategy that has already been approved from all sides, even from Slava Makarov and the original GM team when it was first developed.
– Minimap pinging will be improved upon, similar to how it works in Frontline. This will come with Non-verbal communication improvements.
People who were highly engaged with the company back in 2013 and made plans for the future are no longer working for Wargaming. (Seb: Probably why WoT went bad)
Expedition 2020: we will come up with compensation for tokens that cannot be spent on anything.
Basically in the “Expedition” 2020 it was laid down that the main award is one tank. The second tank is an additional reward. We can see that the audience doesn’t understand this situation. We have it here more towards “all or nothing.”
About the additional token earnings: in connection with the current situation in the world with coronavirus, we’ll see and discuss – I also promise. Offer to finish the game, return the boosters (they are not very compatible with the Steel Hunter) – we will look for a solution, we started the conversation last week. I would like to tell you the details, but they just aren’t there right now.