Scarper

Q & A Session with Kirill Yudintsev re: Current BR, MM & QM. 7/10/15 15:00 GMT

cover_facebook_answers_developers.jpg

 

 

Hello Pilots and Tankers,

 

Please prepare and post you questions here between 15:00 GMT and 18:00 GMT (18:00 MSK - 21:00 MSK) on Wednesday 7th October. The Topic will be unlocked between the hours specified.

 

This topic will be opened 1hr before kickoff, that is 14:00 GMT (1 hours time), where your questions will be pre-moderated, at 15:00 it will be opened fully and any further questions from that time will be fully visible. If you have something you want everyone to see, please wait till 15:00 GMT.

 

 

Important

 

Any posts breaking the forum rules or the following topic specifics will be acted upon:

 

Topic Rules:

  • All questions must relate to the topic in hand, the BR. MM and QM system currently in place or any future development of it.
  • Each user must only post one question at a time, please do not spam the same question over and over.
  • Keep it civil and avoid sarcasm please.

Thank you very much. o7

 

[hr]

 

Current Feedback on BR/MM changes earlier

 

Air Combat - Recent changes in BR's and MM for aircraft based on feedback and statistical analysis are not giving us or you the community what is intended. There will be another change in the near future (probably about  a week). somewhere close to what was previously in place. 
 
Ground Combat - Changes in the ground forces aspect is inconclusive at this time and are still being evaluated, but we will see. Currently we are still analyzing information, stats and feedback, but generally current changes shows some improvements in most metrics of the game, so significant changes are not likely to appear immediately (but we are still examining the data).
 
More details will likely be shared this coming Wednesday.
 
We would like to thank our fans for both identifying some issues and continuing to play the game (in fact, unusually, more played after the changes so we have received an increased amount of data than usual). Please continue to provide us with your valuable and detailed feedback.
 
Many thanks.

 

  • Upvote 18

Share this post


Link to post
Share on other sites

Topic is open for pre-moderated questions, in one hour the topic will be unlocked fully to allow the session to begin. Please be nice guys thanks. o7

Thanks Scarper.

 

 

Hello Mr. Yudintsev,

 

Thank you for this Q&A and your time. I would like to ask why you and the others at Gaijn have decided to implement these most recent BR changes which do the complete opposite of what was requested by many and compress an already squashed BR system. Such changes were unasked for and gives off a misunderstanding of the devs giving off a "we don't care" vibe towards the War Thunder community. 

 

Were the changes simply for testing purposes to further develop the new QM system? Or will it be scrapped altogether given the amount of feedback? And if the recent BR changes were indeed for testing purposes have you considered opening a test server to test such systems instead of putting the changes on the live server which could damage a players enjoyment of the game?

 

Is the Quantitative Matchmaking system the only possible addition to the MM you have currently planned or do you plan to test other systems for the MM?

 

Hope im not asking too many questions at once and if these questions have already been answered then I apologize. Thanks for you and your teams hard work.

 

Regards, YanderesFTW

 

------

 

Edit: (not sure if im allowed to reply in this way but i'll just leave a reply here like this just incase)

 

Hello
First of all - requests of making MM in smaller spread of BRs (like +\- 0.3) can be simply rephrased like 'make MM meet only the same vehicle I use'.
That's very simple conclusion - if you don't want meet any stronger vehicle, than you automatically will meet only your 100% equals (cause weaker vehicles are not allowed too, obviously). In the same time - the only real equals are exact the same vehicles.

Thank you for your reply, but the community was hinting towards more of  a +\- 0.7 MM spread as this would somewhat shorten the distance between the performance of a low tier vehicle (ex, 6.0 BR) against a high tier vehicle (ex, 6.0 BR) which would be better for all then attempting to fight for example of BR 7.0 vehicle in a BR 6.0 vehicle while at the same time most likely won't effect the queue time to much. I think a +\- 0.7 MM spread is worth testing. But your correct, having the spread at something low such as +\- 0.3 would end up being problematic.

 

Another important reason for BR spread is queue waiting time - the more different 'games' you have the more people will be waiting for the game. Waiting for the battle for an hour or even 'just' 15 minutes is not acceptable for most players. Reducing BR spread for +\-0.3, for example, will expose average waiting time for 7-10 minutes. 'Average' means that during 24 hours most of time you'll be waiting much more time.

I can see why some would choose queue time over game play quality but what about the people who don't really care about the queue times? Do you think you could add a feature that could be enabled or disabled that would allow you to fight aircraft that are more similar in performance to yours for those who don't mind long queue times as opposed to those who do?

 

Scarper (Today, 02:19 AM): 

Might be worth posting again m8, hes working his way through them and as im sure you agree, it is a slow process. o7

Alright, thanks again Scarper o7

Edited by YanderesFTW

todace (Posted )

Hello
First of all - requests of making MM in smaller spread of BRs (like +\- 0.3) can be simply rephrased like 'make MM meet only the same vehicle I use'.
That's very simple conclusion - if you don't want meet any stronger vehicle, than you automatically will meet only your 100% equals (cause weaker vehicles are not allowed too, obviously). In the same time - the only real equals are exact the same vehicles.
Otherwise the game will have no gameplay sense of progression (both be metagame terms and skills, cause meeting only one enemy type with one teammate significantly reduces possible tactics and will to get another vehicle).
The exact BR spread depends is driven generally by simple rule - player should meet at least 10 (better - 30) vehicles in battles (not in one battle). With the principle that team have to have a chance of winning.
Another important reason for BR spread is queue waiting time - the more different 'games' you have the more people will be waiting for the game. Waiting for the battle for an hour or even 'just' 15 minutes is not acceptable for most players. Reducing BR spread for +\-0.3, for example, will expose average waiting time for 7-10 minutes. 'Average' means that during 24 hours most of time you'll be waiting much more time.

Changes in BR (as well as more vehicles) were something we wanted for long time but we can't made them before QMM was made, according to a principle of possibility to win.

Scarper (Posted )

Might be worth posting again m8, hes working his way through them and as im sure you agree, it is a slow process. o7
  • Upvote 25
medal medal

Share this post


Link to post
Share on other sites

Although the community was strongly against the suggested BRs before the new ones became live, you still went ahead with the changes, why?

todace (Posted )

I have answered above.
  • Upvote 11
medal medal medal

Share this post


Link to post
Share on other sites

It's obvious that T-10M and Maus are too strong for most 6.7 tanks. How about bring them back to 8.0? If a player with 7.0 plays against 8.0 tanks QMM should limit the amount of this vehicles down to 4.

According to this take the new ammo for the IS-2 Mod 1944 back if you bring Maus and T-10M back.

todace (Posted )

1. any of 6.7 BR tank can (with good skill and proper distance and proper bullet) kill T-10 or Maus. Well, it is hard, but possible.

2. You have not more than 4 of top tanks in enemy team, and in the same time your own tops

3. Planes can kill Maus and T-10 easily

4. Not all battles are with T-10 against 6.7

5. According to stats efficiency (K\D) of Maus havent increased at all, and T-10 performs only a very bit better

6. We've played it a lot on this rank (well, not me personally, I've just reached 4th rank on tanks in USSR on my current account and mostly play US tanks on 2-3rd rank now, or Jets in planes) and find it challenging but playable.

6. We will check all stats in a nearest future again.
  • Upvote 2
medal medal medal medal medal medal medal

Share this post


Link to post
Share on other sites

Dear Mr. Yudintsev,

I am happy to see You here answering questions. I am grateful You have decided to spend time with us, and I am hoping the questionnaire will be fruitful and good-tempered and will be of great  use to You.

I have a question that has to do with Battle Ratings, but I am interested in the grand model, not particular planes.

QUESTION: Would it be possible to get an in-depth explanation of how battle ratings are formed – the calculations and mechanics under the hood, to be precise? I remember Gaijin doing an admirable in-depth analysis about how plane flight models are formed. I am respectfully asking You to consider a similar devblog or a simple questionnaire or post about how BRs are really formed from raw data. I understand simply explaining it in short terms is most likely impossible and beyond the scope of this questionnaire.

I believe such explanation would unshroud some of the mysteries around the BR calculations, seeing how they are often a source of much discussion. Since BRs are formed due to statistics, then I believe it could be helpful to reveal just what statistical tools and analyses are used.

I understand some of this might be business secrets or otherwise sensitive, and You might not be willing to share all the excuisite details of the BR calculation system. Regardless, I hope You could shed some light on this matter.  If you decide to answer or pass the question forwards, please see supplemental.

Thank you in advance.

Respectfully Yours,

El_Perkele

SUPPLEMENTAL: The Battle Rating system is somewhat unique to War Thunder. Based on what we have heard, the BRs are based on statistical analysis. Personally, as a player also interested in and employing statistical analyses in my line of work, I would be interested in knowing as much as possible about what metrics are used to determine battle ratings. I have been toying around with my personal data (I know, n=1) and have noticed some very interesting bimodal and skewness patterns in metrics such as kill-to-death ratio, winrate, missions per kill and so on in my personal data. (Some planes’ aberrant BRs, such as F4U, hint towards a certain set of problems, if the analysis is mostly statistical.) Although most of this is probably normalized in larger dataset, the presence of skewness and bimodality presents interesting questions as to how they are handled in BR calculations. Essentially, what I would be interested is a detailed answer dealing with things such as dealing with non-unimodal distributions, distribution skewness, non-normality and using median and mean as tools. Are You using nonparametric testing or polynomial regression to produce comparable results? How are the players categorized into class factors, or is it one continuous variable? How is the comparison value formed and adjusted? J My questions are endless, but I believe those working with the hard calculations of statistics-based BR system understand perfectly what I am respectfully asking, and I have full trust they can produce a satisfying and coherent reply even under stricter NDA policy.

I hope my question is not too much work, and I believe a throughout explanation of how BRs are  formed will help us all to further developed the ambitious and unique system your company has put a lot of work towards.

If such an answer exists, I am sorry for requesting a double feature and would like to be guided to such an answer.

R.Y.,
E_P

todace (Posted )

Some principles of whole MM I've put above.

About Tanks BRs. Planes BRs are mostly calculated with just pure stats data and community feedback (last part in fact is more important and affects BRs more in RB than stats).

Tanks BRs are also calculated by stats (K\D, Hits, critical hits, assists, life time, amount of sessions played) but also we specifically check that tanks that mret in battle can be hit\killed.

todace (Posted )

Regarding exact formulas and stats data.

First, we have found that different timezones have different stats. In RB that's partly because that players prefer to play on their country's vehicles, so for example, there are much more US vehicles in the battles during US primetime (rest of world have night and sleeping).
Another reason is that very experienced players tend to play on the not 'their' country (just because they are played a lot on them and want something new).
So during US primetime for example, there are more experienced and skilled players on Germany vehicles than on US vehicles.

The exact formulas is something that is we really do not want to share, not because of big business value in them, but because of 'self-fulfilling prophecy' effect - experienced and skilled players (which stats data we analyze independently) will use BR as stats data and affect the stats result.
  • Upvote 2

Share this post


Link to post
Share on other sites
My Question :
 
Arcade / Realistic Planes : Would/Could there be any additional ground target  / ground strike / base domination missions added in the near future with major targets for bombing and strafing like factories, airfields with aircraft, oil refineries, rocket launch sites, military installations , shipyards, submarine pens , railway and railway yards, military trains,  ect spread throughout the maps giving the players more targets/missions to do during the matches vs just capping bases ect. I think if this would be possible it would make the game more realistic/challenging/fun. I know it would be very welcomed by many players to have additional things to attack or protect. , thank you.

todace (Posted )

Our designers really want it and push it.
However most of location artists are now working on maps for tanks. When we'll have 18-20 of them, we'll pay more attention to new mission features.
And now we have 16 of them, so that is going be rather soon (after 1.55 I guess).
  • Upvote 4
medal medal medal

Share this post


Link to post
Share on other sites

can Future MM are more reasonable on Date of vehicles ? and can u extend range of MM ? from current higher 8.3-8.7 to 10.0 ? . Currently MM is unreasonable and very disappointment of player who doesn't have high Br vehicles to drop on with higher performance vehicles and had a hard times to fight with higher performance vehicles . this cause a lot of player to stay out of the games and wait for responded until game balance is more likely to playable to come back into the game . player who have higher BR tank better performance and matching with player who try to grinding to get better vehicles like top end player is facing far harder times  to get score than player whom have higher performance vehicles with new Current BR relative that surely had far more easy time than new - mid player what is your resolve of this problem .  i sure previous MM BR is much better than currently .

todace (Posted )

Well, our stats shows that MORE players played significantly MORE sessions with new BRs (on tanks at least). Not any BR showed decreasing in popularity.

But we'll be regularly check all data.
medal medal

Share this post


Link to post
Share on other sites

Привет Mr. Yudintsev

 

It was reported the BR system is adjusted by game statistics, I assume K/D ratios.  Those (vehicles) with more kill success get modified upward while those who are killed more often are downgraded. 

 

The flaw with this is easy to play units skew the entire system and cause confusion and frustration.

 

Why not use a BR range based strictly on performance and weapons, and THEN have a modifier applied on statistical data? 

 

Then the games are be made in tune to skill, mode, and other variables with out having to constantly adjust the BR as is done now.

todace (Posted )

1. If you have any clear formula in mind of how calculate BR or MM based on performance - provide it please.
We (and me personally) spend hundreds of hours trying to create something that can be based on performance only.
Everything showed significantly poor results.
Until you comparing something close supposing to meet only vehicles of their kind (for example - high-altitude fighters, or bi-planes) you'll get nothing reasonable.
Who is better - P-40\P-63 or B-17? I-153P or Bf-109E-3?
It REALLY depends on what they want to do, what is the goal and whole situation.

So, that's also contains response for how we calculate BRs - not, not only by K\D, otherwise most of bombers (not B-17 - he is good in killing planes) and attackers would have very close to each other BRs

But regarding tanks - we do use their parameters as an input to BR. Cause they have something very comparable - what 'amount' of bullets of their supposed enemies they can handle before they will be down (and how fast and easy they can destroy the opponent, so generally who will be first)
  • Upvote 21

Share this post


Link to post
Share on other sites
  • A few questions, about SB's BRs. Mine are the first three, the  other two people are other SB players. 
    • Will Gaijin consider making SB-air's fighter BR's historically based? (with adjustments for broken FM's, and moving prototypes/paper planes higher)
      • Most other flight sim's use this system and it works splendidly there (e.g. all the IL2's, DCS, Aces High, etc.), but here we have match-ups where a good 1943 fighter like the F4U-1a has the same BR as a 1941 Spitfire Mk Vb, which is absolutely not 'equal' or balanced. Using dates offers an objective measure, for instance nobody will have to argue about the La-5FN's proper BR as it is modeled off a 1944 test-bed version of the La-5; no one will argue about if the Typhoon Mk IB/late is fair at 4.0 as it's a 1944 plane too, and nobody will argue about if the P-47D is competitive with 1944 planes as it will be fighting historical opposition and will have to adapt as they did IRL; there's many more examples, but these highlight the issue. 
    • Would Gaijin consider changing it's statistical analysis if it won't do historically based BR's?
      • The current adjustments mirror using a simple arithmetic mean, by stitching to a more complex model (the specifics are the topic for another thread though) Gaijin can mitigate the way that FoTM planes have their BR's adjusted.  
    • Would Gaijin consider allowing players to have some input on BR's changes?
      • Basically to be used to allow planes to be moved up or down based off the FM's current state. 
    • ZdrytchX
      • Do they really believe that Me 262s are superior to the current immediate post-war jets with their current flight models? 
    • Mr4nders0n
      • Could we at least in Sim have a more performance based BR calculation?
        • I don't care if AB/RB use player based statistics, but such things have nothing to do with a simulation, as which WT claims it is
        • Believes Gaijin won't do Historical BR's. 

 

P.S. Scarper, thanks for your response in the other thread, as you couldn't I rearranged my schedule so I could do it. 

todace (Posted )

This is very old question.
The problem is that there is no such thing as 'historically based BRs'. There were planes that were used for 10+ years without any modifications. Different planes of different countries were never made as equals - they were made trying to be best when performing their battle tasks. Sometimes it required for example 5tons of fuel more than do have their opponents (of course that made their less effective).
German jets were for example in service from 1944 and they can easily bit ANY prop fighter that they can meet. They were made to fight hundreds of B-17 and\or to carry bombs, and they were much more expensive but they are not equals to their enemies.
'Other games' do not have random battles of 400+ planes.
They can:
a) do not care about MM at all
or
b) pick only those planes that are generally close to each other (like Bf109F and La-5F or Bf-109E and Spitfire II).

We do not have such options - we have random battles with random players, all planes must be able to enter battles and we have 400+ of them.

But we are going to create special historical events with QMM that will allow us to try recreate real battles.

todace (Posted )

Me-262s are not superior than late post-war jets (but for example better than MiG-9s). BRs of Me-262 will be changed this week I think (to something closer of what was previously)

todace (Posted )

Current BRs calculation are not using simple arithmetic mean.
  • Upvote 7
medal medal

Share this post


Link to post
Share on other sites
 

1 Question from me:

  • Would it be at least possible to fill the brackets in Enduring Conflict with more or less historical plane sets instead of BR based ones?

todace (Posted )

We'll see.

Currently we are still working on EC itself.
  • Upvote 4
medal medal medal medal

Share this post


Link to post
Share on other sites

What will happen to the F2A-1 BR (both regular and Tach's versions) in realistic battles ?

 

BR change was announced at first

But when the BR changes became effective, no BR change to the Buffalo.

 

So there is actually nothing to revert.

Still the F2A-1 BR is a huge problem in Tier 1.

 

Don't forget that it currently clubs new players in their bi-planes. 

Not a good thing for nobody.

todace (Posted )

thanks for your post
I'll check
  • Upvote 8
medal

Share this post


Link to post
Share on other sites

It would be interesting to know what exactly has to be understood under "what was intended"? Could that be clarified a bit.

 

What was intended, maybe that is something the community misinterpreted, and had false exceptions toward.

todace (Posted )

I hope I've answered in a first post.
medal medal medal medal medal

Share this post


Link to post
Share on other sites

Can we increase the BR range, especially for ground forces? QM is not working in 6.7-7.7 matches since the post war tanks is just too powerful for WWII tanks to handle.

And I don't think increasing the BR range causes the MM wait times to increase, since I'm already waiting about 3-5 minutes for a RB match at 5.7, the same wait times as when BRs were to 8.0 for GF RB. So compressing BRs is not a good solution.

todace (Posted )

In fact our data shows reduce in waiting times.
medal medal medal medal

Share this post


Link to post
Share on other sites

Why Gaijin introduced more compressed BR range instead of what players want?

todace (Posted )

I think I have answered in a first post
  • Upvote 1
medal medal medal medal medal medal

Share this post


Link to post
Share on other sites

Are you planning to make Germans 128 mm cannons get 311 penetration? we all know there tanks slow and this won't effect balance.

todace (Posted )

1.
We are using correct data for penetration of 128mm cannons.

2. We are not balancing tanks with their DMs. They all supposed to have as much realistic DM and cannons as it possible
  • Upvote 3

Share this post


Link to post
Share on other sites

Will Enduring Confrontation get it´s own fixed MM tables for the different brackets?( The current  player performance based BRs from the old modes simply do not reflect the special- or multi-role qualifications of the aircraft in a an enduring confrontation battle as opposed to the old modes aka team deathmatch. Futhermore the aircraft performance balance has to be even finer tuned in EC due to the unlocking over time feature to be fair.)

todace (Posted )

We are planning to replace current SB with EC mode.
So all BR of SB mode will be by EC.
  • Upvote 9
medal medal medal

Share this post


Link to post
Share on other sites
Comment:
As long as your Battle Ratings are not correct (The Base of MM), no matter how much you code it, it will fail.
Even if you have same amount if 7.7's on each team, it doesn't mean it's balanced. The point is what plane does that nation have at 7.7BR. Me-262 C-2b is 7.7 and F-86F-30 is 7.7. F-30 is vastly superior.

You need to start handing out BR's according to the vehicle performance and not player performance.
People do better in A6M5 Otsu (5.3BR) than most people do in P-51D-5 (4.7BR). Why? A6M5 Otsu has player friendly style and most players just want to turn fight. But is A6M5 Otsu really superior to P-51D-5. Definitely not!!!

So here's my question:
• How many decades will it take for you to start assigning BR's according to the vehicles acuall performance rather than player performance?!!!

Further comment:
Players that use vehicle the wrong way shouldn't have 50% chance to beat a player that does use it's vehicle correctly just because your stats read otherwise.




I know I shouldn't ask 2 questions but I thought it was a very important one to include.
Players complained about over compression for over a year and what you did instead was ignorant towards the community. You lowered the max BR cap over impressing the game further more!
Here's the question:
When will you increase the max BR cap of planes/tanks/ships to 10.0BR?

todace (Posted )

I have answered that above.
There is no way to compare performance data of different planes (only in some specific cases like duels)
  • Upvote 11
medal medal medal medal medal medal

Share this post


Link to post
Share on other sites

Are there any plans to rework tier V aircraft battle ratings? currently most jets either club mercilessly or are clubbed helplessly performance wise (komet vs sabres, MiG-9/F-80,F2H vs early 262 and kikka), and the Hawker Hunter will only hurt this if added first. Not to mention that the import Sabres and Migs have the same problem premium planes had being a lower or higher BR than their counterparts.

todace (Posted )

Yep, working on them.

This changes were not balanced and good overall, sorry for that.
  • Upvote 1

Share this post


Link to post
Share on other sites

Do you consider adding a hard line in terms of matchmaking between pre end of WW2 and post war vehicles? The differences in technology which occur are extremely unbalanced, e.g Panthers vs. T-10M; Me262s vs Mig-15s or Sabres. One side struggles to even get themselves into fighting position while the other side has got no real challenge in fighting inferior opponents, no tactics can compensate for the lack of essential vehicle characteristics.

:salute:

todace (Posted )

We have implemented this in the past.
There is no sense in it - in this case it will be better to create different research line.
If planes of different eras can't meet in battle - there are two different GAMEs.
  • Upvote 2
medal medal medal medal

Share this post


Link to post
Share on other sites

I think overall the quantitative matchmaking concept is a good idea for most of the tiers.  However, I have been reading of players experiencing  "teething" issues with it in tier V.  Has the data pointed to a possible cause or resolution yet?

Scarper (Posted )

Hes not ignoring yours, just busy answering and missed it. 8(

Share this post


Link to post
Share on other sites

BR range is 1.0 to 8.0, however it increments x.3, x.7, so instead of 70 levels we really have 21 BR levels.  With the down tier we now have effectively 20 BR levels. 

 

The units in game range from 1936 to 1954, 18 years, 1 BR per year (This sounds very similar to the 22 level Tier system as before).

 

Why force units into such a tight BR range (+/- 1.0) that results in players to gamble they will be in game with over tiered opponent?

 

Have you considered a broad (50 level) BR and then adjust the MM range?

Or allow more lower tier units to make up for high tier opponents (to the BR total for each side is as close to equal as possible)

todace (Posted )

Yes, we had.
Again -
1. we want 20-40 possible opponents AT LEAST.
2. there is no any sens in MM based on the year of production. Po-2 fought in Korean War, German Jets in 1944
  • Upvote 11

Share this post


Link to post
Share on other sites
Guest
This topic is now closed to further replies.

  • Recently Browsing   0 members

    No registered users viewing this page.