61 replies to this topic

[Bowelhacker]

Vellron2005, on 05 May 2020 - 01:36 AM, said:

I think the ONLY realistic way to measure a player's skill is using match score, BUT only in a specific mech variant, and only calculated from that player's performance in that specific mech withinthe last 100 matches.

Because someone can be the boss of saaay, a Timberwolf, but completely suck in a Commando. Or even a different Timberwolf.

KDR and W/L are not good measurements of skill because they can be manipulated and the said player can get carried. Match score cannot be artificially inflated so easily, but is should still be done on a mech-by-mech basis.

Match score is ********. Someone can get 8 kills/500+ damage and still only make 400 MS (see: Kat Baran Ace of Spades.)

[Jack Booted Thug]

dario03, on 05 May 2020 - 01:28 AM, said:

And they all laughed when I said nobody would suspect the newb green mech. Laughed I say, laughed!

stock green for the win!!!

[Dee Eight]

Is anyone going to suggest something that hasn't been said before dozens of times in dozens of threads on this subject ? No ? Then why bother wasting time typing out long theories on how to fix something PGI refuses to publically admit is broken and always has been broken and was originally coded to do exactly what it has done. Its not a skill bar... its a game tenure bar.

[Horseman]

jjm1, on 04 May 2020 - 11:54 PM, said:

I just look at the colour of their mechs.

There are two tiers: Baby puke green and Probable Threat

And the Stealth Atlas says:

Quote

Hold my beer, let's paint a "friendly assault" con on my forehead.

Edited by Horseman, 08 May 2020 - 11:41 AM.

[Horseman]

Nearly Dead, on 05 May 2020 - 11:15 AM, said:

I would prefer the average of (damage / mech weight) per match from the last 100 matches. If they want to refine it further add in AMS damage, TAG damage and UAV damage so that people don't stop doing those useful but expensive (UAV in c-bills, TAG, NARC and AMS in weight) things. I assume they would need to create a new measurement at the end of the match from damage and mech weight to track.

I mean, if you add these things you more or less end up with match score.

Quote

If an Atlas pilot does 1000 damage every match he/she is probably a T1 player and gets bragging rights. If a Locust player does 200 damage every match he/she is probably completely unnoticed and doesn't even realize himself how good he is doing.

The problem with this approach is that not all mechs of the same tonnage - or even variants of the same chassis - are created equal. Going by your concept of using weight, you'd put Locust, Flea and Piranha on the same scale.
What I would call a "good" solution to that is the difference between the player's average score in last X matches in that chassis and the global average for all players playing that chassis.
This would promote players who do better than the average while demoting players who do worse than the average and properly account for the chassis they were using.

[Mech Walesa]

For me it's weapon accuracy and the use of LoS + positioning. This is true for every game not just mwo.

[crazytimes]

I feel it's probably the aggressive use of decaling that really stands out. A good set of stickers really puts a bit of fear into the enemy. The formula is practically "Real tier=10-(decals)"

[Dr Cara Carcass]

thievingmagpi, on 06 May 2020 - 08:33 PM, said:

pretty much this.

if you're good, you and *other good* players know it. that's it. trying to split hairs over stats and niggling over rankings means you're not good. also, as we've seen, just because people have heard of you and *claim* you are good is meaningless as they're often wrong. the amount of "oh you should listen to (insert 30 percentile here) they know what they're doing" is quite common.

Pretty much this, however, the ocr tool for mechwarrior that is coupled to the jarls list gives you an idea of the outcome of the match before you have played it. So some kind of truth is in this. Sort teams by Win/loss and you have better games. If you want to have a genral idea if some is top or bottom trash the jarls lsit is quite ok.

[Nearly Dead]

Horseman, on 08 May 2020 - 12:06 PM, said:

I mean, if you add these things you more or less end up with match score.
The problem with this approach is that not all mechs of the same tonnage - or even variants of the same chassis - are created equal. Going by your concept of using weight, you'd put Locust, Flea and Piranha on the same scale.
What I would call a "good" solution to that is the difference between the player's average score in last X matches in that chassis and the global average for all players playing that chassis.
This would promote players who do better than the average while demoting players who do worse than the average and properly account for the chassis they were using.

I like this, but I think that it is even less likely than my original proposal to be implemented due to the complexity of gathering the data. You would have to look at each mech variant and rank each player in each of them. I played my Nova Cat, Wolverine, Arctic Wolf, and Jagermech today. So I would have to be ranked in 4 separate datasets to compare to players in those chassis. Then consider the versatility of each of them, the Wolverine can be a lot of different beasts depending on how you kit it out.

IMHO we need to propose something to PGI that would be a significant improvement over the current system, and be cheap and easy enough that they could be talked into doing it. It would also have to be PC, ie, not make people feel bad. Which is why I proposed that they not disclose it and just turn tier into a seniority badge.

[Void Angel]

There's only one way to measure player skill.

Victory.

The reason that PSR isn't cutting it is that it doesn't measure victory over time - there's an upward bias in the system. Since the object of the game is to win, a player's skill can only be measured by how well they achieve the goal.

Victory; nothing else matters.

[Horseman]

Nearly Dead, on 10 May 2020 - 04:47 PM, said:

I like this, but I think that it is even less likely than my original proposal to be implemented due to the complexity of gathering the data. You would have to look at each mech variant and rank each player in each of them. I played my Nova Cat, Wolverine, Arctic Wolf, and Jagermech today. So I would have to be ranked in 4 separate datasets to compare to players in those chassis. Then consider the versatility of each of them, the Wolverine can be a lot of different beasts depending on how you kit it out.

IMHO we need to propose something to PGI that would be a significant improvement over the current system, and be cheap and easy enough that they could be talked into doing it. It would also have to be PC, ie, not make people feel bad. Which is why I proposed that they not disclose it and just turn tier into a seniority badge.

Well, there is a solution to that.
PGI already aggregates your stats per chassis. Keeping track of them on a per-month (rather than lifetime) basis isn't that hard, even if you drop statistics older than past full three 3 months.
Keeping track of the global stats of the chassis after getting per-player stats is again quite easy.

Edited by Horseman, 10 May 2020 - 10:37 PM.

[Nearly Dead]

Just had a perfect example of how badly messed up team score is. Domination on River City.

Came down to one of our guys vs two of theirs.

Our guy killed them both in a close fight in the water at the edge of the pee stain for the win.

His final score was 4 killed, 1 assist, 222 damage. He won the game against 2x1 odds in a Commando. A very effective, efficient fighter. Attacked 5, killed 4.

My final score was 1 killed, 6 assist, 755 damage. I scared several enemy players, got lucky once and was killed by the two final enemy players, dying like a fighting bull in an arena unable to hit either of the enemy fast movers with my AC2s even once I put my back to a wall and focused on just staying alive.

Final team scores?

Live Commando: 207

Dead Jagermech: 407

If we measure win loss, I (and 10 other dead players) get the same credit as he does. Wrong.

If we measure raw damage I get 3.4 times the credit he does. Wrong, wrong, wrong, wr.

If we measure damage / mech tons he gets a score of 8.88 and I get a score of 6.26. I think that is fair. We both get credit for what we actually did, and it is in proportion to the team tonnage that we took up.

Just take this simple score, Damage / Mech tons and record it. When you get to 100 games, drop the oldest so you have a rolling 100 game average.

Divide the player base into groups. Earlier I suggested 24 skill groups so you can do 2, 4, 6 or 8 combined skill groups as the matchmaker requires to achieve the Holy Wait Time of Ultimate Importance.

I have seen too many players do zero damage game after game while spectating to believe that being 0.083 of the winning team should give some kind of credit. Especially now with premade groups dropping and stomping teams of solo players.

If you have a premade of three aces and a C player, the C player gets credit for the win even if he is somebodies nephew and doesn't contribute anything but funny comments and absorbing some damage early in the game.

IMHO it should either be D/MT or K/D. And K/D encourages cowardice and non participation in violation of the COC as players run away and hide or hold back to try to survive the match even at the cost of a team loss.

[VonBruinwald]

Nearly Dead, on 11 May 2020 - 05:26 AM, said:

Just had a perfect example of how badly messed up team score is. Domination on River City.

Came down to one of our guys vs two of theirs.

Our guy killed them both in a close fight in the water at the edge of the pee stain for the win.

His final score was 4 killed, 1 assist, 222 damage. He won the game against 2x1 odds in a Commando. A very effective, efficient fighter. Attacked 5, killed 4.

My final score was 1 killed, 6 assist, 755 damage. I scared several enemy players, got lucky once and was killed by the two final enemy players, dying like a fighting bull in an arena unable to hit either of the enemy fast movers with my AC2s even once I put my back to a wall and focused on just staying alive.

Final team scores?

Live Commando: 207

Dead Jagermech: 407

If we measure win loss, I (and 10 other dead players) get the same credit as he does. Wrong.

If we measure raw damage I get 3.4 times the credit he does. Wrong, wrong, wrong, wr.

Matchscore is badly skewed in favour of damage, a player could headshot 8 opponents and easily score less than someone sandblasting all over the place.

It could also be he scored team damage, this can have a massive effect on your final MS and isn't displayed on the team scoreboard. Or maybe he was hiding for most of the match and only coming out when he saw weakened opponents, 50 dmg per kill isn't a lot (that said he might just have outstanding accuracy).


The best stat you could arguably use is W:L, since this measures your contribution to the team as a whole. It accounts for all playstyles, killing mechs, protecting others with ECM/AMS, using comms and anything else you can imagine that adds/detracts to the teams effectiveness. Of course there is a certain random element as to who you're matched with; which people are quick to jump on if their W:L is poor (remember, if they're doing bad it's not their fault (common factor be damned)).

The problem with using none-team stats is someone can hide at the back under ECM, leach of locks and use their team as meatshields while they rack up damage and pad their stats with kills their fallen teammates softened up. On paper they look like a good player because MS/K:D but in reality they're not.

[Horseman]

Nearly Dead, on 11 May 2020 - 05:26 AM, said:

If we measure damage / mech tons he gets a score of 8.88 and I get a score of 6.26. I think that is fair. We both get credit for what we actually did, and it is in proportion to the team tonnage that we took up.

Just take this simple score, Damage / Mech tons and record it. When you get to 100 games, drop the oldest so you have a rolling 100 game average.

Is a Locust equal to a Piranha?
Nope.
Are they the same tonnage?
Yes.
Normalizing by tonnage doesn't work. Normalizing by average chassis performance would.

[Thorqemada]

I took a look at my stats and there are quite some Details like Game Mode Stats, Map Stats, Mech Stats, Weapons Stats.
They give quite some quantitative Information that you can try to harvest for measurement if you are a stats nerd.

I only generalize some (Solaris not included):
6 of 7 Gamemodes i am at a positive W/L (Dominaton worst)
18 of 20 Maps i am at a positive W/L (Therra Therma and Rubelite Oasis worst)
39 of 43 used Mechs i am at a positive W/L
But what does that mean?

Beyond that you can go deeper, like Weapon accuracy, XP earned, C-Bils farmed, Damage done etc.
But in the end as it is a 12 vs 12 what do individual stats say, you are only one of 24 Players and 23 other massively influence your success and PGI is said can not even differentiate between Solo and Group stats which further alters the stats.
So what would that mean?

I have said in other Threads MWO is an Equipment based game and thus the MM needs to be Equipment based wth some Mechanic that gives Greenhorns a step by step increase in challenge (PSR could do that), and a Mechanic that prevents you from being Setup against the same Players over and over so the Results will average out if you are not Utmost Top Tier where you allways win no matter what.

For everthing else you will have to setup Leagues or Challenges that in the end rank you for your success and place you accordingly.

Edited by Thorqemada, 11 May 2020 - 11:44 AM.

[MischiefSC]

There's no debate.

Win/loss. That is literally a measure of your overall impact on winning matches. There's a small bias introduced for group play but the same is true of match score.

The goal of the game is winning. W/l measures that as accurately as possible.

Edited to add -

Damage, even kills does accurately reflect winning a match. Stay in the back with LRMs, killsteal, hide and try to finish off damaged enemies after your team is wiped, there's a swath of ways to inflate all your metrics except one -

W/L, with a large enough sample size, reflects your impact on winning. Only players who aren't actually that good but play with exceptional players and do that for the majority of their matches are outliers in this and that's maybe a dozen players out of thousands. Statistically insignificant. For over 99.9% of players w/l is accurate and for the rest it's still more accurate than anything else.

Edited by MischiefSC, 11 May 2020 - 03:20 PM.

[K O Z A K]

The more smack people talk on the forums, the better they are. I don't even know why this is up for discussion.

[MischiefSC]

Hazeclaw, on 11 May 2020 - 03:17 PM, said:

The more smack people talk on the forums, the better they are. I don't even know why this is up for discussion.

I know that the MM punishes me because my massive, massive epeen from years of forumwarrioring is considered a tonnage advantage. It's like pugging in a 2man of assaults when I'm dropping solo in a medium.

Massive.

[MeanMachinE]

I would go with KMDD:s, if only one metric is needed from the ones in use.

W/L does not reflect individual performance, especially in solo play.
K/D ratio does not really work as well as KMDD's. I have seen a mech get 2 kills with 2 damage overall in the match. Solo kills on the other hand could be considered.
Match score can drop drastically with e.g. excessive team damage, though your weight in the battle can be crucial. You can also get a good damage and not really contribute to the winning (e.g. some missile boats)
Damage done in a match is not a good metric. Especially in the past LRM boats could do a lot of damage without being effective. Pinpoint damage/aiming should be rewarded.

Edited by MeanMachinE, 15 May 2020 - 02:21 AM.

[Kroete]

MeanMachinE, on 15 May 2020 - 01:43 AM, said:

I would go with KMDD:s, if only one metric is needed from the ones in use.

LRM boats get more kmdd then kills.
A fast streaker can inflate the numbers better, with more kills,
A atm veagle is the top of inflating mechs, with much more kills.

All lock weapons, all spreading, the only point they miss is pinpointdamage.
But how will you messure meaningfull pinpointdamage?
By destroyed components?

Edited by Kroete, 15 May 2020 - 03:27 AM.