hideyourkids

Team death ratios- are roflstomps too common?

8 posts in this topic

Has anyone tried modeling the team death ratios? (So if everyone was equally skilled in a match, fix a chance p for any given player to die. Then when a player dies, alter the chance of any player on that team dying by some factor depending on the ratio of alive to dead members on both teams. Of course this gets more complicated with skill level distribution, but I'd be curious to see the results from just taking some weighted average of player skill and assigning every player a chance to die based on that). Using this model, it'd be easy to see why roflstomps occur, especially if a death confers a huge penalty to the chance of members on one team surviving. 

An even simpler model would be this- we can consider each team as 15 bernoulli trials, we just get a binomial distribution B(15,p) (and find p from maximum likelyhood given data). Now if we assume that p is the same for both teams, then it's pretty easy to find the probability of a 15-1 match, or 15-7 match (just assume both teams's deaths are independent multiply binomial probabilities). I think this model is the one people have in mind when the think of "fair"?

I keep reading in the forums that matches are unfair and unbalanced because there are too many roflstomps. Where can one get data on the team death ratios? Is that information stored somewhere? has someone done this analysis?

 

 

Share this post


Link to post
Share on other sites

I don't have any data, but I have the impression that roflstomps are most common at the very low tiers (1-3) and the very high tiers (9-10). 

Share this post


Link to post
Share on other sites

I did this, and posted on the official forums:

 

http://forum.worldoftanks.com/index.php?/topic/482441-blowouts-numerical-results/

 

The results were that for perfectly balanced teams the chance of a blowout is about 20% (or 1 in 5 battles). A blowout was defined as a result worse than 15-5. For badly imbalanced teams the probability rose to about 33%.

 

In the thread someone else posted his personal statistics he has kept and showed that they were similar to what I calculated.

 

Share this post


Link to post
Share on other sites
On 3/17/2016 at 8:47 AM, lt_lolcat said:

I don't have any data, but I have the impression that roflstomps are most common at the very low tiers (1-3) and the very high tiers (9-10). 

It's actually the other way around.  Higher tiers generate more stomps.  My guess is that once the ball gets rolling it's harder to stop.  In the low tiers outnumbered teams can still put up a decent last stand.

 

 

Share this post


Link to post
Share on other sites
2 hours ago, NeatoMan said:

It's actually the other way around.  Higher tiers generate more stomps.  My guess is that once the ball gets rolling it's harder to stop.  In the low tiers outnumbered teams can still put up a decent last stand.

It also seems to happen faster. At 1:55 we're even at 1:1 and moving into what looks to be a decent position. At 2:15 we're down 10:2 and they've started to cap. You blink, and all the green dots are just gone.

It didn't fell as bad back when the other team was still using 75mm and 90mm guns. All those fast 128mm guns out there now, two shots to the face and you're dead, and any open flank rapidly turns into a game-ending crossfire. We used to have time to react to map changes, now it seems like it's over before I can reload. My last four games have all been 10-12 tanks down before 2:30.

Share this post


Link to post
Share on other sites
On 3/22/2016 at 0:40 PM, lesterquaestor said:

I did this, and posted on the official forums:

 

http://forum.worldoftanks.com/index.php?/topic/482441-blowouts-numerical-results/

 

The results were that for perfectly balanced teams the chance of a blowout is about 20% (or 1 in 5 battles). A blowout was defined as a result worse than 15-5. For badly imbalanced teams the probability rose to about 33%.

 

In the thread someone else posted his personal statistics he has kept and showed that they were similar to what I calculated.

 

You are using voodoo math.

 

To start, you put your assumptions before any empirical results, and then you don't have any empirical results to back up your assumptions or model.

 

Other people have tried to force fit lenchesters square law in a similar manner as you, once again an epic fail.

 

There is no simple ( lazy) way around modeling this. It starts with large data mining. I would probably build a markov process designed to be calibrated to sample paths and end results. And then use the resulting empirical geometry to look for simplifications that aren't egregious.

 

 

Share this post


Link to post
Share on other sites

I've collected Margin of victory data (difference in surviving players between teams) from 6k of my games going all the way back to patch 8.0.  It mirrors what Lester posted quite well.

 

Margin per patch (version)

marginversion_zpsgw6fmlme.jpg

 

Margin per Battle Tier

MarginBT_zpszgnax8ii.jpg

 

Margin vs win chance

winchanceMargin_zps1sjgh386.jpg

 

 

 

Share this post


Link to post
Share on other sites

  • Recently Browsing   0 members

    No registered users viewing this page.