138 replies to this topic

[Nightbird]

What you want

Khobai, on 26 June 2018 - 09:53 AM, said:

how does allowing new players to not play with friends without getting stomped by vets improve the new player experience?

Doesn't match what you ask for

Khobai, on 26 June 2018 - 09:53 AM, said:

Which is why I think it makes the most sense to open GQ up to to more casual groups.

The reason you're getting stomped is because the MM can't distinguish between casual and elite groups. This will NOT change if you apply ANY change to ANY mode as long as the change is evenly applied to BOTH causal and elite groups.

This idea as stated in the opening post of this thread solves your problem.

[MischiefSC]

Nightbird, on 26 June 2018 - 09:28 AM, said:

On the contrary, I think nothing is more imporant to MWO than creating good match making before MW5 is released. MW5 however well received is a single player experience. It'll sell for only the price of one mech pack and will be discounted 75% after one year. There will be one shot of revenue, and an influx of new players into MWO much like Battletech once people finish the 50-100 hours of gameplay.

Those players will leave MWO quickly if MM is as dumb as it is today. This applies to QP SQ, GQ, FP, and Solaris. PGI should be desperate to increase the new player retention rate. The MW5 player influx is a one time opportunity. Weapon balancing is important to us min-maxers, but a few % either way on stats is nowhere as important to the new player experience as match making.

I don't know. I think the actual ideal result is 'why not have both'.

You have MW5 for coop vs AI but that will get dull in time, however it will serve to help train new players for MWO which can be seen as the PvP MP for MW5 (hopefully leading to MWO2 in the new engine).

MW5 isn't just single player, it's also coop. That's a critical piece. Coop vs AI would have been great for MWO NPE.

Nightbird, on 26 June 2018 - 03:14 PM, said:

What you want



Doesn't match what you ask for



The reason you're getting stomped is because the MM can't distinguish between casual and elite groups. This will NOT change if you apply ANY change to ANY mode as long as the change is evenly applied to BOTH causal and elite groups.

This idea as stated in the opening post of this thread solves your problem.

The concept being that if there's elite players in good mechs the MM will put them in a match that's even just 6v6 or 8v8 and then split the scrublets into their own 6v6/8v8 instead of a 12v12 with wildly mixed skill ranges and loadouts.

Ironically the bad players, the people who are so bad they're below MWOs already rather embarrassing mean, are still going to lose all the time. There's no mathematical fix for those guys save improvement.

Edited by MischiefSC, 28 June 2018 - 11:31 AM.

[Nightbird]

MischiefSC, on 28 June 2018 - 11:33 AM, said:

I don't know. I think the actual ideal result is 'why not have both'.

You have MW5 for coop vs AI but that will get dull in time, however it will serve to help train new players for MWO which can be seen as the PvP MP for MW5 (hopefully leading to MWO2 in the new engine).

MW5 isn't just single player, it's also coop. That's a critical piece. Coop vs AI would have been great for MWO NPE.

I'm all for both, but MW5 mechanics is going to be very different and AI will be T5 level, so the degree of training wouldn't get a new player over T4 skill. However, if they give MWO a try, it'll be stomp after stomp.The ideal timing is to have MM fixed before then, so that instead of losing 80% of matches they get held at 50%/50%, and that should keep them from saying F-it and quit.

Quote

The concept being that if there's elite players in good mechs the MM will put them in a match that's even just 6v6 or 8v8 and then split the scrublets into their own 6v6/8v8 instead of a 12v12 with wildly mixed skill ranges and loadouts.

Ironically the bad players, the people who are so bad they're below MWOs already rather embarrassing mean, are still going to lose all the time. There's no mathematical fix for those guys save improvement.

Allowing assymetric teams would make match making much faster esp for group queue. Why wait for some arbitrary X vs X match when the pieces available can have a good and fair fight?

As for fixing bad players, the teams will have equal chances of winning so bad players will win more even if they do nothing. This is necessary to improve MWO's overall player experience. It will be much harder to push above 1 WLR because the MM will always be creating opponents on your level.

[MischiefSC]

Nightbird, on 28 June 2018 - 11:57 AM, said:

I'm all for both, but MW5 mechanics is going to be very different and AI will be T5 level, so the degree of training wouldn't get a new player over T4 skill. However, if they give MWO a try, it'll be stomp after stomp.The ideal timing is to have MM fixed before then, so that instead of losing 80% of matches they get held at 50%/50%, and that should keep them from saying F-it and quit.



Allowing assymetric teams would make match making much faster esp for group queue. Why wait for some arbitrary X vs X match when the pieces available can have a good and fair fight?

As for fixing bad players, the teams will have equal chances of winning so bad players will win more even if they do nothing. This is necessary to improve MWO's overall player experience. It will be much harder to push above 1 WLR because the MM will always be creating opponents on your level.

I don't think you'll ever get good balanced asymetric matches. It never worked in TT either; you just only had a couple of players so if player 1 is competent but has 12 inferior mechs and player 2 is dumb but has 8 superior mechs player 1 will still win. This gave the illusion of 'balance' because sometimes the guy with the superior mechs would lose. However that was because each mech wasn't a different player and there wasn't any skill based matchmaking involved.

In MWO focus fire and teamwork are a huge force multiplier. However unless you're dealing strictly with premades who have an established 'group performance Elo' you'll never accurately predict how much teamwork a specific set of players will use together regardless of individual skill.

Having asymetric teams eliminates the value of the more accurate MM by introducing a lot of inaccuracy back into the equation. It would have to be built on the idea of everyone playing full tryhard every match and that's just not realistic for the vast majority of players.

I'm a big fan of a better matchmaker and making it as dynamic as possible but predicting group behavior dynamics is an incredibly twitchy thing. Not everyone competent at a skill or even good in a group is a good leader - someone might be a good leader who's great at building teamwork spontaneously on the fly but is only a mediocre player.

Equal team sizes allows for a predictable set of averages. Individual matches may swing but the average should fall within a guideline set by measurable individual behaviors. However those averages vary by player based on more or less players shifting the value of teamwork and coordination significantly.

I'm game with trying it but dynamic asymetric team sizes probably isn't going to work accurately or consistently - which defeats the purpose of having a more accurate MM.

[Nightbird]

MischiefSC, on 28 June 2018 - 01:51 PM, said:

I'm game with trying it but dynamic asymetric team sizes probably isn't going to work accurately or consistently - which defeats the purpose of having a more accurate MM.

It's hard to put the math into words so I can try an intuitive appeal. Suppose we have a team of 8 EmP in queue. Is there any other group of 8 that the MM can put together in queue that will result in a fair match? The answer is going to be 99.99% no unless there is an 8 man #2 or #3 team in queue at the same time. What about a group of 10? Maybe, if it's two groups of 5 man that are very good. Group of 12? Probably any hodge podge of T1 players would be enough. Group of 16? Even T2-3 players would be enough to give EmP trouble.

I hope it's clear what I am getting at, flexibility will result in consistently better or equal matches to what we have today, and not worse unless the programmer really Fs up.

Edited by Nightbird, 28 June 2018 - 03:19 PM.

[MischiefSC]

Nightbird, on 28 June 2018 - 03:19 PM, said:

It's hard to put the math into words so I can try an intuitive appeal. Suppose we have a team of 8 EmP in queue. Is there any other group of 8 that the MM can put together in queue that will result in a fair match? The answer is going to be 99.99% no unless there is an 8 man #2 or #3 team in queue at the same time. What about a group of 10? Maybe, if it's two groups of 5 man that are very good. Group of 12? Probably any hodge podge of T1 players would be enough. Group of 16? Even T2-3 players would be enough to give EmP trouble.

I hope it's clear what I am getting at, flexibility will result in consistently better or equal matches to what we have today, and not worse unless the programmer really Fs up.

I truly get what you're talking about however you're talking about a fringe case for literally 0.01% of the games population.

It would also mean that you're functionally *requiring* EmP to always be 100% full on comp play all the time in any group size above 4 because the MM will stack them against uneven team sizes.

Probably the best way to handle it is an 'opt-in' where you can select to play with a cbill/xp bonus by being put on a team that's smaller than the other side.

However it's a very cludgy and complex thing to create from a coding perspective that's only going to be applicable to a tiny sliver of the playerbase.

I'm not against the idea; just that you're talking about a lot of effort that's only going to really impact a small group of players.

[Nightbird]

On the population, population =/= representation. The top 1% of players probably represent 10-20% of the drops as people in the bottom hardly drop at all every month. You don't have to be EmP to club most teams you meet either, most div C MRBC teams and above club casual groups easily.

The benefit is fairly straight forward, 100% of players benefit. Even if you're a median skill team winning 50% of matchs, some are going to be clubbing bottoms, some getting clubbed by top teams. Those extreme match ups disappear. If you're a bottom team, you're not going to be clubbed most matches, with the rare fair match against another bottom team.

Also, mech choice is being considered, so if a top team brings a troll deck, the MM will recognize it and adjust. This means you're not stuck always bringing meta.

[MischiefSC]

Nightbird, on 29 June 2018 - 06:32 AM, said:

On the population, population =/= representation. The top 1% of players probably represent 10-20% of the drops as people in the bottom hardly drop at all every month. You don't have to be EmP to club most teams you meet either, most div C MRBC teams and above club casual groups easily.

The benefit is fairly straight forward, 100% of players benefit. Even if you're a median skill team winning 50% of matchs, some are going to be clubbing bottoms, some getting clubbed by top teams. Those extreme match ups disappear. If you're a bottom team, you're not going to be clubbed most matches, with the rare fair match against another bottom team.

Also, mech choice is being considered, so if a top team brings a troll deck, the MM will recognize it and adjust. This means you're not stuck always bringing meta.

I do agree with the potential positives, I don't agree with its total player impact. It doesn't take much of a peek at the Jarls List to see that no, the top 1% is not even 10%, not even 5% of total drops. The majority of drops seem to happen with players who struggle to keep even a 1.0 w/l. It's possible that the top 1% don't even equate to 1% of the total drops in the game. Sorting for total matches played only 9 of the top 300 players for matches played had a match score average that qualified as top 1%. Not what I'd call a proper scientific investigation but while top performers play a lot (have to, it's what makes you a top performer) they're at best representational.

You're falling into confirmation bias. I'm not against the concept - if it can be done in a cost-effective manner and be accurate enough to be a net-positive I'm all for it. Just that it seems like a really intensive concept that's not going to impact a lot of matches.

Edited by MischiefSC, 29 June 2018 - 01:15 PM.

[Nightbird]

MischiefSC, on 29 June 2018 - 01:14 PM, said:

I do agree with the potential positives, I don't agree with its total player impact. It doesn't take much of a peek at the Jarls List to see that no, the top 1% is not even 10%, not even 5% of total drops. The majority of drops seem to happen with players who struggle to keep even a 1.0 w/l. It's possible that the top 1% don't even equate to 1% of the total drops in the game. Sorting for total matches played only 9 of the top 300 players for matches played had a match score average that qualified as top 1%. Not what I'd call a proper scientific investigation but while top performers play a lot (have to, it's what makes you a top performer) they're at best representational.

You're falling into confirmation bias. I'm not against the concept - if it can be done in a cost-effective manner and be accurate enough to be a net-positive I'm all for it. Just that it seems like a really intensive concept that's not going to impact a lot of matches.

Rough math using Jarl's, the first page has a median played of about 4000 games? On page 1000 (around the 50 percentile), median is 500 games. Near the bottom on page 2000, only 20 games played. The first 15 pages (1%, 450 players) has around the same number of games as the bottom 500 pages (25%, 15,000 players).

My confirmation bias is 0?

What about my point that even if a team is winning 50% and losing 50%, it's not fun if half the wins and half the losses are completely stomps. Therefore the benefit is not just to the top teams, but teams in the middle and bottom.

Edited by Nightbird, 29 June 2018 - 01:35 PM.

[MischiefSC]

Nightbird, on 29 June 2018 - 01:32 PM, said:

Rough math using Jarl's, the first page has a median played of about 4000 games? On page 1000 (around the 50 percentile), median is 500 games. Near the bottom on page 2000, only 20 games played. The first 15 pages (1%, 450 players) has around the same number of games as the bottom 500 pages (25%, 15,000 players).

My confirmation bias is 0?

What about my point that even if a team is winning 50% and losing 50%, it's not fun if half the wins and half the losses are completely stomps. Therefore the benefit is not just to the top teams, but teams in the middle and bottom.

I'll reiterate -

I agree 110% (I hate that term) with it being a benefit. I agree that it's worth doing if it's possible and viable in the same scale the rest is viable and possible.

However the rest of your suggestion that I'm worried is getting lost here,

Nightbird, on 19 June 2018 - 11:11 AM, said:

The order as I see it being implemented for solo queue:

1. Population high, make 12vs12 teams
2. Wait until 72 players in queue
3. Launch 3 games, top 24 players based on pilot skill in one game, mid 24 in another etc..
4. Players select map and mode
5. During loading time, server adjusts pilot value based on map and mech(including loadout, skill tree, range, heat efficiency etc), arranges two even value teams
6. Do a tonnage check, swap players if necessary

If population is low, MM can target 10vs10 or 8vs8 reducing the wait to 60 or 48 players.

For QP GQ, assymetric team modifers, group size modifiers comes into play, and tonnage checks removed.

everything but dynamic team size seems absolutely something that can be achieved with a minimal investment which is something PGI could actually do.

Everything BUT asymetric team size would be a huge gain for everyone and an immediate improvement in the experience across all skill levels.

Asymetric team size however is a whole other process and while it has value it smacks of something PGI will look at and say 'yeah, no, that looks hard.'

Split the two concepts. Sell the first. ****, write up a more discrete breakdown for it and I'll Twitter it to Russ and anyone else I can get to listen.

Have the second as a 'now that we've got that in place let's kick it up a notch'.

[Nightbird]

The thing is, there isn't much work for asymmetric teams and much to gain. For one, in order for maps to support 8vs8, 9vs9, 10vs10 etc, some changes have to be made (dynamic HP for structures, timers etc). For another, even with current PSR, Group Queue doesn't pay attention to it because putting the odd sized teams to form 12vs12 in a reasonable amount of time is damned hard. Improving the MM to rate the teams better wouldn't help GQ at all unless the fixed team sizes was removed.

That been said, I don't expect anything out of it. Community support works as well as voting on how to perform surgery.

[Nightbird]

Nightbird, on 29 June 2018 - 02:36 PM, said:

That been said, I don't expect anything out of it. Community support works as well as voting on how to perform surgery.

I also don't expect PGI to understand this proposal. Data mining is a niche profession, with only about 50k members in the US. That's a rate of 0.3% (based on 150 mil workers in the US). Using the same rate for a Canadian company, the odds that there is no one at PGI with an understanding of statistical modelling is (.997)^50 = 86%. PGI has 50 people right?

[MischiefSC]

Nightbird, on 30 June 2018 - 11:03 AM, said:

I also don't expect PGI to understand this proposal. Data mining is a niche profession, with only about 50k members in the US. That's a rate of 0.3% (based on 150 mil workers in the US). Using the same rate for a Canadian company, the odds that there is no one at PGI with an understanding of statistical modelling is (.997)^50 = 86%. PGI has 50 people right?

I don't get why. I haven't had to do my own math in years - I've got a ton of tools and ready formulas I can plug into Excel to do that hard **** for me. At this point I just go boot up a bunch of PCs of people who are not at work and kick off about 30 reports, compile it and turn it in to something that can be made into sexy graphs and pictures.

Getting this data would be more about setting up to capture it than parsing it and implementing it.

Maybe that's why we're in a niche profession - it's not hard to say "this is what's happening, this is the breakdown on outcomes, here's what you can do to make the outcome that would be most beneficial for everyone (that matters to who you're speaking to)".

It's that the response is generally

"Yeah, great idea but that would take work. I think we will just do what we are already doing, maybe with a tiny tweak, and just hope everyone likes it anyway."

Edited by MischiefSC, 30 June 2018 - 12:33 PM.

[Nightbird]

https://mwomercs.com...d-more-tonnage/

Tonnage in groups can be unrestricted with this proposal because mech value is added to pilot value. Your whole team can bring 100 tonners and be matched against a group of all 20 tonners if the lighter group has sufficiently more skilled pilots. Sound fun?

The tonnage limit today is for handicapping skilled teams, but applied to all teams. Sounds as great as it works

[MischiefSC]

Nightbird, on 02 July 2018 - 06:52 AM, said:

https://mwomercs.com...d-more-tonnage/

Tonnage in groups can be unrestricted with this proposal because mech value is added to pilot value. Your whole team can bring 100 tonners and be matched against a group of all 20 tonners if the lighter group has sufficiently more skilled pilots. Sound fun?

The tonnage limit today is for handicapping skilled teams, but applied to all teams. Sounds as great as it works

If the games balance can push the skill tree wider so you have a more consistent performance from people anywhere on the spectrum aside from the absolute extremes this would work better.

[Nightbird]

People had wished the skill tree was deeper, what do you mean by a wider skill tree?

[Nightbird]

https://mwomercs.com...ision-at-least/

Solaris needs mechs and loadouts and pilots to be assigned a statistically computed strength value instead of ELO which assumes a game of only skill. This is a false assumption and doesn't work.

[MischiefSC]

Nightbird, on 02 July 2018 - 08:18 PM, said:

People had wished the skill tree was deeper, what do you mean by a wider skill tree?

Typo, I meant skill curve and player distribution along it.

Solaris needs divisions based on tonnage; a light, medium, heavy and assault division and a single 'open' division.

There isn't anything like the population depth for Solaris needed to support any sort of Elo system. The current divisions just don't work to create any real diversity and just highlights, dramatically, how badly balanced the game is. The difference between a good mech and a bad mech is huge. Insurmountable for all but a tiny handful of players. I think that's why Solaris really fell apart, it's a showcase for how absolutely terrible game balance really is.

[Nightbird]

https://mwomercs.com...ng-is-horrible/

More complaints about MM etc etc