717 replies to this topic

[yrrot]

To clarify, I mean the current average MS used to come up with 11% accuracy from R-squared is flawed because the data going in is bad. You'd have to remove the win bias and look at winning MS average and losing MS averages if you wanted marginally better data.

Edited by yrrot, 14 June 2020 - 03:46 PM.

[Zerex]

Nightbird, on 14 June 2020 - 03:09 PM, said:

Average MS is 11% accurate from the R-Squared if you're quoting from me, and this number you can calculate directly from Jarl's data.

The *current MS* which feeds into the PSR is actually less than 5% accurate, same with Jay Z's system. I've been trying hard to explain this so you'll find different attempts posted all over this thread.

Ok i just had a look at R-Squared, I have no idea where or how to even start working out my system using that, is there any chance you could work it out?

[Nightbird]

yrrot, on 14 June 2020 - 03:46 PM, said:

To clarify, I mean the current average MS used to come up with 11% accuracy from R-squared is flawed because the data going in is bad. You'd have to remove the win bias and look at winning MS average and losing MS averages if you wanted marginally better data.

You can easily remove win bias the data? You have the wins and losses. Take (avgMS * (W + L) - winMS*W - lossMS*L)/(W+L)

Zerex, on 14 June 2020 - 03:59 PM, said:

Ok i just had a look at R-Squared, I have no idea where or how to even start working out my system using that, is there any chance you could work it out?

I mean I calculated it to be .11 and .37 for avgMS and WLR respectively. If you want to check my math you have to do it on your end?

Edited by Nightbird, 14 June 2020 - 04:03 PM.

[Zerex]

Nightbird, on 14 June 2020 - 04:03 PM, said:

I mean I calculated it to be .11 and .37 for avgMS and WLR respectively. If you want to check my math you have to do it on your end?

The system i mentioned doesn't use avgMS or WLR, it uses rankings per match based on MS

[Nightbird]

Zerex, on 14 June 2020 - 04:13 PM, said:

The system i mentioned doesn't use avgMS or WLR, it uses rankings per match based on MS

It doesn't stabilize people at any value of PSR, which is the same problem as the current MS PSR.

[DevinMace]

Can a new stat not be made using AMS, Win/loss, and maybe even tier? People love comparing to sports which is the best method, we need a new stat that works like WAR in baseball. Mainly because Wins and losses are more valuable as you go up in tier, I am sure most people can agree to that? I mean PGi does not have to disclose the tiers of each person however it can be used from "behind the scenes" I guess.

[Bud Crue]

OneTeamPlayer, on 14 June 2020 - 02:41 PM, said:

Remove groups from the solo queue and put them back in their own 8v8 queue

This isn’t on the table, is it? I would certainly vote for it if it is.

[Gozuri]

https://mwomercs.com...mmunity-psr-mm/

Ok, so I went through Nightbird's system and yes, it would work if the matchmaker can consistently match players with the same WLR. What his statistics show is simply that WLR is a better indicator for a person's perceivable ability to win games (aka skill) compared to matchscore.
I thought that was a no-brainer and didn't realize that a statistical analysis was required to show it. Think about it, the matchscore is just a value based on some arbitrary things that someone or a group of someones think a player should be doing to win matches, while W/L actually shows the results of that players ability. The main downside of W/L is that it takes into consideration the random factors out of a player's control, i.e. being matched with bad teammates, but random factors would minimize itself if you have a large enough sample size, i.e. played more games. The fact that you even occupy a slot on your team means that there is one less place on your team for a bad teammate to occupy, so your presence would affect the outcome of the match.

The problem I'm seeing with Nightbird's system is that the matchmaker may not always reliably draw players with equal WLR all the time. Sometimes, they will be low player count and so one team has to have an advantage over the other. The rating system in most other games account for that.

Gozuri, on 14 June 2020 - 09:15 AM, said:

If you need a system, just get the average ratings of both teams and determine potential gain/losses in ratings based on the difference.
E.g.

Red Avg psr: 1000 1010 990
Blue Avg psr: 1000 990 1010
R Potential chg: +10/-10 +8/-12 +12/-8
B Potential chg: +10/-10 +12/-8 +8/-12


Everyone on the same team will have the same potential gain/loss.
Winning against a higher skilled team is more rewarding and losing to them is less punishing.
Show team avg psr and potential gain/loss in the match lobby so players don't feel cheated.

This is, of course, more accurate in smaller team sizes due to effect of 'random factors', but as mentioned earlier, given enough games, it will minimize itself.

PS, if someone wants to go by a different definition of skill, I'd like to see it.

[Gozuri]

Zerex, on 14 June 2020 - 10:52 AM, said:

More over its open to be abused by players win farming by on game modes like Assault, Conquest and Incursion. Which could lead to increased numbers of base race matches instead of mech fights, which i think you'll see players leave the game at that point due to it not b a Mechwarrior game but a racing game in mechs.

Wait...isn't that winning the game? Wasn't that the objective? Why put those objectives there if it was never meant to be achieved? How about developing new team strat to prevent someone stealing a win?

[OneTeamPlayer]

Bud Crue, on 14 June 2020 - 06:06 PM, said:

This isn’t on the table, is it? I would certainly vote for it if it is.

This whole thread wasn't on the table till enough people complained about losing their tier and/or the plan for implementing the W/L based system PGI originally thought of.

Figured while i'm here I might as well shoot for the moon.

[Nightbird]

Gozuri, on 14 June 2020 - 07:19 PM, said:

The problem I'm seeing with Nightbird's system is that the matchmaker may not always reliably draw players with equal WLR all the time. Sometimes, they will be low player count and so one team has to have an advantage over the other. The rating system in most other games account for that.

If you look at the example of a team being made, you'll see that my WLR MM does not need to draw players with equal WLR. It's all about sorting players in the queue correctly by skill and then putting half the talent onto each team.

[Gozuri]

Nightbird, on 14 June 2020 - 07:58 PM, said:

If you look at the example of a team being made, you'll see that my WLR MM does not need to draw players with equal WLR. It's all about sorting players in the queue correctly by skill and then putting half the talent onto each team.

Sry, I'll clarify. In your scenario, both teams will have roughly the same average WLR. There may be times however when for example, there are 21 top players and 3 plebs online. Assuming all players have the same WLR within their respective category, i.e. top/pleb, the matchmaker cannot sort so that both teams will have the same avg WLR. This isn't usually a problem if you have a large enough population to draw from, but it could happen in smaller populations. A pure WLR will not account for this.

In other words, you could still end up playing against opponents with a higher avg WLR than your team. Assuming you lost, the result of that game would still be a -1 on your WLR, or a +1 on a win, doesn't matter the odds.

I will say this though. It is still better than trying to include MS.

Edited by Gozuri, 14 June 2020 - 08:37 PM.

[Nightbird]

Gozuri, on 14 June 2020 - 08:32 PM, said:

Sry, I'll clarify. In your scenario, both teams will have roughly the same average WLR. There may be times however when for example, there are 21 top players and 3 plebs online. Assuming all players have the same WLR within their respective category, i.e. top/pleb, the matchmaker cannot sort so that both teams will have the same avg WLR. This isn't usually a problem if you have a large enough population to draw from, but it could happen in smaller populations. A pure WLR will not account for this.

In other words, you could still end up playing against opponents with a higher avg WLR than your team. Assuming you lost, the result of that game would still be a -1 on your WLR, or a +1 on a win, doesn't matter the odds.

Well, what has a better chance of creating two teams with the same WLR? Using WLR? Or Ignoring WLR?

[Zerex]

Nightbird, on 14 June 2020 - 04:40 PM, said:

It doesn't stabilize people at any value of PSR, which is the same problem as the current MS PSR.

Elaborate on this further please, 12 out of the 24 players don't lose or gain PSR on my system, as soon as their match results start hitting their predicted outcome they plateau and that play forever stays at that PSR until their performance shows an increase or decrease

[Gozuri]

Nightbird, on 14 June 2020 - 08:36 PM, said:

Well, what has a better chance of creating two teams with the same WLR? Using WLR? Or Ignoring WLR?

Perhaps there has been some misunderstanding on my stand. I am NOT advocating for the use of WLR, only that the results, i.e Win/Lose, of an individual match should affect a player's ratings. This is what is done in most other games.
There are a few problems with a purely WLR based ranking system. WLR will take into consideration every game you have ever played, even those that no longer reflect your current skill level. A rating system, only shows your current perceivable skill level. You can still climb ratings at a normal rate if you improve, but it will be difficult to climb in WLR if you've already played a few thousand games.

Gozuri, on 14 June 2020 - 09:15 AM, said:

If you need a system, just get the average ratings of both teams and determine potential gain/losses in ratings based on the difference.
E.g.

Red Avg psr: 1000 1010 990
Blue Avg psr: 1000 990 1010
R Potential chg: +10/-10 +8/-12 +12/-8
B Potential chg: +10/-10 +12/-8 +8/-12


Everyone on the same team will have the same potential gain/loss.
Winning against a higher skilled team is more rewarding and losing to them is less punishing.
Show team avg psr and potential gain/loss in the match lobby so players don't feel cheated.

Or in your terms, instead of a loss counting as a -1, if you're unfortunate enough to be on the lower avg psr team, it instead counts as a -0.8 or something, depending on how large the difference is.

I'll give you a scenario. Say the matchmaker decides to consecutively put you in a disadvantageous team 10 matches back to back. You lose all 10 matches. In WLR, you'll get 0/10. If you use a rating system, assuming you have the same disadvantage through all the matches resulting in the same -0.8 per loss and +1.2 per win, it will be something like a 2/10 in your terms.
If you're wondering how is this different, its that you won't suddenly get a round that is too easy after a spate of bad luck, or a game that is too hard after a series of lucky wins.

Edited by Gozuri, 14 June 2020 - 09:50 PM.

[Zerex]

Gozuri, on 14 June 2020 - 07:25 PM, said:

Wait...isn't that winning the game? Wasn't that the objective? Why put those objectives there if it was never meant to be achieved? How about developing new team strat to prevent someone stealing a win?

This is what happens in a team based PvP game when 100% f the rewards are based on the win.

How long will the player base hang round if this starts happening even in 10% of matches?

Also it doesn't matter if on paper Nightbirds system is perfect or not they haven't took the human factor into account. They will do anything to game the system.

[Gozuri]

Zerex, on 14 June 2020 - 10:00 PM, said:

I don't know this game but I think I understand your point. So this is my argument.
The main problem that you brought up is with the gamemodes and has little to do with the Matchmaker.
Objective rushing is not as easy as you make it sound to be in MWO. Out-capping in conquest is a valid strat in conquest, provided the enemy doesn't kill your team off 1st. There are no significant objectives to accomplish in domination or skirmish save for 'kill them all', so we can safely exclude those from the discussion. That leaves incursion and assault.

The way incursion bases are places, most of the time, you will meet the opposing team otw to the enemy base. That leaves you with a decision to either quickly destroy their base, return home fight them in your own turf, or meet them on the field then and there. I'd imagine some of the slower guys would prefer to defend the base, where there are turrets to help you, or get chomped up by a few faster moving mechs as he attempts to rush base. The main problem with this is that its more likely to stalemate, which is why power cells need to be more significant, to encourage players to come out and fight over it.

As for Assault, its similar to incursion but without the power cells. I don't have a solution for this save for moving the bases closer to each other so that it is easier to return to defend it.

If we were gonna do a MS system, we might as well just remove all gamemodes and play only skirmish because non of the objectives actually matter. And even if you did do this, a W/L rating system would still be more accurate than an MS one.
Also, if you're argument is that objectives will provide a significant amount of score points, then how is that any different from W/L in the context of playstyles? You still rush objectives to get the scores right?

Edited by Gozuri, 14 June 2020 - 10:43 PM.

[Zerex]

Gozuri, on 14 June 2020 - 10:37 PM, said:

As for Assault, its similar to incursion but without the power cells. I don't have a solution for this save for moving the bases closer to each other so that it is easier to return to defend it.

Funny how you have to then balance game modes because of the NEW PSR system that was meant balance player skill in matches.

Gozuri, on 14 June 2020 - 10:37 PM, said:

There are no significant objectives to accomplish in domination or skirmish save for 'kill them all', so we can safely exclude those from the discussion. That leaves incursion and assault.

I'm sorry? Domination is the only game mode i have ever payed where the matched ended without a single shot being fired and i have had that happen twice.

Edited by Zerex, 14 June 2020 - 10:53 PM.

[Gozuri]

Zerex, on 14 June 2020 - 10:49 PM, said:

Funny how you have to then balance game modes because of the NEW PSR system that was meant balance player skill in matches.

Read the whole post carefully.

Edit:
I'll try to be clearer. My point is that whatever the problem you're proposing, its with the gamemode. The gamemode has its own problems. You can remove it completely, or attempt to fix it.
Playing a matchscore system in Assault is no different from playing in Skirmish, so removing Assault mode would make no difference. In W/L, you can remove those modes and it would be more accurate than MS.
However, I do not personally like the idea of removing game modes, hence these are just some things that could be done to fix them in a distant future. Or just temporarily remove them while working on a later fix. Or leave it for a while to see if its actually as broken as people think it is.
Game modes are supposed to add variety to a game, something that it fails to do in the current system.

Zerex, on 14 June 2020 - 10:49 PM, said:

I'm sorry? Domination is the only game mode i have ever payed where the matched ended without a single shot being fired and i have had that happen twice.

And why do think you that is? Perhaps its because players would rather not risk their mechs to prevent a loss when it would reflect badly on their score? We are currently in a score based system if you haven't realized. If you have a method that involves capping the domination point without actually fighting off the opponent, perhaps you'd have a better point to make.

Edited by Gozuri, 14 June 2020 - 11:20 PM.

[Zerex]

Gozuri, on 14 June 2020 - 11:00 PM, said:

And why do you that is? Perhaps its because players would rather not risk their mechs to prevent a loss when it would reflect badly on their score? We are currently in a score based system if you haven't realized.

As all players on the losing team all scored 0 MS they all lost the maximum PSR under this system just like they would in Nighthawks