Jump to content

Stomp Stats


38 replies to this topic

#1 VonBruinwald

    Member

  • PipPipPipPipPipPipPipPipPip
  • The Undisputed
  • The Undisputed
  • 3,460 posts
  • LocationRandis IV

Posted 20 February 2021 - 04:37 PM

Back in June Paul gave us some stomp stats and what he defined one as.

View PostPaul Inouye, on 03 June 2020 - 12:09 PM, said:

But first, let's look at some numbers. We mentioned stomp stats that are matches that end with a score of 12-4 or worse for the losing side. While this metric isn't perfect, it does give us a good baseline for comparison purposes.


Since this latest patch I've taken to logging all my matches as I was curious as to where we're at. Back in June the valves were tuned to an alleged 30% stomp-rate. Of course the valves have been relaxed since then....

So after 100 matches let's see where we're at:

Posted Image

Yep, that's right, a whopping 43% of matches ended in a stomp! Oddly enough, I haven't seen a single 12-0 since the patch where I had three of them on the last session I played preceding.

You might also notice it says "Kill Difference", 8 games ended in a "objective victory", that is to say the losing team still had mechs standing when the match ended. In those cases if the winning team has 8 or more kills than the enemy then I've counted it as a stomp, in truth that was only 3 out of the 8 matches and I've yet to encounter a game where the winning team has the least mechs (might have to make a special case of those).

I'm going to keep logging, and intend to update after 7 days.

#2 crazytimes

    Member

  • PipPipPipPipPipPipPipPip
  • Ace Of Spades
  • 1,374 posts

Posted 20 February 2021 - 05:32 PM

I stand by my previous claim that "close" games are indicative that everyone ran around doing their own thing, trading kills, and not working together. "Stomps" fall into two broad categories - one team nascared better than the other, or one team stayed as a team, focused fire, communicated and had a good time at the other teams expense.


I don't see trying to artificially create a situation where every game is "close" is somehow something to work towards. I care a bit about some kind of rough skill and tonnage balance, but other than that the outcome should be what the outcome is. Trying to balance a game to make trading kills rather than teamwork a end goal isn't a win for me.

#3 PocketYoda

    Member

  • PipPipPipPipPipPipPipPipPip
  • Shredder
  • Shredder
  • 4,141 posts
  • LocationAustralia

Posted 20 February 2021 - 05:42 PM

View Postcrazytimes, on 20 February 2021 - 05:32 PM, said:

I stand by my previous claim that "close" games are indicative that everyone ran around doing their own thing, trading kills, and not working together. "Stomps" fall into two broad categories - one team nascared better than the other, or one team stayed as a team, focused fire, communicated and had a good time at the other teams expense.


I don't see trying to artificially create a situation where every game is "close" is somehow something to work towards. I care a bit about some kind of rough skill and tonnage balance, but other than that the outcome should be what the outcome is. Trying to balance a game to make trading kills rather than teamwork a end goal isn't a win for me.


If one team works well and another is clue less, yes stomp.. but if both teams are good there generally isn't a stomp. but if both teams are clueless again generally no stomp. Its a double edged sword..

Edited by Samial, 20 February 2021 - 05:42 PM.


#4 Heavy Money

    Member

  • PipPipPipPipPipPipPipPip
  • The Marauder
  • 1,275 posts

Posted 20 February 2021 - 05:51 PM

I've seen very few stomps caused by the winning team working together very well. I've seen a ton of stomps caused by the losing team spreading out and not even standing near enough each other to be in the same battle, let alone do actual teamwork.

In other words, stomps are not so much caused by the virtues of the winner as by the total failures of the losers.

If this is the case, then attempting to fix it would be an exercise in trying to distribute terrible players.

#5 Anomalocaris

    Member

  • PipPipPipPipPipPipPip
  • Bad Company
  • 671 posts

Posted 20 February 2021 - 06:00 PM

View PostHeavy Money, on 20 February 2021 - 05:51 PM, said:

In other words, stomps are not so much caused by the virtues of the winner as by the total failures of the losers.

If this is the case, then attempting to fix it would be an exercise in trying to distribute terrible players.


I think there is a lot to suggest this statement is generally true. The distribution of skill in MWO doesn't look like a Bell Curve IMO, it's a short tail on the high skill side and very thick long tail on the low skill side, partially due to the skill level of incoming players.

From what I've observed, a low skill group aggregating in Soup Queue is likely to have a bigger overall effect on team success than a high skill group is. IOW, a low skill group is more likely to cause a loss/stomp than a high skill group is to cause a win/stomp. Anecdotal of course, and PGI will never share any data with us.....

#6 Heavy Money

    Member

  • PipPipPipPipPipPipPipPip
  • The Marauder
  • 1,275 posts

Posted 20 February 2021 - 06:17 PM

View PostAnomalocaris, on 20 February 2021 - 06:00 PM, said:


I think there is a lot to suggest this statement is generally true. The distribution of skill in MWO doesn't look like a Bell Curve IMO, it's a short tail on the high skill side and very thick long tail on the low skill side, partially due to the skill level of incoming players.

From what I've observed, a low skill group aggregating in Soup Queue is likely to have a bigger overall effect on team success than a high skill group is. IOW, a low skill group is more likely to cause a loss/stomp than a high skill group is to cause a win/stomp. Anecdotal of course, and PGI will never share any data with us.....


Yeah. So the issue is, how much are stomps caused by the randomness of the matchmaker? And how much by randomness of player choices? Let's consider a couple scenarios:

Say we have a group of 24 people, 12 of which are good for their tier and 12 of which are bad. They are then randomly distributed into teams. The team which gets more of the better players will generally win, within some margin. This is the model of how things work that most people seem to be operating on when discussing this stuff on the forums. The solution is to find ways for the matchmaker to better identify who is good and who is bad, and then distribute them more evenly. But what if this isn't what's actually causing the stomps?

Now consider that we have 24 people of all the same skill level, so their distribution between the teams doesn't matter. You drop them into a match. Each player on each team then randomly decides to go in direction A or B. On one team, it splits about evenly. On the other team, it splits 75/25%. So you end up with part of the map where 50% of team 1 is fighting 75% of team 2, and then you have the rest of the players scattered about, perhaps fighting each other. Team 2 will pretty much always win this, even though all the players in the match are about equally skilled players.

Now think about matches you've all been in. How often do you know that you're probably going to lose within the first 30 seconds just based on who is leaving spawn in which direction?

Of course, we can say that good players will know to stick together, and very good players will stick together AND do teamwork. But we also know that even good players do lots of strange stuff for various reasons. (For example, my performance is based heavily on time of day and how much i'm drinking. If you tried to look at my data and draw some conclusions, it won't work without controlling for this. And this is true of a lot of players.)

So in other words, people talk and think about this issue as if matches are decided based on distribution of player skill without taking into account the possibility of random factors. It'd be great if we had more data about stomp rates at different tiers. Obviously, stomp rates of t1's playing t5's is probably mostly skill difference and doesn't need much explaining. But what's the stomp rate of t5's vs t5's, and t1's vs t1's?

#7 Anomalocaris

    Member

  • PipPipPipPipPipPipPip
  • Bad Company
  • 671 posts

Posted 20 February 2021 - 06:46 PM

View PostHeavy Money, on 20 February 2021 - 06:17 PM, said:


So in other words, people talk and think about this issue as if matches are decided based on distribution of player skill without taking into account the possibility of random factors. It'd be great if we had more data about stomp rates at different tiers. Obviously, stomp rates of t1's playing t5's is probably mostly skill difference and doesn't need much explaining. But what's the stomp rate of t5's vs t5's, and t1's vs t1's?


You can really go down a rabbit hole if you wanted to.

But as a matter of engineering, you start with the biggest, easiest to solve problems first. We have a rudimentary skill measure in PSR (certainly not as good as it could be). So you select 24 players for a match and divide them as equally as possible.

Once you've done that, you have to say, "what's next?" Do we try and improve our skill measure, or do we start trying to account for other factors? For example, if a better measure of skill is off the table, do we attempt to balance teams by mech class, weight and even chassis? If there are 2 MC-IIs in the queue with similar skill, do we split them between teams? And where in the team building process does this decision get made?

And so on, and so on......

PGI has made it pretty clear there are some things they won't do (any more work on PSR....), so I tend to look at things they haven't rejected yet.

#8 Heavy Money

    Member

  • PipPipPipPipPipPipPipPip
  • The Marauder
  • 1,275 posts

Posted 20 February 2021 - 06:54 PM

View PostAnomalocaris, on 20 February 2021 - 06:46 PM, said:


You can really go down a rabbit hole if you wanted to.

But as a matter of engineering, you start with the biggest, easiest to solve problems first. We have a rudimentary skill measure in PSR (certainly not as good as it could be). So you select 24 players for a match and divide them as equally as possible.

Once you've done that, you have to say, "what's next?" Do we try and improve our skill measure, or do we start trying to account for other factors? For example, if a better measure of skill is off the table, do we attempt to balance teams by mech class, weight and even chassis? If there are 2 MC-IIs in the queue with similar skill, do we split them between teams? And where in the team building process does this decision get made?

And so on, and so on......

PGI has made it pretty clear there are some things they won't do (any more work on PSR....), so I tend to look at things they haven't rejected yet.


That's true. I guess my point is, how do we know when we're done improving the system? What's the proper stomp rate if teams are perfectly balanced in every other way? A lot of people would say 0%. So I guess my point is that's a giant assumption because it doesn't count for random factors and winner-takes-all effects.

I've seen games where the higher the tier of play, the more the game comes down to whoever makes the first mistake. Generally they are a standoff for awhile because everything is so equal. Then someone messes up, and now one team has a 1% advantage, but that breaks the tie and quickly gets capitalized on. ALL victories are stomps in those games because its so balanced and people are so good.

MWO of course has a billion factors going on and isn't a complete example of this. But we've all seen matches that went like that: nobody dies for ages, then an entire team crumbles. It was a super close match where everyone is nearly dead, but was still a "stomp" by the usual definition.

#9 Flyby215

    Member

  • PipPipPipPipPipPipPip
  • The Devoted
  • The Devoted
  • 894 posts
  • LocationThunder Bay

Posted 20 February 2021 - 09:24 PM

I record all my games looking for trends; here's my results so far: 46% stomp in a group out of 121 matches; 40% solo out of 82 matches for me this month . 39.8% stomp in group out of 236 matches, and 48.9% stomp out of 43 matches in solo queue for January.

Indeed there's lots of factors. That does not mean I am in any way happy and/or satisfied with percentages that high.

EDIT: Added total matches played for data context.

Edited by Flyby215, 21 February 2021 - 08:48 AM.


#10 PocketYoda

    Member

  • PipPipPipPipPipPipPipPipPip
  • Shredder
  • Shredder
  • 4,141 posts
  • LocationAustralia

Posted 20 February 2021 - 11:29 PM

View PostLockheed_, on 20 February 2021 - 09:50 PM, said:

There's another variable that you could throw into the mix. With no direct influence on what map is played, people can end up with horrible load outs for a certain map. the brawling assault on polar highlands or the LURM boat on Solaris comes to mind and besides the actually lack of suitability for the map, there's the psychological factor where people pretty much give up mentally right at the beginning of the match.


Also one that seems to plague me is networking issues, freezes, bad ping times and hit scan issues due to servers can also very much effect the out come of the match..

Having customers disconnect (usually an Assault) and others never getting in as the match starts already puts a team on a losing footing.

Sounds like a cop out true but this stuff happens nearly every single match for me.

Edited by Samial, 20 February 2021 - 11:31 PM.


#11 whoosher

    Rookie

  • 7 posts

Posted 21 February 2021 - 03:36 AM

View Postcrazytimes, on 20 February 2021 - 05:32 PM, said:

Trying to balance a game to make trading kills rather than teamwork a end goal isn't a win for me.


If the end goal is all about teamwork then let premade teams face only premade teams.

#12 Anomalocaris

    Member

  • PipPipPipPipPipPipPip
  • Bad Company
  • 671 posts

Posted 21 February 2021 - 04:17 AM

View PostHeavy Money, on 20 February 2021 - 06:54 PM, said:


MWO of course has a billion factors going on and isn't a complete example of this. But we've all seen matches that went like that: nobody dies for ages, then an entire team crumbles. It was a super close match where everyone is nearly dead, but was still a "stomp" by the usual definition.


Fair points.

However, I think we can agree that trying to increase competitiveness in matches (however that might occur or be measured) is a good thing. And compared the data PGI released almost a year ago during the queue merge and PSR reset, nothing they've done so far has made it any better. And, in fact, even according to their own data its gotten worse. Maybe a lot worse if the anecdotal evidence posted here is true (people experiencing stomp rates of 40%+).

#13 Willard Phule

    Member

  • PipPipPipPipPipPipPipPipPip
  • The 1 Percent
  • The 1 Percent
  • 3,920 posts
  • LocationThe Omega Company compound on Outreach

Posted 21 February 2021 - 06:02 AM

View Postwhoosher, on 21 February 2021 - 03:36 AM, said:


If the end goal is all about teamwork then let premade teams face only premade teams.


They tried that, but it turned out that not all premade teams are created equal and their population tanked because, oddly enough, the less skilled groups got tired of being farmed by the better ones.

Unfortunately for us solos, those comp group guys know how to communicate with the Developers on their official communication board at Reddit and got them to allow groups in the solo queue again. That's right, again. They did it once before and it was an unmitigated disaster and ended up with the creation of the group queue in the first place.

#14 The pessimistic optimist

    Member

  • PipPipPipPipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 2,377 posts

Posted 21 February 2021 - 06:11 AM

View PostWillard Phule, on 21 February 2021 - 06:02 AM, said:


They tried that, but it turned out that not all premade teams are created equal and their population tanked because, oddly enough, the less skilled groups got tired of being farmed by the better ones.

Unfortunately for us solos, those comp group guys know how to communicate with the Developers on their official communication board at Reddit and got them to allow groups in the solo queue again. That's right, again. They did it once before and it was an unmitigated disaster and ended up with the creation of the group queue in the first place.


Instead of making FP the team mode and leaving quick play for people who just want a casual mode. And the comp guys have comp play and 1v1 and 2v2 Solaris. PGI pretty much dose what ever those guys want. Leave people want to blow on some steam and blow up some mechs for fun out in the cold. And the soup mode gets abused as well

Edited by SirSmokes, 21 February 2021 - 06:20 AM.


#15 VonBruinwald

    Member

  • PipPipPipPipPipPipPipPipPip
  • The Undisputed
  • The Undisputed
  • 3,460 posts
  • LocationRandis IV

Posted 21 February 2021 - 06:32 AM

View PostHeavy Money, on 20 February 2021 - 06:17 PM, said:

Yeah. So the issue is, how much are stomps caused by the randomness of the matchmaker? And how much by randomness of player choices? Let's consider a couple scenarios:


Rather than considering individual scenarios/matches you have to look at the overall trend. I was actually disappointed I didn't have any 12-0 matches when gathering this data. Going from Paul's numbers:

View PostPaul Inouye, on 03 June 2020 - 12:09 PM, said:

Prior to queue merge, our sample set gave us this:

1123 Stomps out of 3755 matches in an average 24hr period resulting in a 29.91% stomp rate.

Post queue merge, our sample set gave us this:
935 Stomps out of 2751 matches in an averages 24hr period resulting in a 33.99% stomp rate.

[After tweaks to Tier/PSR and tonnage balance in the match maker]:
1057 Stomps out of 3453 matches in an average 24hr period resulting in a 30.61% stomp rate.

Essentially, overall stomp numbers increased by only 0.7% compared to pre-queue merge.


Essentially, with proper MM tweaks we have a 30% stomp rate, but since the valves were relaxed we appear to have shot up to 43%.

What that indicates is that the +13% in stomps is due to poor MM decisions. How much of that initial 30% is down to the MM is a different question.



View PostFlyby215, on 20 February 2021 - 09:24 PM, said:

I record all my games looking for trends; here's my results so far: 46% stomp in a group; 40% solo for me this month. 39.8% stomp in group, and 48.9% stomp in solo queue for January.


How big is your dataset? Curious to see how the charts compare with regards to distribution. If you got it all typed out and can copy/paste PM me.

#16 JediPanther

    Member

  • PipPipPipPipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 4,087 posts
  • LocationLost in my C1

Posted 21 February 2021 - 06:43 AM

After watching recent matches it still comes down to two things.

1. How many meta mechs a side has.
2. How much team work if any a side uses.

I find the side that gets to three kills first wins almost all the time. Where are the stats on that? How high is the percent of a 0-3 loosing team wining over that 3-0 coordinated team?

#17 The pessimistic optimist

    Member

  • PipPipPipPipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 2,377 posts

Posted 21 February 2021 - 06:46 AM

View PostJediPanther, on 21 February 2021 - 06:43 AM, said:

After watching recent matches it still comes down to two things.

1. How many meta mechs a side has.
2. How much team work if any a side uses.

I find the side that gets to three kills first wins almost all the time. Where are the stats on that? How high is the percent of a 0-3 loosing team wining over that 3-0 coordinated team?

It's also game modes in the way game modes are set up. Currently the preferred way to win a match is just destroy the enemy team regardless of objectives. Change game modes man the primary way to win is just not eliminating the other team. Top of maybe reducing the number of players from 12 to 8 on each team. Right now there's very little incentive for people to do anything other than death ball an NASCAR

Edited by SirSmokes, 21 February 2021 - 06:48 AM.


#18 Dozer6

    Member

  • PipPipPipPipPip
  • Big Brother
  • Big Brother
  • 139 posts
  • LocationStripping lasers off my Archer to fit 4 more tons of LRM'S

Posted 21 February 2021 - 06:54 AM

View PostHeavy Money, on 20 February 2021 - 06:17 PM, said:


Yeah. So the issue is, how much are stomps caused by the randomness of the matchmaker? And how much by randomness of player choices? Let's consider a couple scenarios:

Say we have a group of 24 people, 12 of which are good for their tier and 12 of which are bad. They are then randomly distributed into teams. The team which gets more of the better players will generally win, within some margin. This is the model of how things work that most people seem to be operating on when discussing this stuff on the forums. The solution is to find ways for the matchmaker to better identify who is good and who is bad, and then distribute them more evenly. But what if this isn't what's actually causing the stomps?

Now consider that we have 24 people of all the same skill level, so their distribution between the teams doesn't matter. You drop them into a match. Each player on each team then randomly decides to go in direction A or B. On one team, it splits about evenly. On the other team, it splits 75/25%. So you end up with part of the map where 50% of team 1 is fighting 75% of team 2, and then you have the rest of the players scattered about, perhaps fighting each other. Team 2 will pretty much always win this, even though all the players in the match are about equally skilled players.

Now think about matches you've all been in. How often do you know that you're probably going to lose within the first 30 seconds just based on who is leaving spawn in which direction?

Of course, we can say that good players will know to stick together, and very good players will stick together AND do teamwork. But we also know that even good players do lots of strange stuff for various reasons. (For example, my performance is based heavily on time of day and how much i'm drinking. If you tried to look at my data and draw some conclusions, it won't work without controlling for this. And this is true of a lot of players.)

So in other words, people talk and think about this issue as if matches are decided based on distribution of player skill without taking into account the possibility of random factors. It'd be great if we had more data about stomp rates at different tiers. Obviously, stomp rates of t1's playing t5's is probably mostly skill difference and doesn't need much explaining. But what's the stomp rate of t5's vs t5's, and t1's vs t1's?

Yup, if you see the gin and tonic on the console in my cockpit we're in trouble.

#19 JediPanther

    Member

  • PipPipPipPipPipPipPipPipPip
  • Legendary Founder
  • Legendary Founder
  • 4,087 posts
  • LocationLost in my C1

Posted 21 February 2021 - 06:55 AM

View PostSirSmokes, on 21 February 2021 - 06:46 AM, said:

It's also game modes in the way game modes are set up. Currently the preferred way to win a match is just destroy the enemy team regardless of objectives. Change game modes man the primary way to win is just not eliminating the other team. Top of maybe reducing the number of players from 12 to 8 on each team. Right now there's very little incentive for people to do anything other than death ball an NASCAR


You also have to add on the facts:

The game rewards damage (greatly) over any thing else.
Reward for an objectives like base capture is nothing at 100cp 0 xp when you can get more just by doing ten damage only.
Spawns are horrible;being in assult spawn in dom mode so far out in your 40-50kph mech means you will hardly be out of the spawn grid before the match is over if you don't have a light on the p stain.
Map design is focused around a specific location or feature reinforcing nascar behavior like "pride rock" on Canyon Network D3.

#20 Dozer6

    Member

  • PipPipPipPipPip
  • Big Brother
  • Big Brother
  • 139 posts
  • LocationStripping lasers off my Archer to fit 4 more tons of LRM'S

Posted 21 February 2021 - 07:03 AM

View PostSirSmokes, on 21 February 2021 - 06:46 AM, said:

It's also game modes in the way game modes are set up. Currently the preferred way to win a match is just destroy the enemy team regardless of objectives. Change game modes man the primary way to win is just not eliminating the other team. Top of maybe reducing the number of players from 12 to 8 on each team. Right now there's very little incentive for people to do anything other than death ball an NASCAR

Agree with you 100% Smokey boy!
Every match becomes skirmish, especially on weekends when i believe there are more C.O.D. mindset kiddies on running around all yologggetwrektlookatmystatsk/dqqtryhard mode.
They may be a decent player but get killed by a group of sub par players that stay together and focus fire very quickly.

And with the above mentioned mindset makes it very hard to get a group together that THINK about what they are doing and work together.
Nevermind combat doctrines and principles...





1 user(s) are reading this topic

0 members, 1 guests, 0 anonymous users