Jump to content
Treeburst

So how random are the teams, really?

Recommended Posts

Posted this on the official forums. Need some smart folks to poke some holes in this. My numbers seem too small to me.

Ok, I'll ask anyone who knows statistics to point out any flaws in my logic here. I'm sure there will be some. If we assume that the NA population has a normally distributed WR (which it does not, there is a much longer tail out to the right, but I think it is close enough) with a mean WR of 48.83 and a std dev of 3, then approximately 58.84% of the population falls below 49.5% WR (XVM displays 49.5-50.49 as 50).

So the odds of any given player on your team being 49% or below is 58.84%. If we also make the assumption that each player is independent, then the odds of one player being on your team does not affect the odds of another player (so this excludes platoons, although I'm not convinced in the grand scheme of things this matters), then we have a collection of independent, "random" selections.

For independent events, the probability is pretty easy to calculate, since you can just multiply the individual probabilities. So for the odds of getting 14 teammates below 50% (well, below 49.5 really) the math looks like (.5884)*(.5884)*(.5884)...*(.5884). Or (.5884)^14. That comes out to .000596234767. So there is just about 1/2 of 1% chance of that happening, if I did my math right.

So what are the odds of that happening 2 games in a row? I believe you can treat the games as independent events as well, so you should be able to just multiply the probabilities again. Now two nights ago I had this happen 4 games in a row (1 tier nine game, 2 tier 7 games, and a tier 4 game, so it wasn't the same bunch of folks all getting in the same games). The odds of that happening seem like it would be (.000596234767)^4, or 0.000000000000126377333. Well, to put it in percentages, 0.0000000000126377333%.

Am I making any gross mathematical errors here? I've got a statistician working across the hall from me that I could bounce this off as well if anyone really feels like I messed up.

To put that last number in perspective, the odds of hitting the jackpot in the Powerball are 1/175,223,510. That is 0.000005707%.

So unless I did the math tragically wrong, you are something like 45,000 times more likely to win the lottery than you are to be the only player with a winning record on your team 4 games in a row. Well, if the teams were random anyway. Hows THAT for bad luck?

Link to post
Share on other sites

It really depends on which tier you are in, you are for example, much likely to get into a shitty game at tier 5 than tier 10 as the players are simply different.

Link to post
Share on other sites

They're random.

Randomly rigged. :D

 

Still, you have to consider the server population. The average is 49%, so there are a lot of players with that winning percentage. A bit random, but should be normal.

Link to post
Share on other sites

It really depends on which tier you are in, you are for example, much likely to get into a shitty game at tier 5 than tier 10 as the players are simply different.

 

That is indeed the big hole in Treeburst's theory. Quality players accumulate at higher tiers (particularly in tier IX and X). Newbies and perennially bad players are more frequent at lower tiers (particularly tiers I-IV).

 

If your statistics are supposed to reflect the player distribution truthfully you have to find a way get the correlation between average tier played and winrates into your reasoning.

 

Without any statistics it comes down to something like this:

  • At tier I-IV winrates are a flawed indicator for the quality of players since a large part of the players have such a low number of games that their winrates are not yet statistically significant.
  • At tier V-VI an average player will have ~48% winrate. 50%-ers are already quality players who can carry teams.
  • At tier VII-VIII the average player should have ~50% winrate. However numbers at these tiers are to some degrees distorted by a large number of baddies or newbies who buy into these tiers via premium tanks.
  • IX-X: from my experience players with green overall stats make up the bulk of most teams. Winrates are in the low 50ies (~52%).
  • Another factor to consider may be that at tier VIII and below deviations under the average seem more common  than deviations above the average. Whereas in high tier games I get the impression that horrible players and great players appear in a more balanced ratio.
Link to post
Share on other sites

For independent events, the probability is pretty easy to calculate, since you can just multiply the individual probabilities. So for the odds of getting 14 teammates below 50% (well, below 49.5 really) the math looks like (.5884)*(.5884)*(.5884)...*(.5884). Or (.5884)^14. That comes out to .000596234767. So there is just about 1/2 of 1% chance of that happening, if I did my math right.

 

...

To put that last number in perspective, the odds of hitting the jackpot in the Powerball are 1/175,223,510. That is 0.000005707%.

 

...

So unless I did the math tragically wrong, you are something like 45,000 times more likely to win the lottery than you are to be the only player with a winning record on your team 4 games in a row.

 

*snip*

 

Tried to disprove this using the law of large numbers and failed hard.  You have shit luck.

Link to post
Share on other sites

The people at different tiers have different W/R probability distributions. There would be much more <49.5%ers in tier those tier 4 and 7 games you were in, and much less in the tier 9.

 

The people that join your game 2 games in a row is also not independent, since most people will just queue right up after the game ends. They don't have to queue up the same tier either. With 2 tier spread, they could easily end up in your team again.

 

Platoons sort of screw up your math as well, because say player A and B are platooning together. Player A is a 45%er. P(PlayerB <= 49.5% | PlayerA = 45%) is probably greater than 58.84%, since bad people tend to play together.

 

As you said, your normal distribution assumption may not actually hold for this particular exercise, it's probably skewed to the left with a fatter right tail.

 

 

 

 

 

 

 

 

 

 

But yeah, your luck does sort of suck.

Link to post
Share on other sites

On the EU server at least, the median is substantially greater than the mean.

 

Source:  http://eumireo.wordpress.com/

 

Also, if there is a skilled platoon playing in your tiers at the same time, they are more likely to be on the other team. (Their team has 12 slots you could land in; the other team has 15 slots

Link to post
Share on other sites

Also, if there is a skilled platoon playing in your tiers at the same time, they are more likely to be on the other team. (Their team has 12 slots you could land in; the other team has 15 slots

 

This ... is such an insight.

Link to post
Share on other sites

I truly believe it goes by map. That's why it's never random. MM with actively rotate players around the map pool to create "diversity". that's why it often stacks the teams like no tomorrow.

 

for instance, south coast... 80% chance i'll be on the 40% OR LESS team. on that map.. consistently terrible chances. same on severogorsk.

 

meanwhile, on westfield, i'll often be on the stacked team. almost never do I get westfield repeatedly.

 

often, south coast is shoved down my throat 2-3x in a row. 

 

 

Link to post
Share on other sites

I truly believe it goes by map. That's why it's never random. MM with actively rotate players around the map pool to create "diversity". that's why it often stacks the teams like no tomorrow.

 

for instance, south coast... 80% chance i'll be on the 40% OR LESS team. on that map.. consistently terrible chances. same on severogorsk.

 

meanwhile, on westfield, i'll often be on the stacked team. almost never do I get westfield repeatedly.

 

often, south coast is shoved down my throat 2-3x in a row. 

 

Cite your sources and data. Anecdote holds no weight here.

Link to post
Share on other sites

All I know is, whenever I play tier 10s, there's a full Relic platoon on the other team, and two of my team's tier 10s are AFK.

 

Every damn time.

 

Oh well, at least I got to kill Loz once or twice.  That's always fun.

Link to post
Share on other sites

Cite your sources and data. Anecdote holds no weight here.

 

What are you talking about and what are you negging it for? I just told you my data. My source is me. You are free to run your own so we can compare note, because im sure you have access to all the inner workings of the game yourself to do it otherwise. Instead you opt to post stupid shit just for the sake of trolling.

 

This place is getting overrun with forumers it seems. No wonder even garbad wont come here no more. Im done, enjoy the official forums 2.0. Now with more stupid.

 

 

Link to post
Share on other sites

 It simply cannot be "random", because you're already starting with "rigged" numbers and a limited pool. 

 

All MM can do is try to be somewhat random with the players on hand within it's parameters (the rigged numbers), which means sure- you're going to see the same knuckleheads if you and they play the same tier bracket (and probably type of vehicle) repeatedly.   A low server population should skew it even more towards the likelyhood of seeing them repeatedly..

Link to post
Share on other sites

You didn't factor in platoons. The two/three players are not randomly selected individually but rather selected as a whole, therefore skewing.

Link to post
Share on other sites

What are you talking about and what are you negging it for? I just told you my data. My source is me.

 

Isn't this like the definition of anecdote?

 

If you want to actually make an argument:

- state a clear hypothesis

- write your observations in a notebook or otherwise record them (properly, no cheating)

- analyze your data to see whether the observations are correct. In this case, you need to show that it is beyond coincidence that teams on X map are bad more often, most likely using statistical means, since they are fairly accepted.

 

If you don't do this, your ideas will be dismissed as crackpot theories (as they should be) and in this case neg-bombed. Your claim is rather extraordinary, and will require the extraordinary evidence I am saying you should collect.

Link to post
Share on other sites

What are you talking about and what are you negging it for? I just told you my data. My source is me. You are free to run your own so we can compare note, because im sure you have access to all the inner workings of the game yourself to do it otherwise. Instead you opt to post stupid shit just for the sake of trolling.

 

This place is getting overrun with forumers it seems. No wonder even garbad wont come here no more. Im done, enjoy the official forums 2.0. Now with more stupid.

 

I didn't neg you, nor am I trolling. Don't throw out wild accusations and assumptions that have no basis in reality.

 

Having only your own experience with no hard data to back it up is the very definition of anecdote. What I expect to see out of data, even if it IS all from your own experiences (which is fine), is a log of every game with a sample size of at least 100 (which is tiny, but doable for a single person) with the map listed, time of day, screenshots of each team, etc. This isn't too much to ask for the claim you are making.

 

If you aren't going to do that then please do not throw around your confirmation bias as fact. That's for the official forums, not here, and that's why you're being neg repped by everybody else.

Link to post
Share on other sites

You are assuming the battle rate of bad players is the same as good ones.
 
Bad players die quicker so they may get more battles per hour, which may or may not be balanced out by the number of battles they play per day.
 
Anyone have any figures on battles per day for unicums vs terribad players?

Link to post
Share on other sites

Also, if there is a skilled platoon playing in your tiers at the same time, they are more likely to be on the other team. (Their team has 12 slots you could land in; the other team has 15 slots

 

No, because the matchmaker will try to match the number of platooned tanks per side, so if one team has a 3 man platoon, the other will have at least one platoon as well (sometimes it uses 2x2 if it can't find a 3 man).  So you still have a roughly even chance, when solopubbing, of the good platoon being on your side vs. the other.

 

 

 

Of course, what this means is that if one team has a purple platoon, the other team is more likely to get an average to bad platoon, because there are more average to bad players, so if the good platoon is on the opposite team it's going to be harder to counter them with your three bads than it might have been if those slots had been rolled seperately and pulled in one or more other good.

Link to post
Share on other sites

You totally ignore platoons, and MM treats them as single entity.

Add to this that platoons can rig the MM using the 2-sec hack.

Link to post
Share on other sites

If two platoons click Battle very close to each other, they are highly likely to be matched against each other.

 

Especially during off peak hours.

 

 

This is how a lot of competitions have been rigged.

Link to post
Share on other sites

No, because the matchmaker will try to match the number of platooned tanks per side, so if one team has a 3 man platoon, the other will have at least one platoon as well (sometimes it uses 2x2 if it can't find a 3 man).  So you still have a roughly even chance, when solopubbing, of the good platoon being on your side vs. the other.

 

This rule may not apply, when the server is low populated.  The server is likely to be low-populated over a consecutive sequence of battles, thereby increasing the odds that the OP (solo-pubbing, more likely to be off-peak) gets screwed by the MM.

Link to post
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.


  • Recently Browsing   0 members

    No registered users viewing this page.

×
×
  • Create New...