155 replies to this topic

[Gagis]

Your side in a team is random, which means wins and losses have a large luck element.

Luck however, is the same for everyone, and cancels out over time. Thats why long-term statistics are a perfectly good way to track skill even in games with a large luck based element.

For similar reasons, a good player in Magic the Gathering might have 60% odds to win instead of 100% odds to win against a worse opponent, thanks to the large luck-based part of the game. Over a fairly small number of matches, the best players (and best decks) still clearly stand out since bad luck and good luck are equal for every player.

Edited by Gagis, 13 April 2022 - 06:08 AM.

[Horseman]

feeWAIVER, on 13 April 2022 - 05:40 AM, said:

Question,
Do you think emphasizing W/L over match score will change player behavior?
Specifically Assault Game Mode being played for cap more than kills?

Simply: no. They play this game to make pew-pew at giant robots, and that is also the activity that provides the biggest rewards during the match. You'd have to massively shift the reward payouts towards win by objective to change that.

[Nightbird]

feeWAIVER, on 13 April 2022 - 05:40 AM, said:

Question,
Do you think emphasizing W/L over match score will change player behavior?
Specifically Assault Game Mode being played for cap more than kills?

I ask because honestly that last assault event created some really fun games when both teams were on the same page about what the objective was.

No, because how people play is more guided by rewards than anything else. A win with no damage gives pretty much nothing in terms of x-bills, xp, MS, to make people shoot each other.

[Nightbird]

DaZur, on 13 April 2022 - 05:50 AM, said:

Our discussion is essentially useless if you truly believe the peripheral impact of the team in a team game is irrelevant...

Of course an individuals quality of play contributes to the probability of a win. That said, the composite quality of a teams play has a much stronger correlation to that same win.

This is by definition correlation =/= causation.

You can also read this real life example in basketball:

https://www.pivotana...post/net-rating

Just by being on the court, every player affects the team's offense and defense performance and this is reflected in his stats. Since players are seldom in a whole game, the stat is adjusted per 100 possessions, but it's the same as per 100 games for MWO. Every point in basketball for the team is a win, and every point for the other team is a loss.

You can argue that the rest of the team affects the gain or loss of points more, this is true, but over many games this washes out and the only impact is the player himself.

Edited by Nightbird, 13 April 2022 - 08:34 AM.

[DaZur]

Nightbird, on 13 April 2022 - 08:32 AM, said:

You can also read this real life example in basketball:

https://www.pivotana...post/net-rating

Just by being on the court, every player affects the team's offense and defense performance and this is reflected in his stats. Since players are seldom in a whole game, the stat is adjusted per 100 possessions, but it's the same as per 100 games for MWO. Every point in basketball for the team is a win, and every point for the other team is a loss.

You can argue that the rest of the team affects the gain or loss of points more, this is true, but over many games this washes out and the only impact is the player himself.

Thanks for showing me this... This I find interesting.

The only niggling thing I would see between that example and MWO is teams in MWO are not a constant. (Each drop is largely random dependent upon the depth of the draw pool). You'd likely have to apply a lengthy and broad ANOVA test... But as you and other demonstrate emphatically that over time the deviations will wash out regardless.

Edited by DaZur, 13 April 2022 - 08:48 AM.

[SFC174]

DaZur, on 13 April 2022 - 08:48 AM, said:

Thanks for showing me this... This I find interesting.

The only niggling thing I would see between that example and MWO is teams in MWO are not a constant. (Each drop is largely random dependent upon the depth of the draw pool). You'd likely have to apply a lengthy and broad ANOVA test... But as you and other demonstrate emphatically that over time the deviations will wash out regardless.

Liked for being willing to consider alternative viewpoints and even change your mind. Would that it was not so rare in this world.

[Nightbird]

DaZur, on 13 April 2022 - 08:48 AM, said:

Thanks for showing me this... This I find interesting.

The only niggling thing I would see between that example and MWO is teams in MWO are not a constant. (Each drop is largely random dependent upon the depth of the draw pool). You'd likely have to apply a lengthy and broad ANOVA test... But as you and other demonstrate emphatically that over time the deviations will wash out regardless.

The same link talks about synergy rating, which is the effect of when two players are on a team together as compared to when they are alone.

The would completely resolve the issue of the difference in performance individually versus in groups in MWO that Horseman talked about, but you can forget about PGI programming it since it would mean tracking another stat for people you frequently group with.

[Magnus Santini]

Alright, let's see if you can explain this to the rest of us. Are you saying that a WLR-based matchmaker works better because it is more closely looking at the player's usual performance, where an AMS-based one is looking instead at that but also at things that are more or less related to the performance but not very strongly (like lance in formation points)?

[Nightbird]

Magnus Santini, on 13 April 2022 - 09:59 AM, said:

Alright, let's see if you can explain this to the rest of us. Are you saying that a WLR-based matchmaker works better because it is more closely looking at the player's usual performance, where an AMS-based one is looking instead at that but also at things that are more or less related to the performance but not very strongly (like lance in formation points)?

Yes, using the basketball analogy, it would be like trying to balance teams using a metric based on how often a player dribbles the ball, scratches his butt, or grunts. Sure, there are other important things in there, but the respective weights of each action is incorrect making the whole AMS metric unreliable.

[Knownswift]

DaZur, on 13 April 2022 - 05:50 AM, said:

Our discussion is essentially useless if you truly believe the peripheral impact of the team in a team game is irrelevant...

Of course an individuals quality of play contributes to the probability of a win. That said, the composite quality of a teams play has a much stronger correlation to that same win.

This is by definition correlation =/= causation.

"It's just my team that is bad"

Even in such a case where a player is continually carried to victory that itself is evident in the publicly available data.

Your use of that adage is completely incorrect. There is a STRONG correlation between in game performance and winning.

[DaZur]

Knownswift, on 13 April 2022 - 02:31 PM, said:

"It's just my team that is bad"

Even in such a case where a player is continually carried to victory that itself is evident in the publicly available data.

Your use of that adage is completely incorrect. There is a STRONG correlation between in game performance and winning.

You keep throwing that "publicly available data" thing around like it's a magic wand or something... lol. This magic data, by itself, nowhere can demonstrate a player being carried by their team or not. You'd have to collect and apply the single users data as well as all the players they've played with over a broad sample (ANOVA test) ... Which I don't think even Nightbird has the intentional fortitude to compile.

Correlation... "yes". STRONG correlation... "no". A single players correlation contribution is 1 to 11 and well.... the teams is 11 to 1. If I have to explain how the teams composite contribution is more palpable in a team game, than a single players... Well, your just being stubborn for effect.

Edited by DaZur, 14 April 2022 - 04:48 AM.

[Nightbird]

DaZur, on 14 April 2022 - 04:46 AM, said:

You keep throwing that "publicly available data" thing around like it's a magic wand or something... lol. This magic data, by itself, nowhere can demonstrate a player being carried by their team or not. You'd have to collect and apply the single users data as well as all the players they've played with over a broad sample (ANOVA test) ... Which I don't think even Nightbird has the intentional fortitude to compile.

Correlation... "yes". STRONG correlation... "no". A single players correlation contribution is 1 to 11 and well.... the teams is 11 to 1. If I have to explain how the teams composite contribution is more palpable in a team game, than a single players... Well, your just being stubborn for effect.

Can't respond to you because you have no idea what an ANOVA test is.

You're also using the word correlation wrong.

You're sounding like a kid trying to use big words... if you know what they mean, do some math and show us...

[An6ryMan69]

Nomad Tech, on 29 March 2022 - 10:14 PM, said:

To make the tiers better overall you need to one remove groups and two stop mixing and matching different tiers together.

This here would provide the best player experience, but at the cost of wait times.

It's not rocket science.

The mixing of groups and skill levels really is a problem, I'm not sure why guys minimize this.

Example-

A few weeks back I introduced my 15 year old to MWO, we got him all set up and I watched/coached him through the training grounds and all of his cadet missions. I clearly remember at one point, while he was still a cadet with maybe a dozen drops to his name, and could barely figure out basic mech mechanics while under fire, he was attacked by what was obviously a group drop...and he was specifically double teamed by TWO reds both with "Ace of Spades" titles, battering him a the same time. So, the lowly cadet got his *** chewed off in about two seconds by guys he should really never run into on the field, got right pissed off, and didn't come back to the game for about three weeks. A fine example of how match making should never work, and how it can be so tough to bring in new players.

Thanks all.

Edited by An6ryMan69, 14 April 2022 - 09:07 AM.

[Nightbird]

An6ryMan69, on 14 April 2022 - 09:06 AM, said:

This here would provide the best player experience, but at the cost of wait times.

It's not rocket science.

The mixing of groups and skill levels really is a problem, I'm not sure why guys minimize this.

Example-

A few weeks back I introduced my 15 year old to MWO, we got him all set up and I watched/coached him through the training grounds and all of his cadet missions. I clearly remember at one point, while he was still a cadet with maybe a dozen drops to his name, and could barely figure out basic mech mechanics while under fire, he was attacked by what was obviously a group drop...and he was specifically double teamed by TWO reds both with "Ace of Spades" titles, battering him a the same time. So, the lowly cadet got his *** chewed off in about two seconds by guys he should really never run into on the field, got right pissed off, and didn't come back to the game for about three weeks. A fine example of how match making should never work, and how it can be so tough to bring in new players.

Thanks all.

This wouldn't help because people do make alt accounts to seal club and if you tank the first few matches, the current PSR Tier mechanism allows them to play a lot of matches before needing to create a new account.

If those people were in a group, they can also shoot each other in the leg to drop their Match Score to 0 and prevent increasing in Tiers.

The current system has too many problems.

[DaZur]

Nightbird, on 14 April 2022 - 08:20 AM, said:

Can't respond to you because you have no idea what an ANOVA test is.

You're also using the word correlation wrong.

You're sounding like a kid trying to use big words... if you know what they mean, do some math and show us...

Sigh... I thought we were getting along.

I made it clear I was not a mathematician nor a statistician far earlier in this thread. I'm 54 years old, haven't had a need to actually apply what I learned in a single statistics semester in high school (37 years ago if anyone's counting)... If I'm using the wrong vernacular while describing my thought process... feel free to call me on it. Don't really care...

Neither my age or ego has room having a piss fight with anyone, least of all you. To that end, I have nothing to prove to you, so I will not be "showing you my math"...

Edited by DaZur, 14 April 2022 - 10:41 AM.

[Nightbird]

DaZur, on 14 April 2022 - 10:40 AM, said:

Sigh... I thought we were getting along.

I made it clear I was not a mathematician nor a statistician far earlier in this thread. I'm 54 years old, haven't had a need to actually apply what I learned in a single statistics semester in high school (37 years ago if anyone's counting)... If I'm using the wrong vernacular while describing my thought process... feel free to call me on it. Don't really care...

Neither my age or ego has room having a piss fight with anyone, least of all you. To that end, I have nothing to prove to you, so I will not be "showing you my math"...

I also made my position clear, I will not leave wrong math statements unchallenged. If you want to see a correlation ANOVA analysis between past match score or past WLR against future WLR, see here: https://mwomercs.com...and-suggestion/

If you understand that you don't understand a topic, feel free to ask a question instead of making an incorrect statement.