Jump to content


Team Win Rate and Victory!

win rate statistics your team sucked

  • Please log in to reply
26 replies to this topic

Mere_Anarchy #1 Posted 03 December 2014 - 08:51 PM

    First Sergeant

  • Players
  • 32057 battles
  • 1,118
  • [III]
  • Member since:
    07-07-2014

Thought I would offer up this little tidbit to give you some reading material while you are waiting for the server restart and your chance to play Malinovka for the first time or to start grinding the new line of British mediums.

 

The final bit of information I can pull from my sample of games involves matchmaking.  Specifically, I was curious to see what effect the overall Win Rate (WR) of a team would have on their chances of winning a battle. 

 

From the larger dataset I gathered I was able to look at the results of 35 matches played from Tiers 2/3 to 10.  I excluded battles with Tier 1 vehicles because – well, because it’s Tier 1.  Only one of these matches resulted in a draw.  The analysis did not take into account the mix of vehicles (heavy, medium, TD), AFK’s, platoons, or fail platoons (too complicated and too small of a dataset). 

 

Here’s what I found:

 

 

Mean WR

Median WR

WR Range

Mean Tier

Wins

50.59

50.80

35.5-71.1

6.39

Losses

48.54

48.24

32.3-68.4

6.35

 

Graphs of the Data

Spoiler

 

  • The winning teams’ mean Win Rate was significantly higher than that of the losing side.

Spoiler

 

  • The team with the higher mean Win Rate won 83% (29/35) of the matches; the team with the higher median Win Rate won 89% (31/35) of the matches.

Spoiler

 

  • There was no significant difference in the mean Tier composition of teams

Spoiler

 

  • The one draw had teams with average Win Rates of 47.04 and 48.71 and median Win Rates of 46.65 and 48.97, respectively.

 

I would not take these results to be completely representative of all WoTBlitz matches for two primary reasons:

  • the sample size is a bit on the small side. 

  • I can’t control for is the effect of my own participation.  I played in each one of the matches and obviously had an impact on the overall results.  It seems possible that my participation and overall WR (54%) could (and likely did) bias these results.  In that respect the mix of players in each match is not completely random.  But, given that I am only one of 7 players on a team (and am a long way from being a dominant player) I suspect the overall impact is minimal. 

 

Regardless of these factors, there is no doubt that there is a strong pattern that is well outside of the bounds of random variation: in a limited sample of my own battles my team’s overall win rate played a very large role in the outcome of a battle.   This really isn’t that earth-shattering of a revelation – its pretty much common sense.  I was surprised, however, with how strong the relationship was. 

 

So what might account for the ~10-15% of the time that the lower WR teams were victorious?  This is where a host of other relevant factors likely play a role: tank up grades, crew training, ability of a particular player with the tank they are using, the map, lag, mix of vehicles, platoons, fail platoons, AFK’s, etc. These are very difficult if not impossible to account for and measure. 

 

Just a final note – this is not in any way intended to be an argument for or against the current matchmaking system and whether or not player ability should also be used when setting up matches.  

 

Thanks and I hope to see you on the battlefield soon!

 

For those of you who are interested – the actual statistical analyses are in the spoiler below.

Spoiler

 


 

 


The_Gister #2 Posted 03 December 2014 - 08:53 PM

    Senior Sergeant

  • Players
  • 10051 battles
  • 784
  • [PRAMO]
  • Member since:
    04-27-2013

Wow... cool.

 

 

 

Spoiler

 


   

Just Wait until Clan Wars come out...

          


Zach_la #3 Posted 03 December 2014 - 08:59 PM

    Junior Sergeant

  • Players
  • 10147 battles
  • 150
  • Member since:
    06-27-2014
Wow, that's awesome you put that together! Very interesting and I like the graphs. +1

 

 

Ambassador for Teufel Hunden


hughjass07 #4 Posted 03 December 2014 - 09:00 PM

    Senior Sergeant

  • Club Wargaming
  • 13316 battles
  • 534
  • [HJG]
  • Member since:
    06-26-2014
Very interesting information!  Now, if we had an app on iOS/android that could access the data and quickly calculate the median win rate of each team at battle start, would we not have a crude beginning to XVM for WoT BLITZ?

On Twitter @HughbertJass

Though likely I go to my Doom, my heart is valiant, my conscience clear,

and before it is my time, by God they will know that I was there!


DrWylde #5 Posted 03 December 2014 - 09:04 PM

    First Sergeant

  • Players
  • 23471 battles
  • 1,353
  • [RBLN]
  • Member since:
    04-03-2014
Always suspected this to be true... now, it appears proven [ with perhaps some sampling error ].  Thanks Mr. Statistics!

IS-7, IS-4, IS-8, ST-I, IS-3, IS-6, KV-4

Obj. 268, Obj. 704, ISU-152

E 100, E 75, Tiger II

T-62A, T-54, T-44

T30, M103


LAPPELduvide #6 Posted 03 December 2014 - 09:05 PM

    PhD in Dakimakura Engineering and Design M.S. in Waifu Studies

  • Players
  • 24288 battles
  • 7,043
  • [ANEKI]
  • Member since:
    06-26-2014
You really like WoTB!

┬┴┬┴┤ (◕ᴗ◕✿├┬┴┬┴┤(ಠ﹏ಠ├┬┴┬

>>>Join [MAHOU/ANEKI] Discord!!!<<<

Best ANIME CLAN in the NA and a nice community for weebs. 


Aventre3 #7 Posted 03 December 2014 - 09:06 PM

    Senior Sergeant

  • Players
  • 24876 battles
  • 654
  • [III]
  • Member since:
    07-15-2014

Again great job!

 

I don't think MM should consider player skill. As frustrating as it maybe some time to watch a t95 and is7 sniping from the back, the randomness of battle and unpredictable teams is all apart of the game. Though I think that company battles and clan wars will provide the chance to play on more "even" footing. 


For advice and review of Premium Tanks, please see this post.


glass2707 #8 Posted 03 December 2014 - 09:17 PM

    First Sergeant

  • Players
  • 30004 battles
  • 3,070
  • [GRIM]
  • Member since:
    09-27-2014
The sample size is somewhat limiting but it is interesting.  Would be even more interesting to see how much the difference in average win rates mattered.  Like a 1 point difference made X, but a 5 plus difference made it a lock.  The percentages seem too high to me that simply the highest mean won 80%, though you get that with small samples.  Last nights loss fest on my T-44 turned a 75% win rate on that tank to around 55%.  this was immediately after upgrading the engine also so did that upgrade make the tank or me worse, doubtful.  It was likely the result of a small sample size.  I likely wasn't as good as the sample before nor as bad as that days sample.  Though who knows maybe I was actually worse and playing poorly that day for some reason.  

gffcom #9 Posted 03 December 2014 - 09:19 PM

    REDACTED

  • Players
  • 34810 battles
  • 5,323
  • [SRT-]
  • Member since:
    07-04-2014
Thanks! Your statistics posts are always informative and valued!

                                                         

Xs: IS-7, IS-4, FV215b, T110E5, E100, E50M, Obj 140, T-62A, M48A1 Patton, FV4202, Leopard 1, Obj 268

                                                                  SPARTA 

                     

                                                           


_TheMessiah #10 Posted 03 December 2014 - 09:20 PM

    First Sergeant

  • Players
  • 22825 battles
  • 1,475
  • [BNCR]
  • Member since:
    06-24-2014
:popcorn:

Mere_Anarchy #11 Posted 03 December 2014 - 09:21 PM

    First Sergeant

  • Players
  • 32057 battles
  • 1,118
  • [III]
  • Member since:
    07-07-2014

View PostThe_Gister, on 03 December 2014 - 02:53 PM, said:

Wow... cool.

Spoiler

 

Thanks?  

Spoiler

View PostZach_la, on 03 December 2014 - 02:59 PM, said:

Wow, that's awesome you put that together! Very interesting and I like the graphs. +1

Glad you liked it - thanks!

View Posthughjass07, on 03 December 2014 - 03:00 PM, said:

Very interesting information!  Now, if we had an app on iOS/android that could access the data and quickly calculate the median win rate of each team at battle start, would we not have a crude beginning to XVM for WoT BLITZ?

Thanks - like your videos BTW.  I suppose that would work - the access to the data would be the problem.  That and the temptation to go AFK if it looks like your team can't fight their way out of a wet paper bag.

View PostDrWylde, on 03 December 2014 - 03:04 PM, said:

Always suspected this to be true... now, it appears proven [ with perhaps some sampling error ].  Thanks Mr. Statistics!

Happy to help.

View Post_eL33t_, on 03 December 2014 - 03:05 PM, said:

You really like WoTB!

Not as much as I like statistics...

View PostAventre3, on 03 December 2014 - 03:06 PM, said:

Again great job!

 

I don't think MM should consider player skill. As frustrating as it maybe some time to watch a t95 and is7 sniping from the back, the randomness of battle and unpredictable teams is all apart of the game. Though I think that company battles and clan wars will provide the chance to play on more "even" footing. 

Thanks Aventre3!  Last one for a long time.  I should actually do some real work for a change.  I agree - as frustrating as MM can be, I like the chaos.  


Edited by vanpeenen, 03 December 2014 - 09:25 PM.

 

 


Captain_Dread #12 Posted 03 December 2014 - 09:22 PM

    Sergeant

  • Players
  • 40203 battles
  • 301
  • [III]
  • Member since:
    07-06-2014
Just my observation, which may likely be flawed, but if I score significant damage on one game, the next game there seems to be a better chance I will be high tier, and vice versa.  This would make sense, that you would want better players at the higher tiers to make the match more equal.  It also explains why some players always complain about being low tier, look at their stats.  This would be just one of the many factors, though, and could be null if only a certain number of that tiered tanks were in the que.

 

"It's not the size of the dog in the fight, it's the size of the fight in the dog."  Mark Twain

 

 


HisMaj #13 Posted 03 December 2014 - 09:32 PM

    First Sergeant

  • Players
  • 6851 battles
  • 2,001
  • Member since:
    07-17-2014
looks like I will only be on the winning side ~10% of the time...

I'm back! 

You still remember me right? 

No?

:(


dimzzy #14 Posted 03 December 2014 - 09:58 PM

    First Sergeant

  • Players
  • 38488 battles
  • 4,179
  • [-1]
  • Member since:
    06-29-2014

Do I understand this correctly that:

 

>> The team with the higher mean Win Rate won 83% (29/35) of the matches; the team with the higher median Win Rate won 89% (31/35) of the matches.

 

means that a team of good players will loose to a team of average players + very good player?



sheriffsherman #15 Posted 03 December 2014 - 10:57 PM

    First Sergeant

  • Players
  • 24534 battles
  • 4,674
  • Member since:
    10-06-2014

View Postdimzzy, on 03 December 2014 - 01:58 PM, said:

Do I understand this correctly that:

 

>> The team with the higher mean Win Rate won 83% (29/35) of the matches; the team with the higher median Win Rate won 89% (31/35) of the matches.

 

means that a team of good players will loose to a team of average players + very good player?

The way I interpret the information is that a solid team will overcome a mismatch in tier vehicles 80% of the time and 89% of the time when the tiers are even.  Please correct me if I'm wrong, but either way this is clear proof that a few WR points makes a huge difference in the qualiy of the player and teams.  I did take some statistics in college but that was a long long time ago.  Perhaps we can get more detailed analysis from the OP.


FIRST TO 200 TANKS ON BLITZ!!!

I'm a gen 1 gamer, been crushing bad guys since the 1970s


Mere_Anarchy #16 Posted 03 December 2014 - 11:19 PM

    First Sergeant

  • Players
  • 32057 battles
  • 1,118
  • [III]
  • Member since:
    07-07-2014

View Postdimzzy, on 03 December 2014 - 03:58 PM, said:

Do I understand this correctly that:

 

>> The team with the higher mean Win Rate won 83% (29/35) of the matches; the team with the higher median Win Rate won 89% (31/35) of the matches.

 

means that a team of good players will loose to a team of average players + very good player?

 

I would probably just say that the team that has the best overall win rate will win most of the time.  In this case median is a better way to measure what the 'overall' or group WR really is.  Having one really good player on your team could make the average, or mean, WR higher, but if the other players are just average or worse, and the other team has a bunch of overall better players they will likely win.  The two percentages I give that you quoted are related.  Two of the winning teams out of the 35 matches had a lower average win rate than their  competitors which made it seem like the worse team won.  But, if you look at the median WR for those two teams it actually was higher for the winning team.

 

 It all depends on the mix of players, but in some ways it could be the opposite of what you say.  For example, one of the losing teams had a player with a 68% WR, but because the rest of the team wasn't that good they didn't win.


 

 


DisordeR_ #17 Posted 03 December 2014 - 11:23 PM

    Senior Sergeant

  • Players
  • 30953 battles
  • 624
  • [-BA-]
  • Member since:
    06-27-2014
Just wanted to say thank you for your posts - well done.

dimzzy #18 Posted 04 December 2014 - 02:03 AM

    First Sergeant

  • Players
  • 38488 battles
  • 4,179
  • [-1]
  • Member since:
    06-29-2014

View Postvanpeenen, on 03 December 2014 - 11:19 PM, said:

 

I would probably just say that the team that has the best overall win rate will win most of the time.  In this case median is a better way to measure what the 'overall' or group WR really is.  Having one really good player on your team could make the average, or mean, WR higher, but if the other players are just average or worse, and the other team has a bunch of overall better players they will likely win.  The two percentages I give that you quoted are related.  Two of the winning teams out of the 35 matches had a lower average win rate than their  competitors which made it seem like the worse team won.  But, if you look at the median WR for those two teams it actually was higher for the winning team.

 

 It all depends on the mix of players, but in some ways it could be the opposite of what you say.  For example, one of the losing teams had a player with a 68% WR, but because the rest of the team wasn't that good they didn't win.

 

Oh, that's true--it is just the opposite of what I said: median is less affected by outliers so higher victory probability for the team with better median WR means that a team of good players should beat a team of not so good players + very good player. I think I also just realized that MM should not take player stats into account because as you would become better in the game it will become much more frustrating to you--it would be expected that you should carry the team every time...

glass2707 #19 Posted 04 December 2014 - 04:02 AM

    First Sergeant

  • Players
  • 30004 battles
  • 3,070
  • [GRIM]
  • Member since:
    09-27-2014
The variety of the games is what keeps it interesting.  Sometimes you cruise to victory, sometimes you struggle, and sometimes(for us non unicorns) there is nothing you can do.

Mere_Anarchy #20 Posted 04 December 2014 - 02:59 PM

    First Sergeant

  • Players
  • 32057 battles
  • 1,118
  • [III]
  • Member since:
    07-07-2014

View Postglass2707, on 03 December 2014 - 03:17 PM, said:

The sample size is somewhat limiting but it is interesting. Would be even more interesting to see how much the difference in average win rates mattered. Like a 1 point difference made X, but a 5 plus difference made it a lock. The percentages seem too high to me that simply the highest mean won 80%, though you get that with small samples. Last nights loss fest on my T-44 turned a 75% win rate on that tank to around 55%. this was immediately after upgrading the engine also so did that upgrade make the tank or me worse, doubtful. It was likely the result of a small sample size. I likely wasn't as good as the sample before nor as bad as that days sample. Though who knows maybe I was actually worse and playing poorly that day for some reason.

 

I thought about looking at just that fact - I'm sure there probably is some sort of threshold.  But, with the small sample I had it wasn't possible.  The other thing that would be interesting to see is the margin of victory (measured by tanks destroyed/surviving) relative to WR differences.  But - that is way beyond anything I feel like attempting.  I've experienced the same thing with my tanks and win rate.  I've gotten to the point that after I have them fully upgraded I figure that I need about 150 or so battles before their win rates stabilize.  I thought I was doing great in my E100 when I first got it b/c my win rate was really high (at least for me); but after about 100-125 battles my win rate became a bit more typical for me (low 50% ish).

 

View Postgffcom, on 03 December 2014 - 03:19 PM, said:

Thanks! Your statistics posts are always informative and valued!

 

Thanks - glad you find them interesting.

 

View PostCaptain_Dread, on 03 December 2014 - 03:22 PM, said:

Just my observation, which may likely be flawed, but if I score significant damage on one game, the next game there seems to be a better chance I will be high tier, and vice versa.  This would make sense, that you would want better players at the higher tiers to make the match more equal.  It also explains why some players always complain about being low tier, look at their stats.  This would be just one of the many factors, though, and could be null if only a certain number of that tiered tanks were in the que.

 

Its an interesting idea - I honestly don't know if this is the case or just random variation.  My understanding is that the match making doesn't take this stuff into account, but I honestly don't know if WG has ever actually made public the specific algorithm they use.  At least from my very initial analysis it seems like getting a balance of tiers seems to be the main emphasis.

 

View PostHisMaj, on 03 December 2014 - 03:32 PM, said:

looks like I will only be on the winning side ~10% of the time...

 

Not necessarily - depends on how the rest of your team is.  That's what makes it interesting is that its the specific mix of factors that ultimately determine winning or losing.  I always prefer to win - but overall like the battles best that I know I did a solid job (whether win or lose).  The ones that I win when I go out right away with little damage I find very unsatisfying.  I hate losing, but if I did a solid amount of damage and played to my utmost I can at least take away that consolation.

 

View Postsheriffsherman, on 03 December 2014 - 04:57 PM, said:

The way I interpret the information is that a solid team will overcome a mismatch in tier vehicles 80% of the time and 89% of the time when the tiers are even.  Please correct me if I'm wrong, but either way this is clear proof that a few WR points makes a huge difference in the qualiy of the player and teams.  I did take some statistics in college but that was a long long time ago.  Perhaps we can get more detailed analysis from the OP.

 

The interesting thing about this was I didn't really run into any serious tier mismatches in the games I analyzed.  They were all ridiculously closely matched.  This is not to say that I haven't been in battles with the crazy imbalances.  It would be interesting to see how those would play out relative to WR.  One of the battles I did record had concep in it as a Tier 8 with Tier 9 and 10 tanks.  Not surprisingly he ended up doing the most damage on our team.  Just goes to show how one really skilled player can overcome the Tier handicap.

 

View Postjmstichler, on 03 December 2014 - 05:23 PM, said:

Just wanted to say thank you for your posts - well done.

 

Thanks - glad you like them.  I appreciate the feedback!


 

 





1 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users