101 replies to this topic

[Vlad Ward]

VonBruinwald, on 20 January 2021 - 11:29 AM, said:

Fixed.

And it's still better overall for the game.

[Knight Captain Morgan]

VonBruinwald, on 20 January 2021 - 09:58 AM, said:

Should have limited group size to 2 OR 4 in group queue, this would have broken up the larger groups stomping and helped the MM fix matches faster since 2 and 4s can be easily combined to get 12.

I agree, but that's not what they did and all the "groups" praised them for it... since what they did let "groups" farm solo pugs which is what the "groups" were gunning for all along...

Edited by Knight Captain Morgan, 20 January 2021 - 12:51 PM.

[Bennesto]

Knight Captain Morgan, on 20 January 2021 - 12:49 PM, said:

I agree, but that's not what they did and all the "groups" praised them for it... since what they did let "groups" farm solo pugs which is what the "groups" were gunning for all along...

No, they did not. This is a fairy tale.

[GARION26]

VonBruinwald, on 20 January 2021 - 11:29 AM, said:

If memory serves, Their final number was 1% and their definition was a 10 kill difference.

I hear it bandied about on the forum quite a bit so I've been trying to find the actual frequently referenced difference
Here's the post I can find:
https://mwomercs.com...anges-jun-2020/

"Essentially, overall stomp numbers increased by only 0.7% compared to pre-queue merge."
" stomp stats that are matches that end with a score of 12-4 or worse for the losing side. "

The baseline rate of stomps prior to the Queue merger was just under 30% we added 0.7% to that with the final version of queue system - and that was prior to PSR reset/update - which presumably would have reduced stomps.

That 0.7% increase on a background of 30% would be very hard to notice as a player - you'd have to play statistically around 142 matches to have it impact one of your matches based on a Number Needed to Harm calculation (remember it's percentile based rates if you do the math yourself)
https://en.wikipedia..._needed_to_harm

Now whether queue merge has worsened quality for other reasons then stomps - that's a different discussion but the quoted stomp percent change is so small it's hard to imagine it being apparent in actual perceived quality of matches.

Edited by GARION26, 20 January 2021 - 02:03 PM.

[Vlad Ward]

GARION26, on 20 January 2021 - 01:29 PM, said:

I hear it bandied about on the forum quite a bit so I've been trying to find the actual frequently referenced difference
Here's the post I can find:
https://mwomercs.com...anges-jun-2020/

"Essentially, overall stomp numbers increased by only 0.7% compared to pre-queue merge."
" stomp stats that are matches that end with a score of 12-4 or worse for the losing side. "

The baseline rate of stomps prior to the Queue merger was just under 30% we added 0.7% to that with the final version of queue system - and that was prior to PSR reset/update.

That 0.7% increase on a background of 30% would be very hard to notice as a player - you'd have to play statistically around 142 matches to have it impact one of your matches based on a Number Needed to Harm calculation (remember it's percentile based rates if you do the math yourself)
https://en.wikipedia..._needed_to_harm

Now whether queue merge has worsened quality for other reasons then stomps - that's a different discussion but the quoted stomp percent change is so small it's hard to imagine it being apparent in actual perceived quality of matches.

Thank goodness someone brought facts into play.

12-4 is actually a pretty liberal definition of a stomp, given how games tend to collapse after the first few kills.

[VonBruinwald]

GARION26, on 20 January 2021 - 01:29 PM, said:

[snip]

Thanks. Was going from memory and didn't think to look through announcements:

Paul's numbers (If you didn't read/understand the thread):

Pre-merge stomp rate was 29.91%
Final stomp rate was 30.61% (+0.7)

These numbers are based on a 24hr period sample.

This was all pre-PSR reset and the stomp rate/PSR monitoring was to be followed up on.... that was October '20...

[GARION26]

VonBruinwald, on 20 January 2021 - 02:04 PM, said:

Thanks. Was going from memory and didn't think to look through announcements:

You are welcome. I found it last week because I kept hearing people on this fervently referencing 'data' that no one actually linked to or could give actual specifics of.

I only found it through google using various search terms. If there is another announcement or post out there that people are vaguely recalling (or misremembering) then I didn't find it.

It's just so odd that certain statements of 'PGI said X happened by their own numbers' are made so definitively - the data doesn't actually seem to show it.

[VonBruinwald]

GARION26, on 20 January 2021 - 02:10 PM, said:

You are welcome. I found it last week because I kept hearing people on this fervently referencing 'data' that no one actually linked to or could give actual specifics of.

I only found it through google using various search terms. If there is another announcement or post out there that people are vaguely recalling (or misremembering) then I didn't find it.

It's just so odd that certain statements of 'PGI said X happened by their own numbers' are made so definitively - the data doesn't actually seem to show it.

The 1% I had in mind came from the Combined queues - final discoveries thread which pre-dates that one by a month.

Short of going through each PGI employees account and going back through their posts one-by-one you're not going to find much. Most of their comments are throw away responses in response to questions asked and some bits of information come out through twitter and discord... tracking it all down would be a pain.

One thing we do know is this was all due for review 3 months ago. Using data from the first 24hrs doesn't necessarily provide a representative example once the changes have set in.

I suggest someone with a Twitter asks Russ to do that review.

[Anomalocaris]

GARION26, on 20 January 2021 - 01:29 PM, said:

I hear it bandied about on the forum quite a bit so I've been trying to find the actual frequently referenced difference
Here's the post I can find:
https://mwomercs.com...anges-jun-2020/

"Essentially, overall stomp numbers increased by only 0.7% compared to pre-queue merge."
" stomp stats that are matches that end with a score of 12-4 or worse for the losing side. "

The baseline rate of stomps prior to the Queue merger was just under 30% we added 0.7% to that with the final version of queue system - and that was prior to PSR reset/update - which presumably would have reduced stomps.

That 0.7% increase on a background of 30% would be very hard to notice as a player - you'd have to play statistically around 142 matches to have it impact one of your matches based on a Number Needed to Harm calculation (remember it's percentile based rates if you do the math yourself)
https://en.wikipedia..._needed_to_harm

Now whether queue merge has worsened quality for other reasons then stomps - that's a different discussion but the quoted stomp percent change is so small it's hard to imagine it being apparent in actual perceived quality of matches.

You forgot one thing. After that post, Paul eventually loosened the matchmaker back up due to increased match making time (you'll have to search twitter for that I believe). Hence one reason why we have such horrible tonnage imbalances in so many matches. Of course they haven't posted anything about matchmaking or match quality since then.

Regardless, thanks to PGI's stretching of the truth about putting resources into the game again (which was pretty much hiring Daeron to put a familiar face on their efforts to boost playerbase pre-acquisition), the playerbase is now operating at peak player numbers twice that of the historic low in February (almost 3x as much in December, but holiday freebie boosts don't really count). Average numbers (excluding December) are also twice as high. And the playerbase numbers are at June 2019 levels (and in fact close to August 2018 numbers according to Jarl's).

At any given time the minimum number of players online exceeds 650 (2x steam numbers - Russ says non-steam players can account for even more than 50%) and the maximum exceeds 2200. Even if groups make up only 10% of players, that's enough for 3-9 group only matches at any given time (or 4-14 if we go 8v8 which PGI has shown can be done easily). And given that there was no perceived need to merge the queues back in June 2019, I suggest there is no need now. Unless you believe that player numbers will crater again once people figure out that PGI fed them a line of b.s. about game improvements.

And here's the kicker. If you believe that groups are 10% or less of the community, then we shouldn't be catering to them by borking the solo queue matchmaker trying to mix them in, they're basically insignificant. If you believe they are a substantially larger part of the community, then there is clearly more than enough of them for a group only queue. Pick your poison.

It's pretty easy to see a 3 mode MWO

1. Solo queue - drop whenever and whatever you want. Random teams go.
2. 8v8 group queue - groups of 2 or 4. Play with a friend, drop with a lance, practice your small unit tactics
3. 12v12 faction - Any size, for serious players only Drop deck and tactics/strategy at the highest level.

Why one might even say that seems like a logical progression for players as they get more skilled/serious at the game. And if you allowed solo opt-ins to group queue, I'm sure you'd get some small percentage of solo players participating, which makes it even more viable.

[The pessimistic optimist]

Anomalocaris, on 20 January 2021 - 06:02 PM, said:

1. Solo queue - drop whenever and whatever you want. Random teams go.
2. 8v8 group queue - groups of 2 or 4. Play with a friend, drop with a lance, practice your small unit tactics
3. 12v12 faction - Any size, for serious players only Drop deck and tactics/strategy at the highest level.

Yup

[GARION26]

Anomalocaris, on 20 January 2021 - 06:02 PM, said:

You forgot one thing. After that post, Paul eventually loosened the matchmaker back up due to increased match making time (you'll have to search twitter for that I believe). Hence one reason why we have such horrible tonnage imbalances in so many matches. Of course they haven't posted anything about matchmaking or match quality since then.

Agreed there were some tweaks to the matchmaker and it's possible stomp rate has gone up significantly with those subsequent changes. But I can't find ANY data from PGI that says that.

Here's Paul's twitter account. He doesn't as far as I can see have any mention of stomp rate on there after the data I've linked to year.https://twitter.com/paul_inouye?lang=en

One thing is very clear - we had a 30% stomp rate PRIOR to introducing groups. That's a very large percent of matches. It's possible groups made that worse significantly (the data we have seen don't show that.) But it's very clear that stomp rate prior to groups is very high - possibly for reasons intrinsic to MWO others have noted. Possibly because the matchmaker was never that good.

If stomp rate went up post groups did it go up by enough of a percentage for it to be actually noticeable in the number of games played by the average person? If it was a 10% absolute rate increase (twice the temporary worst case increase in the data I can actually find referenced) you'd have to play a lot of matches per night to have an extra stomp (ten matches, and for a stomp against a particular player roughly 20). And it's pretty unlikely you'd notice that difference until you got to a much higher number of drops.
I'm not arguing here that we should or shouldn't have groups or what the best format is. I am saying if we're going to talk about it - at least we should understand the actual data we have been presented. Certainly if there is data I've missed in my search lets look at it and inform our discussions with actual numbers.

Note if we think the average drop has a group of 4 or a group of 2 and 3 then at least 33% of the people in any given match are 'in groups' that's a pretty big chunk of the population.

@VonBruinwald agree I'd like to see more of the data. 1% increase vs 0.7% is pretty much the same either way so your recall is fine. Either way it's a very small absolute increase on a background rate of around 30%

I agree it's hard it would be find all the little data nuggets hidden in various formats - but at least we can use the actual data we can find to inform these discussions. I'd love to see an update.

Data methodology in general for the samples we are quoting isn't great the 24 hour look back and look forward is a very brief period to look at for outcomes that have a lot of variables (weekend matchmaker accuracy is probably better then weekday, non event day is probably better then event, post new mech sale worse then without recent sale)

Edited by GARION26, 21 January 2021 - 09:29 AM.

[The pessimistic optimist]

GARION26, on 21 January 2021 - 09:24 AM, said:

Agreed there were some tweaks to the matchmaker and it's possible stomp rate has gone up significantly with those subsequent changes. But I can't find ANY data from PGI that says that.

Here's Paul's twitter account. He doesn't as far as I can see have any mention of stomp rate on there after the data I've linked to year.https://twitter.com/paul_inouye?lang=en

One thing is very clear - we had a 30% stomp rate PRIOR to introducing groups. That's a very large percent of matches. It's possible groups made that worse significantly (the data we have seen don't show that.) But it's very clear that stomp rate prior to groups is very high - possibly for reasons intrinsic to MWO others have noted. Possibly because the matchmaker was never that good.

If stomp rate went up post groups did it go up by enough of a percentage for it to be actually noticeable in the number of games played by the average person? If it was a 10% absolute rate increase (twice the temporary worst case increase in the data I can actually find referenced) you'd have to play a lot of matches per night to have an extra stomp (ten matches, and for a stomp against a particular player roughly 20). And it's pretty unlikely you'd notice that difference until you got to a much higher number of drops.
I'm not arguing here that we should or shouldn't have groups or what the best format is. I am saying if we're going to talk about it - at least we should understand the actual data we have been presented. Certainly if there is data I've missed in my search lets look at it and inform our discussions with actual numbers.

Note if we think the average drop has a group of 4 or a group of 2 and 3 then at least 33% of the people in any given match are 'in groups' that's a pretty big chunk of the population.

@VonBruinwald agree I'd like to see more of the data. 1% increase vs 0.7% is pretty much the same either way so your recall is fine. Either way it's a very small absolute increase on a background rate of around 30%

I agree it's hard it would be find all the little data nuggets hidden in various formats - but at least we can use the actual data we can find to inform these discussions. I'd love to see an update.

Data methodology in general for the samples we are quoting isn't great the 24 hour look back and look forward is a very brief period to look at for outcomes that have a lot of variables (weekend matchmaker accuracy is probably better then weekday, non event day is probably better then event, post new mech sale worse then without recent sale)

Can't we just take the quick play options back down to 8v8 and keep faction the 12v12 mode. Have 8v8 solo quick play and 8v8 group quick play queue. It will speed up match making and make it easier to get better match making and make faction the mode to play if you want larger matches make it more unique and different from quick play.

[Anomalocaris]

GARION26, on 21 January 2021 - 09:24 AM, said:

Agreed there were some tweaks to the matchmaker and it's possible stomp rate has gone up significantly with those subsequent changes. But I can't find ANY data from PGI that says that.

I'll try and find it for you, but I might have to exceed my 30 min weekly time allotment for forum shenanigans to do so

Keep in mind too that after all the matchmaker tweaks you would still see an extra stomp every 30 games vs. the pre-merge queue. You can argue about whether or not that's perceptible, but it is a failure for PGI because they said that match quality (which they defined as stomp rate) would improve. It didn't.

But again, I'll try and find the posts about loosening the matchmaker again. Appreciate the discussion.

[Anomalocaris]

Brief twitter search pulled this up. But it lacks context. I'll look more tomorrow.

https://twitter.com/...066324230541313

[GARION26]

Anomalocaris, on 21 January 2021 - 12:05 PM, said:

I'll try and find it for you, but I might have to exceed my 30 min weekly time allotment for forum shenanigans to do so

Keep in mind too that after all the matchmaker tweaks you would still see an extra stomp every 30 games vs. the pre-merge queue. You can argue about whether or not that's perceptible, but it is a failure for PGI because they said that match quality (which they defined as stomp rate) would improve. It didn't.

But again, I'll try and find the posts about loosening the matchmaker again. Appreciate the discussion.

Yeah I remember them loosening the matchmaker (and can find the forum posts here) but that had to do with PSR reset and a very tight PSR matchmaker initially (with a very concentrated around post reset tier 3 population) rather then the groups per se. They didn't present data on stomps in that tweet or that I can find.

Just to clarify the math - see the link to a number needed to harm calculation it is definitely not 30 matches to get one extra stomp from the data we have. It's a 0.7% increase on a background rate of around 30% out of 100 % of matches

If you have a 0.7% absolute increase in stomps - you would need roughly 142 total matches to get one extra stomp. (see number needed to harm calculation)
If its a 1% absolute increase on stomps you would need roughly 100 total matches to get one extra stomp.
From a player perspective only half of those stomps are the unpalatable 'we had no chance matchmaker is broken' compared to the less exciting but still perceived often as fun 'they had no chance, great game played everyone.'


That's on a background rate of having a stomp game out of every 3 or 4 matches you played prior to group incorporation.

I don't think it's plausible that people are actually noticing an extra stomp out of 142 games with a background rate as high as ours was at baseline.

Do you have a link to PGI saying stomp rates would drop after group Queue was merged? That wasn't something I saw when digging through the data.

[Vlad Ward]

Paul discussed the Tier/Weight/Time balancing here:

https://mwomercs.com...al-discoveries/

[VonBruinwald]

Vlad Ward, on 21 January 2021 - 03:17 PM, said:

Paul discussed the Tier/Weight/Time balancing here:

https://mwomercs.com...al-discoveries/

Already covered:

That link is from May.
There was more detailed information released in June,
And the MM Was further opened in July which made things worse and invalidated those numbers.

"Current" setting were due for review in October but that never happened...

[Anjin]

GARION26, on 20 January 2021 - 11:10 AM, said:

You realize the current matchmaker puts a group in both teams as the first matchmaking priority right?

Why are there so many matches then where there's 2 groups on one side and none on the other?

[Hauptmann Keg Steiner]

Anjin, on 21 January 2021 - 11:21 PM, said:

Why are there so many matches then where there's 2 groups on one side and none on the other?

[Citation Needed]

Are you sure there's no group on the other side? You don't have to be in the same unit, or even a unit, to group up.

[Anjin]

Hauptmann Keg Steiner, on 21 January 2021 - 11:46 PM, said:

[Citation Needed]

Are you sure there's no group on the other side? You don't have to be in the same unit, or even a unit, to group up.

Going from the experiences where I'm on a team with no groups (asked) and there are multiple obvious groups on the other side. Though admittedly (not asked). Can't speak for the opposite situation at all, where I know there are two groups on my team and no obvious ones on the other team, as I usually don't end those matches annoyed.

But yes, a lot of the experience is only anecdotal, as I don't usually ask about groups at the beginning of a match. But it happens a fair amount.