Jump to content


Replay analysis - what does this tell us about the MM?

Lets do science statistics

  • Please log in to reply
399 replies to this topic

Homer_J #161 Posted 12 July 2018 - 05:24 PM

    Field Marshal

  • Beta Tester
  • 29674 battles
  • 31,317
  • [WJDE] WJDE
  • Member since:
    09-03-2010

View PostLordMuffin, on 12 July 2018 - 03:32 PM, said:

 

Indeed, and this is due to the non-skillbased-matchmaker we have,

 

More to do with the non skilled playerbase.  Which is to be expected in what is after all a casual video game.  I certainly don't want anything where I actually have to try my hardest every battle.

m1x_angelico #162 Posted 12 July 2018 - 06:16 PM

    Warrant Officer

  • Player
  • 23119 battles
  • 873
  • [-VETO] -VETO
  • Member since:
    01-04-2015

View PostHomer_J, on 12 July 2018 - 05:24 PM, said:

More to do with the non skilled playerbase.  Which is to be expected in what is after all a casual video game.  I certainly don't want anything where I actually have to try my hardest every battle.

 

Well if the non-skilled player base would be allocated properly to the teams, we wouldn't actually have as many MM related treads on the forum.

 

Block Quote

 If you ask (me, at least) what is the best thing to do in order to win more, the answer is something that is very simple, but probably second only to "learn how to use a mouse and keyboard" in the result it provides:

Make the biggest positive impact you can,  as early as you can.

 

This would hold true if the remaining 14 players on your team would apply it. However, if the players don't or can't apply it, then it really only matters whether you skill level equals the combined skill level of the enemy players who are now rushing your position because your unskilled team mates are dead.



duijm #163 Posted 12 July 2018 - 06:21 PM

    Captain

  • Player
  • 44723 battles
  • 2,045
  • [RANGX] RANGX
  • Member since:
    10-18-2011

View Postm1x_angelico, on 12 July 2018 - 05:16 PM, said:

 

Well if the non-skilled player base would be allocated properly to the teams, we wouldn't actually have as many MM related treads on the forum.

 

 

This would hold true if the remaining 14 players on your team would apply it. However, if the players don't or can't apply it, then it really only matters whether you skill level equals the combined skill level of the enemy players who are now rushing your position because your unskilled team mates are dead.

 

No it wont. With Sbmm al winrates will go to 50% so all beter player will complain and leave the game.

If most players cant pull their weight in battle they really dont deserve better player to help boost their winrate.



Long_Range_Sniper #164 Posted 12 July 2018 - 06:24 PM

    General

  • Player
  • 33393 battles
  • 9,153
  • [-MM] -MM
  • Member since:
    04-04-2011

View Postm1x_angelico, on 12 July 2018 - 05:16 PM, said:

Well if the non-skilled player base would be allocated properly to the teams, we wouldn't actually have as many MM related treads on the forum.

 

Allocated using which method?



Homer_J #165 Posted 12 July 2018 - 07:02 PM

    Field Marshal

  • Beta Tester
  • 29674 battles
  • 31,317
  • [WJDE] WJDE
  • Member since:
    09-03-2010

View Postm1x_angelico, on 12 July 2018 - 06:16 PM, said:

 

Well if the non-skilled player base would be allocated properly to the teams, we wouldn't actually have as many MM related treads on the forum.

 

 

You've not actually looked at Baldrickk's stats have you?



OreH75 #166 Posted 12 July 2018 - 07:16 PM

    Captain

  • Player
  • 48807 battles
  • 2,289
  • [OXIDE] OXIDE
  • Member since:
    05-29-2013

View Postm1x_angelico, on 12 July 2018 - 06:16 PM, said:

 

Well if the non-skilled player base would be allocated properly to the teams, we wouldn't actually have as many MM related treads on the forum.

 

Well if WG stopped sending player names to the games and only displayed the tanks in the game so xvm would become useless we wouldnt actually have as many MM related threads on the forum.



Baldrickk #167 Posted 12 July 2018 - 08:08 PM

    Field Marshal

  • Player
  • 30494 battles
  • 14,619
  • [-TAH-] -TAH-
  • Member since:
    03-03-2013

View Postm1x_angelico, on 12 July 2018 - 06:16 PM, said:

 

Well if the non-skilled player base would be allocated properly to the teams, we wouldn't actually have as many MM related treads on the forum.

 

 

This would hold true if the remaining 14 players on your team would apply it. However, if the players don't or can't apply it, then it really only matters whether you skill level equals the combined skill level of the enemy players who are now rushing your position because your unskilled team mates are dead.

Players are distributed fairly which is fine with me.

I don't want a skill based MM. Which would be inherently unfair. 

There are strongholds and clan wars for those seeking that experience. 

 

As for the second part, the whole point is to decide the battle before your allies can die off.

You know how people deride redline camping until the team is dead, because despite being able to pick up damage at the end, they've lost the battle? 

The idea is to reverse that.

If you can cause the enemy to take early losses before your team take too many, your team is that much closer to that point where they can dominate the battle with ease. You only need a few extra kills in most cases as described in my earlier post.

 

View PostOreH75, on 12 July 2018 - 07:16 PM, said:

Well if WG stopped sending player names to the games and only displayed the tanks in the game so xvm would become useless we wouldnt actually have as many MM related threads on the forum.

Yeah, but it would feel like you are playing a solo game more. 

Even more importantly though,  the forumite mod would stop working! 


 

Etre_ #168 Posted 12 July 2018 - 09:44 PM

    Second Lieutenant

  • Player
  • 52703 battles
  • 1,336
  • [WEBOB] WEBOB
  • Member since:
    04-21-2014

View PostBaldrickk, on 08 July 2018 - 02:51 AM, said:

World of Tanks

 

It's a big game, with plenty of people playing it, but is is fair?

.

.

.

 

 

We found the guy who's in charge of powerpoints, excels and so on. 

Funny thing is, all despise these powerpoints but hey, what you gonna do ... a manager has to manage something. Is all in the artistic impression. 

 

 

Really now. Some explanation of the terms here wouldn't hurt. Like what are  μ, σ  and how are relevant ? Or you expect everyone around here to have notions of statistics. Also, make the god damn labels visible. Can't  read s...t !

Your chart skill are lacking bro ! I bet most of the people quit  reading after the first 2-3 charts. I did at least. 

 

Some extras:

How did you extract all the info from the replays ? (edit, saw it now, that github link)

The rating I suppose is the present rating not the one at the moment when the replay was taken. If so, are those charts involving ratings, relevant ? 



TANKOPPRESSION #169 Posted 12 July 2018 - 10:24 PM

    Warrant Officer

  • Player
  • 43488 battles
  • 826
  • [-SJA-] -SJA-
  • Member since:
    04-25-2012

View PostEtre_, on 12 July 2018 - 08:44 PM, said:

 

We found the guy who's in charge of powerpoints, excels and so on. 

Funny thing is,

 

Funny thing is , at the start of reading your post my heart was stopping . I thought you was going to say  : We found the guy who's in charge of the Matchmaker :

True story .



Homer_J #170 Posted 12 July 2018 - 10:53 PM

    Field Marshal

  • Beta Tester
  • 29674 battles
  • 31,317
  • [WJDE] WJDE
  • Member since:
    09-03-2010

View PostEtre_, on 12 July 2018 - 09:44 PM, said:

 

We found the guy who's in charge of powerpoints, excels and so on. 

Funny thing is, all despise these powerpoints but hey, what you gonna do ... a manager has to manage something. Is all in the artistic impression. 

 

 

Really now. Some explanation of the terms here wouldn't hurt. Like what are  μ, σ  and how are relevant ? Or you expect everyone around here to have notions of statistics. Also, make the god damn labels visible. Can't  read s...t !

Your chart skill are lacking bro ! I bet most of the people quit  reading after the first 2-3 charts. I did at least. 

 

Some extras:

How did you extract all the info from the replays ? (edit, saw it now, that github link)

The rating I suppose is the present rating not the one at the moment when the replay was taken. If so, are those charts involving ratings, relevant ? 

 

If you don't like the evidence because it does not fit your belief system then say so.

Zhul87 #171 Posted 13 July 2018 - 08:09 AM

    Corporal

  • Player
  • 21598 battles
  • 123
  • [TEC] TEC
  • Member since:
    05-31-2013

With the gathered data as a base, how would people rate the likelyhood of, lets say, 5-10 (mostly) consecutive wins followed by consecutive defeats of similar size, within 50 games and what would be the chance that such streaks would appear more than once in those 50 games?

 

Example:

V V V V V V D D D D D.....D D D D D D V D V D D D V V D V V V V V V V V V



jabster #172 Posted 13 July 2018 - 08:40 AM

    Field Marshal

  • Beta Tester
  • 12598 battles
  • 24,026
  • [WSAT] WSAT
  • Member since:
    12-30-2010

View PostZhul87, on 13 July 2018 - 07:09 AM, said:

With the gathered data as a base, how would people rate the likelyhood of, lets say, 5-10 (mostly) consecutive wins followed by consecutive defeats of similar size, within 50 games and what would be the chance that such streaks would appear more than once in those 50 games?

 

Example:

V V V V V V D D D D D.....D D D D D D V D V D D D V V D V V V V V V V V V

 

I dunno, what do you think?

Baldrickk #173 Posted 13 July 2018 - 08:40 AM

    Field Marshal

  • Player
  • 30494 battles
  • 14,619
  • [-TAH-] -TAH-
  • Member since:
    03-03-2013

View PostZhul87, on 13 July 2018 - 08:09 AM, said:

With the gathered data as a base, how would people rate the likelyhood of, lets say, 5-10 (mostly) consecutive wins followed by consecutive defeats of similar size, within 50 games and what would be the chance that such streaks would appear more than once in those 50 games?

 

Example:

V V V V V VD D D D D.....D D D D D D V D V D D D V V D V V V V V V V V V

Streaks happen naturally in a random system.

Beyond that, everything from internet conditions to mental state can have an effect on the outcome too, and one day you could be at the top of your game, while the next you end up half a second late to each decision you make.

 

The newest graph, showing just the raw battles in order can be used to see if there is a trend in the strength of the teams you are matched with/against.

The outcome is a little harder as it is only known for battles watched until the end.

 

And short streaks of 5?  I'd be more suprised worried if they didn't turn up regularly.

It would be evidence towards a system trying to control the results as opposed to streajs being an argument against it. 



Simeon85 #174 Posted 13 July 2018 - 09:01 AM

    Colonel

  • Player
  • 1097 battles
  • 3,593
  • Member since:
    04-19-2013

View PostBaldrickk, on 12 July 2018 - 04:36 PM, said:

I'd like to see it done as an experiment. 

 

 

The annoying thing is it wouldn't be hard, just tweak the maps to work for 15v15, slip them into the main rotation (could limit it to higher tiers initially if there were worries about view ranges/slower vehicles at lower tiers) and see what happens.

 

Especially as Grand Battles have largely been a failure, it would at least mean the map work didn't go to waste. 

 

But I suspect what Kozzy said is correct, stuff like awareness, view range, camo etc. comes into play and the average bob wouldn't cope. 



Kozzy #175 Posted 13 July 2018 - 09:23 AM

    Major

  • Player
  • 38855 battles
  • 2,705
  • [RINSE] RINSE
  • Member since:
    06-29-2011

As not to derail Baldrikk's excellent thread I have started a new one to continue the discussion involving grand battle maps here:

 


Edited by Kozzy, 13 July 2018 - 09:24 AM.


Baldrickk #176 Posted 13 July 2018 - 05:55 PM

    Field Marshal

  • Player
  • 30494 battles
  • 14,619
  • [-TAH-] -TAH-
  • Member since:
    03-03-2013

View PostEtre_, on 12 July 2018 - 09:44 PM, said:

 

We found the guy who's in charge of powerpoints, excels and so on. 

Funny thing is, all despise these powerpoints but hey, what you gonna do ... a manager has to manage something. Is all in the artistic impression. 

 

 

Really now. Some explanation of the terms here wouldn't hurt. Like what are  μ, σ  and how are relevant ? Or you expect everyone around here to have notions of statistics. Also, make the god damn labels visible. Can't  read s...t !

Your chart skill are lacking bro ! I bet most of the people quit  reading after the first 2-3 charts. I did at least. 

 

Some extras:

How did you extract all the info from the replays ? (edit, saw it now, that github link)

The rating I suppose is the present rating not the one at the moment when the replay was taken. If so, are those charts involving ratings, relevant ? 

 

Aparrently, Powerpoint and Excel is the best way to analyse a lot of data. Who knew?

 

μ and σ are standard variables representing mean and standard deviation.  You should be aware of what a mean is if you have finished primary school.  I can forgive standard deviation as it's a bit more of an advanced topic.  But you are sat in front of a computer connected to the internet, and google would have given you https://en.wikipedia...al_distribution as the first result had you searched for "μ σ graph".

Essentially, they describe the centre point and the width of the distribution shown in the graph - the orange line representing the "best fit".

At least you know now, this was a learning experience for you.

 

The labels are perfectly visible at 1:1 scale. Try zooming in?  I made the graphs large though because, well, there are a lot of data points on some of those graphs, and I wanted to be able to actually show them.

By all means, make changes to  the chart generation or raise issues against the project on github if you think it should be better.

 

The rating is indeed the current rating of each player.  Yes, it is still relvalent, because although not perfectly historically accurate, the matchmaker doesn't know whether players will continue to play in the future and how their stats will change if they do.  From a statistical point of view, any one of those players could become a Unicum later on, but there is an equal chance that any of them might improve that much.

 



m1x_angelico #177 Posted 13 July 2018 - 07:00 PM

    Warrant Officer

  • Player
  • 23119 battles
  • 873
  • [-VETO] -VETO
  • Member since:
    01-04-2015

View Postduijm, on 12 July 2018 - 06:21 PM, said:

No it wont. With Sbmm al winrates will go to 50% so all beter player will complain and leave the game.

If most players cant pull their weight in battle they really dont deserve better player to help boost their winrate.

 

So, you think all better players would leave the game, because they would have more challenge in the game? Although I don't agree with such conclusion, what you are actually saying, the more dumb WG makes this game, the more roflstomps there are, more better players will come? Well, that definitively one way to go...

 

View PostLong_Range_Sniper, on 12 July 2018 - 06:24 PM, said:

Allocated using which method?

 

If you recall, we already had this discussion. We didn't have final conclusions what would be the best approach but then we don't have to. In the end, the method that would be used will be developed by people who are actually paid to solve such issues and code it into the game. Unfortunately, I'm not one of them. 

 

View PostHomer_J, on 12 July 2018 - 07:02 PM, said:

You've not actually looked at Baldrickk's stats have you?

 

If we are talking about 20k+ games, off course I did. However, MM has changed in the meanwhile, so I'm reserving my final judgment until I get sufficient no. of my replays trough the tool. The results may show that my current perception/opinion is correct or incorrect.

 

View PostOreH75, on 12 July 2018 - 07:16 PM, said:

Well if WG stopped sending player names to the games and only displayed the tanks in the game so xvm would become useless we wouldnt actually have as many MM related threads on the forum.

 

I'm not sure how exactly XVM is affecting that 2/3rds of my or the opposing team die in 3-4 minutes. So, people would not see that they were in a "bad" team based on the statistics showed by XVM, but unless they are totally retarded, they would be able to deduce it based on the fact that 2/3rds of the team are dead so fast.

 

View PostBaldrickk, on 12 July 2018 - 08:08 PM, said:

Players are distributed fairly which is fine with me.

I don't want a skill based MM. Which would be inherently unfair. 

There are strongholds and clan wars for those seeking that experience. 

 

As for the second part, the whole point is to decide the battle before your allies can die off.

You know how people deride redline camping until the team is dead, because despite being able to pick up damage at the end, they've lost the battle? 

The idea is to reverse that.

If you can cause the enemy to take early losses before your team take too many, your team is that much closer to that point where they can dominate the battle with ease. You only need a few extra kills in most cases as described in my earlier post.

 

Although it is your right to think that MM is fair and not to want SBMM, I disagree with some of the things you said.

 

Why is SBMM unfair? Let us take football. When a certain team starts playing it plays in the lower leagues. When it gets better it automatically progresses to a higher league. Would it be fair for Barcelona to play with a 5th league Scottish team? Yes, it would be - to Barcelona, cause they would wipe the floor with them. Can you imagine Barcelona players saying they don't want to play in the Spanish Primera because it is not fair to play against Real Madrid. This is what you are effectively saying.

 

I do agree there are  strongholds and clan wars where you are more likely to find such balance. However, you can't play all tier tanks in this game modes. So, to grind the tanks I'm forced to go into a random match to be potentially paired with either Barcelona or a Scottish team. 

 

Now let us say that you are equally paired with Barcelona or a Scottish team, so your average is 50%. But, how it is fair to club baby seals or be the baby seal being clubbed, just because everyone's average with even out at 50%. Why not make the matches a bit more challenging and interesting? Why not make me wonder if this the game we win, or this is the game we loose?

 

I mean, we can agree or disagree on SBMM and it effect, or even if it's actually possible to make it operational, but I think WG at least owns us a try. If it fails, it fails.

 

EDIT: Dont mind the typos


Edited by m1x_angelico, 13 July 2018 - 07:02 PM.


Long_Range_Sniper #178 Posted 13 July 2018 - 07:21 PM

    General

  • Player
  • 33393 battles
  • 9,153
  • [-MM] -MM
  • Member since:
    04-04-2011

View Postm1x_angelico, on 13 July 2018 - 06:00 PM, said:

If you recall, we already had this discussion. We didn't have final conclusions what would be the best approach but then we don't have to. In the end, the method that would be used will be developed by people who are actually paid to solve such issues and code it into the game. Unfortunately, I'm not one of them. 

 

Which is why I mention the lack of a metric for SBMM that can be used. Because without that metric nobody can say either way as to whether a future application of SBMM would be good for the game. A bit like trying to claim absolutely that the game is rigged or isn't rigged without the actual numbers to show that's the case. I seem to remember you being fairly strong on the fact that people can't make an absolute conclusion because of the lack of data/evidence for rigging of MM.

 

In which case surely the same applies to any future SBMM because neither you, nor me, nor anyone else seems to be able to propose an effective skill metric.

 

If such a metric was available that when applied to SBMM ensured my efforts in game were commensurately rewarded in terms of XP/credits earned each game, AND my queue times were the same as any other player then I'd gladly look at that. A bit like people putting forward evidence that the game is rigged. I'd look at that as well.

 

But surely you must agree that in the absence of a suitable skill metric nobody can say that SBMM would be good for the game. Because if you use existing skill metrics and the current MM it would appear that the downsides outweigh the positives.



duijm #179 Posted 13 July 2018 - 07:35 PM

    Captain

  • Player
  • 44723 battles
  • 2,045
  • [RANGX] RANGX
  • Member since:
    10-18-2011

View Postm1x_angelico, on 13 July 2018 - 06:00 PM, said:

 

So, you think all better players would leave the game, because they would have more challenge in the game? Although I don't agree with such conclusion, what you are actually saying, the more dumb WG makes this game, the more roflstomps there are, more better players will come? Well, that definitively one way to go...

 

Although it is your right to think that MM is fair and not to want SBMM, I disagree with some of the things you said.

 

Why is SBMM unfair? Let us take football. When a certain team starts playing it plays in the lower leagues. When it gets better it automatically progresses to a higher league. Would it be fair for Barcelona to play with a 5th league Scottish team? Yes, it would be - to Barcelona, cause they would wipe the floor with them. Can you imagine Barcelona players saying they don't want to play in the Spanish Primera because it is not fair to play against Real Madrid. This is what you are effectively saying.

 

Now let us say that you are equally paired with Barcelona or a Scottish team, so your average is 50%. But, how it is fair to club baby seals or be the baby seal being clubbed, just because everyone's average with even out at 50%. Why not make the matches a bit more challenging and interesting? Why not make me wonder if this the game we win, or this is the game we loose?

 

 

No this is not what I am saying. For most players the challenge would be gone if because Sbmm their winrate would drop to 50%. No matter how hard you try you wont win more because mm would activly will give you worse teams if you are good and good teams if you are bad.

If bad and good players get same teams and the other side always a team with the same skill they will all win 50%.

 

Your Barcalona statement is not correct.  It would more like you are Messi. Barca wins so we remove modric and give you a noob player in return. Do you win again we remove again a good player from your team and give you a other noob.....just as long till you/Barca is just as good as all the other teams.  If you try harder and win more (challenge?) you will even get more noobs and your (messi) winrate will go down.

Does that sound fair?

 


Edited by duijm, 13 July 2018 - 07:36 PM.


Baldrickk #180 Posted 13 July 2018 - 07:36 PM

    Field Marshal

  • Player
  • 30494 battles
  • 14,619
  • [-TAH-] -TAH-
  • Member since:
    03-03-2013

View Postm1x_angelico, on 13 July 2018 - 07:00 PM, said:

 

So, you think all better players would leave the game, because they would have more challenge in the game? Although I don't agree with such conclusion, what you are actually saying, the more dumb WG makes this game, the more roflstomps there are, more better players will come? Well, that definitively one way to go...

 

 

If you recall, we already had this discussion. We didn't have final conclusions what would be the best approach but then we don't have to. In the end, the method that would be used will be developed by people who are actually paid to solve such issues and code it into the game. Unfortunately, I'm not one of them. 

 

 

If we are talking about 20k+ games, off course I did. However, MM has changed in the meanwhile, so I'm reserving my final judgment until I get sufficient no. of my replays trough the tool. The results may show that my current perception/opinion is correct or incorrect.

 

 

I'm not sure how exactly XVM is affecting that 2/3rds of my or the opposing team die in 3-4 minutes. So, people would not see that they were in a "bad" team based on the statistics showed by XVM, but unless they are totally retarded, they would be able to deduce it based on the fact that 2/3rds of the team are dead so fast.

 

 

Although it is your right to think that MM is fair and not to want SBMM, I disagree with some of the things you said.

 

Why is SBMM unfair? Let us take football. When a certain team starts playing it plays in the lower leagues. When it gets better it automatically progresses to a higher league. Would it be fair for Barcelona to play with a 5th league Scottish team? Yes, it would be - to Barcelona, cause they would wipe the floor with them. Can you imagine Barcelona players saying they don't want to play in the Spanish Primera because it is not fair to play against Real Madrid. This is what you are effectively saying.

 

I do agree there are  strongholds and clan wars where you are more likely to find such balance. However, you can't play all tier tanks in this game modes. So, to grind the tanks I'm forced to go into a random match to be potentially paired with either Barcelona or a Scottish team. 

 

Now let us say that you are equally paired with Barcelona or a Scottish team, so your average is 50%. But, how it is fair to club baby seals or be the baby seal being clubbed, just because everyone's average with even out at 50%. Why not make the matches a bit more challenging and interesting? Why not make me wonder if this the game we win, or this is the game we loose?

 

I mean, we can agree or disagree on SBMM and it effect, or even if it's actually possible to make it operational, but I think WG at least owns us a try. If it fails, it fails.

 

EDIT: Dont mind the typos

To reply to my bit specifically:

 

I'm not against leagues. Not from a fairness perspective, and I wouldn't mind being matched in one.

 

There are some significant differences between teams in football and in WOT though.

 

Football teams are selected and remain as a team. In WOT std battles, it's likely you will never see any of those players again.

 

How would you handle platoons in league games? Would you have bad players pulled up (fail toon) or pull good players down (seal clubbing time!)

 

Football games are prearranged and people will be ready for the aranged game. Will there be enough people at the top and bottom of the skill scale to make games quickly enough?

 

At the moment,  stats are global.  People compare cleanly with each other. 

Is a person with bad stats in a higher league better or worse than a player with good stats in a lower league?






2 user(s) are reading this topic

0 members, 2 guests, 0 anonymous users