Think you're always Uptiered? Take a deeper look.

[YoMama2 1,232]

On 7/4/2017 at 2:33 PM, JagerShooter said:

I'm obviously missing something. Without BR's of individual players vehicles shown in the post battle summary, how are you able to determine what you are  up against. Why has Gaijan decided to shield this information from us. Was it to stifle all of the controversy surrounding this post as well as many others. Just asking why you all have not had a problem with this omission.  

Incidentally my previous post, never received a Moderator or developer response as well.        

 

Boy I hear you... it was SOOOOO MUCH EASIER to look at the post battle summary.   In its infinite wisdom of breaking things that weren't broken, Gaijin removed the post battle stats right as I embarked on this quest.    Gaijined again, what can I say.   I'm used to it by now.   So I must PAUSE my fighting after my first death to count everyone for a few minutes and then get back into the action.   Yes, it does affect my game performance.

[YoMama2 1,232]

On 7/4/2017 at 2:44 PM, Quacki said:

 

 

I scan the player table in the 20 seconds just before the battle starts. If I take longer to figure out min and max BR, I'd use some time into the beginning of the battle. But Yomama2 is hardcore: He stops mid-battle to do the full count to take the full data and I commend him for his effort. It really is a shame Gaijin removed the post-battle informations! (also see below) It is a mystery to me as to why it was removed; I generally don't buy in the conspiracy theory that they did so to obfuscate or hide some imbalance, but on the other hand I think I never saw any official statement.

 

 

Obviously this is too much hassle, but I would be interested in the data from both teams. I would like to see wether MM tries to balance fail squads. As far as I can see, it doesn't, but then again, I do not pay too much attention to fail squads.

 

In my past surveys (I've done A LOT of them) I have looked at both teams.   I can tell you that the MM does NOT seek to balance the two teams with respect to: a) the BR's of the tanks assigned to one team or the other, or b) the ranks of the players randomly assigned to both teams.   When you look deep, you see some very imbalanced sorts, but this is because it's done randomly and random sorts produce outliers.   I've seen matches with 5 marshals on one side and none on the other.   I've seen matches where one team had 3 top dogs and the other didn't have ANY tanks that were top dogs or any that were .3 below the td's!   I've seen a match that had five fail squad members with 1.0 BR tanks in a 4.7 match.   (They might as well have stayed in hangar.)

 

But it DOES seem to balance out organized squads presenting themselves to battle.   In other words, if two organized squads are ready to go, the MM will NOT put them on the same side, it'll put them on opposite sides.   It does nothing to weigh whether those squads are well balanced though.   Meaning one squad could be 4 WINFX spaded and aced marshals and the other could be Capt. Joe Blow and his 3 stock lieutenant buddies.

[Predatorr13 179]

Yes, it does match squad vs squad without respect to their BR or if it is made of same squadron team mates. Playing 5.7 we get HuntU (usually at 6.7), WinFx (at 5.7 or 6.7), or 7eg (at 6.7 or 5,7) for example. But using 3 random team mates the squad is treated just the same...it matches you with another squad of 4 (or 3 if 4 isn't available and so on). You do appear to get better matches when your squad is 3 rather than 4. 

[Maj_Fox@psn 521]

Question.

 

If you have all 5.0 vehicles and the Reds have a 5.7 in their Line ups but only drive out 4.7 are you up tiered? 

 

The reason I ask is that it may be more important to look at the vehicles used and not all of those in the line ups.

 

Cheers 

 

 

[SneakySausage 786]

I think it's most important to look at what they could have taken out as compared to what they did.

 

Sometimes people bring out a lower vehicle for tactical reasons. Speed, gun depression/elevation and so forth. The fact that a 6.7 was In the lineup, and could have been used is in my opinion what should matter whether considering if you've been ''uptiered''

[JagerShooter 64]

I thought that everyone comes to the battle with their respective line up based on the highest BR ranked vehicle. Gaijan then places all of those 'highest " BR's into a MM formula (what ever that may be) and spits out battle map and game player line up. I believe without regard to the BR rated vehicles remaining in your lineup, which may be the same or a lower BR.

Is Gaijan also formulating all of the players recorded abilities as well and then placing them in respective order of team play. If that is how we get battle ranked, then that's one hell of an algorithm............

 

[YoMama2 1,232]

11 hours ago, *Maj_Fox said:

Question.

 

If you have all 5.0 vehicles and the Reds have a 5.7 in their Line ups but only drive out 4.7 are you up tiered? 

 

The reason I ask is that it may be more important to look at the vehicles used and not all of those in the line ups.

 

Cheers 

 

 

 

4 hours ago, JagerShooter said:

I thought that everyone comes to the battle with their respective line up based on the highest BR ranked vehicle. Gaijan then places all of those 'highest " BR's into a MM formula (what ever that may be) and spits out battle map and game player line up. I believe without regard to the BR rated vehicles remaining in your lineup, which may be the same or a lower BR.

Is Gaijan also formulating all of the players recorded abilities as well and then placing them in respective order of team play. If that is how we get battle ranked, then that's one hell of an algorithm............

 

 

Since the MM makes all of its decisions BEFORE the match starts, it cannot tell what vehicles the players will eventually choose once play begins.   And the best vehicle available may not necessarily be the highest rated one in your rack either.   When I play 3.3 German, I never use the Pnz III's, which suck,  I like to come out swinging with fast 2.7 vehicles like the Pnz II H.   But that's immaterial for our purposes here.

 

There are thousands more variables that go into figuring out who will WIN any given match, actual vehicles used being one.   Spading, acing of crews being others.   Personally, I think that the single biggest determinant of who wins a match vs. who loses is this:  which side has the best four guys?   I don't think that the other 12 contribute much one way or the other.   If your top guys are better than their top guys, regardless of if you're in a real squad or not, you'll win.

 

But we're not interested in who will win these matches, we're interested only in the pre-match favorability of the sort.   MM only looks at the top 3 vehicles in your rack, averaging them for planes, picking the top tank for tanks, and then it matches them up accordingly.  After that, the MM's job is done and it's up to the players to shake everything out.

Edited July 7, 2017 by YoMama2

[Quacki 355]

On 6.7.2017 at 5:54 AM, YoMama2 said:

 

In my past surveys (I've done A LOT of them) I have looked at both teams.   I can tell you that the MM does NOT seek to balance the two teams with respect to: a) the BR's of the tanks assigned to one team or the other, or b) the ranks of the players randomly assigned to both teams.   When you look deep, you see some very imbalanced sorts, but this is because it's done randomly and random sorts produce outliers.   I've seen matches with 5 marshals on one side and none on the other.   I've seen matches where one team had 3 top dogs and the other didn't have ANY tanks that were top dogs or any that were .3 below the td's!   I've seen a match that had five fail squad members with 1.0 BR tanks in a 4.7 match.   (They might as well have stayed in hangar.)

 

But it DOES seem to balance out organized squads presenting themselves to battle.   In other words, if two organized squads are ready to go, the MM will NOT put them on the same side, it'll put them on opposite sides.   It does nothing to weigh whether those squads are well balanced though.   Meaning one squad could be 4 WINFX spaded and aced marshals and the other could be Capt. Joe Blow and his 3 stock lieutenant buddies.

 

Ok, that is interesting. I actually thought there was some kind of sophisticated measure behind the occasions I saw imbalanced teams in terms of BRs, for example by having better (higher rankedas in more XP) players in lower BRs to offset the higher ranked tanks on the other side. After one particular steamroll I recorded the player ranks in order to see if this could explained the onesidedness, but there was nothing remarkable. Takes a lot of time to record, though.

 

About the squad balancing, this I also noticed. One afternoon some time ago we played in our squad of three and were constantly matched against a COBET squad which kicked our asses all over the place.

[SneakySausage 786]

It is completely random teams, only balances and by balances I mean split it so that organised squads (if there are multiple) onto different sides. 

 

This can be defeated tho, as 

A) no other squads in que

B) you can multi squad If your wearing clan tags, and have other non clan members with your squads (4 tags max) 

[cheatsreported 30]

It's a fallacy to believe Gaijin uses your skills otherwise you wouldn't be matched with so many people that have 30k games and are expert as well as 100 level. The fact is if they were using real skill parameters then it wouldn't matter if you were up tiered because your skills would make up the difference. Instead they use "victories/battles" giving you your "vehicle battles ratio" which is actually a useless number because it doesn't take into account your teammates may not have been that seasoned or they didn't use any kind of strategy so you got killed multiple times. I see people that I play against that have a battle ratio of 55% bu their kill ratio is 190%. I just got killed by a girl that had a Panther II, her "battle/kill ratio" was 100%, 57 /57 so I looked at her other vehicle stats and most were over 100% but she was only a level 36. I'm a 60 but since I keep getting paired with people that have much better stats I can't move to another tank or get expert in my tank. It sucks but I'm sure their response is we take everything into account but they really don't. If they actually did take the "battle/kill ratio" into account a lot of people would be banned because their "third party skills" would certainly stand out. 

[Querulous1 3,276]

The randomness of the BR match up on either side can be maddening. Its often funny to see people complain in chat that one side has 2 or more top tier than the other, as if this was a bug.

[vraffini@psn 31]

10 minutes ago, Querulous1 said:

The randomness of the BR match up on either side can be maddening. Its often funny to see people complain in chat that one side has 2 or more top tier than the other, as if this was a bug.

Totally agree + people complaining about long queue time. Like completely balanced match + short queue = not possible.

[cheatsreported 30]

I just got killed in a battle by the sam tank I was using, difference was he was 100 level expert but what I don't get is why none of my shots mattered hitting his armor in weak spots. Should not matter what rank he was or if he was expert. Definitely a flaw in this game.

[SneakySausage 786]

Dunning Kruger in affect

[Panssarikolonna 1,086]

4 hours ago, cheatsreported said:

I just got killed in a battle by the sam tank I was using, difference was he was 100 level expert but what I don't get is why none of my shots mattered hitting his armor in weak spots. Should not matter what rank he was or if he was expert. Definitely a flaw in this game.

It's called RNG

[Panssarikolonna 1,086]

4 hours ago, cheatsreported said:

when you see a player that's new to the game and has 3700 kills in 3200 battles with 1200 victories the numbers don't lie. Just got killed by an Object 120, player was level 15 with 2105 kills in 3100 battles so 67% kill ratio and on the 120 she was 31 kills in 37 battles which is an 84% kill ratio.

What?

3700 kills on 3200 battles the ratio is 3.7 to 3.2 meaning little bit over 1 kill per game.

Even I have 1.6 kills per game and I'm not that good.

[cheatsreported 30]

Gaijin should really be looking into some of the players tonight, shooting through mountains for the kill. I'm sorry but the match ups are really getting pathetic, 4 MAUS in one battle all on the enemy side. Their logic for the entire game is going out the window taking what was a fun game and making it a blood bath for us. Most of these guys would crap in their pants if they ever served and had to actually fight but they think its cool cheating. Got news for ya, the enemy would end your life the first 30 seconds your feet hit the ground because you are a loser. You may be able to play the game and think you're awesome but in my book you will be a loser all your life, you have no skills. People can tell me I'm not very good but they haven't actually played with me. Fact is I usually lead and make some good kills but there's always the loser that hides across the board that kills me. Yep, not much face to face action always a wanna be sniper that really shows how demented some of these players are. I really like how many times I've been killed using the dame tank as my opponent hitting him in vulnerable spots multiple times and then getting killed with his one shot exactly where I hit him. The Leopard that takes hits from me when I know he should be dead and then kills me with one shot. These are the losers because I've killed them before but certain players are really wanting to test banning to the limit. I just hope at some point they are gone forever. Gaijin just needs to listen a little better an look a little harder.

[cheatsreported 30]

My point with the 3799 kills is at level 15 every vehicle was nearly 100%. You are being stupid if you don't see the numbers. It takes some time, especially with an object 120, it bounces and its hard to aim fast. Most of the tanks require some getting used to and you don't get 86% kill ratio just starting and over 100% on other tanks besides. How does someone that just stqarted the game have over 200 vehicles? Spending lots of money but it still takes some time to develop the skill. 

[SneakySausage 786]

@takkuta you best ignore him. His mind is gone

[cheatsreported 30]

The logic is flawed no matter where you look. We had 2 goals captured, I was capturing the third while 14 teammates defended the other 2 goals yet the game was called and we lost. Can't really understand their logic but the bar showed we were out and the enemy still had life. Didn't understand that call at all.

[Panssarikolonna 1,086]

If you have enough money you can have all crews aced, have all premium tanks, have all mods installed on the tanks.

It only needs  bunch of euros or dollars to get them.

 

It's not secret that some top dogs has several accounts.

[TheBeast30 0]

On 26/06/2017 at 21:26, YoMama2 said:

After arguing ceaselessly with the guys who insist they are always "uptiered 8 out of 10 times" regardless of what BR they play, I finally decided to do something to prove that this wasn't the case.    Sure, the saner and more mature individuals among us know that this isn't happening frequently, that most of the time it's a pure case of observer bias.   These guys aren't actually keeping records, they're just going by their faulty memories and selectively fixating on the bad matches, while forgetting the good ones.

 

Then there are the guys who don't really even know what uptiering is.   They think that ANY time they see a more powerful tank than themselves, they've been "uptiered".    They don't realize that only four MM possibilities are open to them:   "Runt" - where your tank is 1.0 below the top tanks, "Low Placement" - where your tank is .7 below the tops, "High Placement" - where you are only .3 below the tops, and "Top Dog" - where your rack's BR is the highest in the game.   These second group of guys think that anything but Top Dog is uptiering.  In reality the two lower placements are "bad" and the two upper ones are "good".

 

But even this doesn't tell the full story, because to really know if you've been shoved into an Uptiered or difficult match, you have to look one level deeper at the meta data associated with the match -- you have to look at the BR's of all the enemy players in the match.   This doesn't give you all the data you might want to gauge match difficulty, but it gets you past the false impressions that just looking for top dogs will give you.

 

In order to compare match difficulty of all tank AB matches at all BR's, I built a metric formula.   Here's how it works.   Count all the BR's of all the enemy players in your match, log them in spreadsheet and multiply the numbers of players at each BR by 0 if they are your same BR, +.3 or -.3 if they're one notch above and below you, +.7 or -.7 if they're two notches, and +1.0 or -1.0 if they're 3 notches up or down, etc., then add them all up for an average each game.   If the match is exactly balanced around you, the average will be 0.   If a match is favorable for you, you'll see a low number like -8.2, if the match is truly uptiered badly against you'll see a positive number like +6.5.   The worst reasonably possible match would be about +12.5, the best reasonably possible match would be about -19.6.

 

Then all you do is play a buttload of games and tally up all the matches, looking for an average of averages.   This is what I'm attempting to do at three different BR's that are reputed to up or downtiered viciously.   Let's examine my data to date.

 

First, I'm not disputing that some BR's are uptiered or downtiered more heavily than others.   If you play 8.3 tanks, you're down or evenly tiered 100% of the time.   If you play 1.0 tanks, you're uptiered all the time.   And in the middle you tend to get a mix.    What I'm disputing is the anecdotal rumor offered by whiners.    Until we put some data behind things, anecdotal rumor is rampant, but it's worth diddly squat.    The first 3 BR's I'm considering are 4.7 American (rumored to either be Top or Runt, nothing inbetween for a general trend towards uptiering), 3.7 Russian (Rumored to be heavily uptiered), and 3.0 British (rumored to be heavily downtiered).   While compiling data on these 3 BR's, I never changed the vehicles in my rack.   I tried to minimize the variables in my inputs.   I also tried to mix up my play times because I have seen trends based on playtimes.   Certain nations are awake at different times of the 24 hour cycle and they favor different tanks.   In order to have meaningful data, you have to play at least 40 of these games with a static rack, changing nothing else but times of day.   I'm not there yet and hopefully I'll get there before another patch changes everything.    But what I do see is eye opening.  

 

Consider the 4.7 American rack first -

 

If we were judging by the old method, this rack would have 11 out of 20 matches uptiered, so slightly unfavorable.   But scrutinizing with the deeper look method reveals something very different.   Over 20 matches it has an average of -2.59, which is very favorable.   And counting all the players above and below my BR reveals the same thing:  149 playes below, only 83 above.   These deeper look impressions correspond with my general feeling that American 4.7 vehicles do quite well.    Two other things jump out when looking at this data:  1) very few players at 4.0 and 2) a bunch of players at 3.7.    These observations have deeper meaning, namely: your odds of up or downtiering placement are heavily dependent on the POPULARITY OF VEHICLES ABOVE AND BELOW YOU.   You are pushed and pulled by the masses wanting to play around you.

 

Now let's look at that "popular" 3.7 BR with a Russian Rack.  It's rumored to be heavily uptiered.

 

 

The old method says this rack was uptiered 13 out of 15 games -- about the worst you can expect.   We should be getting slaughtered.   Yet this wasn't the case.   Russian 3.7's do quite well anyway.   Is this due to Russian bias or something else?   I propose that it's "something else".   A deeper look reveals only a pinkish +2.53 average rating, disfavorable, but much less than you'd expect.   Why is this?   Because most of the vehicles carry my exact same 3.7 BR rating.   110 of them out of 240.   And many of the games (like games 1, 3, 4, 7, 8, 9, 10, 13, and 15) which should be uptiered games, actually have more players at or below my BR than above it.

 

Once again, I note that 3.7 has the most players, and 4.0 is empty most of the time.   Interesting.

 

Now let's look at the rumored "downtiering" at 3.0 British.

 

 

The old method says that this BR level is heavily downtiered with only 6 out of 15 matches uptiered.   But once again, the deeper look paints a different picture.  It shows VERY FEW tanks at our same level, with 132 below and 73 above.   Again 4.0 seems to be an unpopular BR so few matches get fully bumped up to it and once even, there were NO 4.0 tanks present in a 4.0 match!    Lots of 3.7 tanks are around, and a fair number of 2.7's as well.   Here's what happens when you play 3.0 -- you get downtiered a lot, and when you do, you rule.   When you see 3.7 tanks, they wipe you out.   Why does this rack get a lower average score than the 4.7 rack above it?   Because even though it has proportionally more tanks below our BR level, the aren't as far below us.  The popular BR level below us is only .3 below, not 1.0 below as in the case of the 4.7 rack.   The 4.7 rack saw more tanks that were proportionally much weaker than themselves.   The 3.0 rack sees a lot of tanks that are barely weaker than themselves.

 

Conclusions -   A superficial look at just the BR level of the top tanks present in match does not give an accurate picture of relative uptiering upon sorting.  A deeper look at the BR levels of all the enemy tanks present in the match is needed in order to calculate uptiering upon sorting.   Those deeper looks reveal that match uptiering is rarely as severe as it seems at first glance.   They are usually more balanced towards the middle.

 

That said, a study of the deeper data can reveal sweetspots and dryspots at certain BR's which can be exploited.  

 

Hypothesis: 4.0 tanks are rarely played, as are 5.0 and 3.0 tanks.    With this in mind, a player playing 4.0 tanks might expect to see 4.7 and 3.7 tanks frequently, but little at the extremes.   If he sees few 5.0 tanks, he would then expect the 4.7 tanks to be top dogs who are capped frequently and thus he will enjoy a slight advantage over his 3.7 brethren below him.

you know its not true, stop make them advertise, the matchmaking now realy suck

[xBILLYJACKx 144]

I have all 3.0s across the board. I decided to take one away and add the Panzer IV F2 which is at 3.7. All of the sudden I'm fighting guys in 4.7 tanks. Yeah funk that. I'll take my toys and go home.

[blastedryan 967]

9 hours ago, TheBeast30 said:

you know its not true

Prove it with numbers not feelings then