349 replies to this topic

[Tarogato]

Mystere, on 01 February 2018 - 10:56 PM, said:

Although I still think W/L is the best measure over the long term, a better metric for measuring "effective" damage would be helpful for those who think otherwise.

YueFei, on 01 February 2018 - 10:52 PM, said:

I think the Jarl's List should put more weight toward W/L. I agree with Mischief that W/L is, at the end of the day, the only thing that really matters.

WLR absolutely should never be used as a performance metric so long as PGI's leaderboard combines solo and group queue.

For instance, in December I had only a 1.52 WLR, because I played basically 100% solo queue. But January I had a 2.72 WLR, because I played group queue instead - I didn't magically become a 179% better pilot over the course of one month. So no, WLR absolutely will not become part of the Jarl's List scoring mechanic until PGI separates their leaderboards properly. Right now match score is the best we have to go by, though it is flawed as well...

Xavori, on 01 February 2018 - 04:44 PM, said:

Players who score higher on the leaderboards want to score higher on the leaderboards. That's really it.

You can play on teams thereby eliminating most of the randomness I talk about. You can only play the meta mechs. You can intentionally pad your damage stats which will end up boosting your other stats (except kills, those are still pretty random since short of having your teammates avoid kill shots, not much you can do there).

But if you play solo queue, or you play whatever mech you feel like playing regardless of whether it's a top tier or fully skilled or not, then your stats will suffer because you will play with more bad players than good, and in a team based game, much of your success (in MWO 87.5% of it) is tied to your team.

There is some truth to your first line - if you want to get high match scores, you run certain types of mechs and builds. For me in particular, it's high-alpha long-range laservom. Like cHLL/cERML. Nothing else performs as well as that for me.

But the notion that solo queue brings your stats down is completely false. Well, depending on what level player you are maybe. But if you're above average player, than joining group queue with other above average players means that you all are competing to kill things - your group queue mates steal your damage and your kills and kill enemies more efficiently, so everybody's scores will be lower. Whereas in solo queue, you generally are by yourself and everybody else is potatoes, so you will typically score higher. If you want to game the leaderboards, avoid group queue. And then tweet Russ to separate solo/group on the leaderboards. Because a combined leaderboard is broken and I'm tired of that shjt.

Bombast, on 01 February 2018 - 04:33 PM, said:

Wait, we're getting an XVM?

Man, people are going to be pissed.

Yeah, I don't think it's a good idea. It could potentially cause a LOT of problems, and I could even see it driving more people out of MWO. Either it finally reveals to people how potato everybody is and they are so disheartened they leave MWO, or they start quitting out of matches before they start because they see the odds are stacked against them, or they rage at potatoes because they can know see who is the most potato and scream their terrible stats into their ears, hence driving the potatoes out of the game. It's all around a bad idea. But there's nothing we can do to stop it... the means is available. It's nothing more than a couple afternoon's coding project for somebody who really wants to be the Oppenheimer of MWO...

Sebaztien Hawke, on 01 February 2018 - 03:52 PM, said:

...If I started checking these stats I would have to care about whether I survived the game or not. Also, you’d never experiment or use a new and unskilled mech. I better leave it alone.

But if the top person only has 84 games, are they not a very experienced player using a separate account? Doesn’t that break the whole leaderboard? I know my stats or performance now would be much improved if I erased my first 500 games or something...

Actually, that top person scored 600 average match score over those 84 games, in 100% light mechs. That is beyond exceptional. Obviously they are a smurf (at the very least... ).

I just got an idea how to nerf smurf scores though... I'll run it by Scuro and see if we can implement it soon. Should be fool-proof.

[Mystere]

Tarogato, on 02 February 2018 - 05:25 AM, said:

WLR absolutely should never be used as a performance metric so long as PGI's leaderboard combines solo and group queue.

You're definitely right about separating solo and group stats.

[Dr Cara Carcass]

Bombast, on 01 February 2018 - 05:28 PM, said:

You're telling me. I'm at 1,392 (97.39%), and I can't even shoot straight.

Yeonne Greene, on 01 February 2018 - 05:40 PM, said:

There actually is an awfully high number of ultra-super-mega potatoes in this game.

FFS, I'm still in the top 3% and I've been doing mediocre.

Same here, I recently took a stock summoner with ersmal srms ppc and lbx on polar highlands and came out top of my team.
First i thought damn no chance in this thing. Then i used it a few more games. Same results. Some evening you can dominate in everything, some evenings you get emp, 228, and what not on in the same enemy team game after game and you actually need a decent build to compete.
Yolo que is about numbers of games played, with the problem, that the vast majority of mwo players seems to play with less than 10 fps, gamepdas or joysticks. On average you will find so manny ultrapotatoes that you overall get a decent matchscore and end up high on the list.
If the amount of potatoes would be less, i think the upper part of the list would also seperate more, since inflaion of dmg and therefore matchscore would be harder.

[Zergling]

No single stat should ever be used as a performance metric; I prefer a combination of Wins/Losses, Kills/Deaths, Kills/Battle and Average Match Score.

If a player scores high in all of those, then they are very likely a genuinely good player.

And one of the changes I'd like to see with Jarl's List is Kills/Battle being added.

Edited by Zergling, 03 February 2018 - 02:31 AM.

[MischiefSC]

@Tarogato -

Sort the list by w/l. Scrape everyone with sub 200 total matches.

Way better representation (though still flawed) than match score. Look at the names involved.

Sure, there's still some fallout and exceptions but far fewer in w/l.

For an example look up everyone in the MWOWC. Sort by match score (effectively damage/match) and by w/l. See what you get.

Mixed group/pug stats makes it an iffy metric to rank people by but that only effects a very small population segment. Basing it off average damage/match skews for a much, much larger population segment.

Edited by MischiefSC, 03 February 2018 - 04:24 PM.

[Ghogiel]

MischiefSC, on 03 February 2018 - 04:22 PM, said:

@Tarogato -

Sort the list by w/l. Scrape everyone with sub 200 total matches.

Way better representation (though still flawed) than match score. Look at the names involved.

Sure, there's still some fallout and exceptions but far fewer in w/l.

For an example look up everyone in the MWOWC. Sort by match score (effectively damage/match) and by w/l. See what you get.

Mixed group/pug stats makes it an iffy metric to rank people by but that only effects a very small population segment. Basing it off average damage/match skews for a much, much larger population segment.

Match score does the same thing as w/l in solo and group.

The jarls list is for just for the lolz and isn't sorting anything very well.

In solo queue I personally can average 450-650 ms depending how much I game it.

In group queue it's not nearly as easy to farm much over the 400 mark because of how the MM works and having to compete with good players.

W/L in solo will be 2-3 if you are decent. Usually most people top out at 10-20 in solo and group, with more people being able to game in group, but with the actual best W/ls are from solo, eith I think 2-3 exceptions one iirc being fission. tier up is at 99w/l in group this month too so there is that as well.

Edited by Ghogiel, 04 February 2018 - 09:59 AM.

[MischiefSC]

Ghogiel, on 04 February 2018 - 09:55 AM, said:

Match score does the same thing as w/l in solo and group.

The jarls list is for just for the lolz and isn't sorting anything very well.

In solo queue I personally can average 450-650 ms depending how much I game it.

In group queue it's not nearly as easy to farm much over the 400 mark because of how the MM works and having to compete with good players.

W/L in solo will be 2-3 if you are decent. Usually most people top out at 10-20 in solo and group, with more people being able to game in group, but with the actual best W/ls are from solo, eith I think 2-3 exceptions one iirc being fission. tier up is at 99w/l in group this month too so there is that as well.

Sort the list. By match score you get a pretty big segment of sub 2.0 w/l in the top 100 - scrub everyone under 200 matches. Most of the actual players in the finals in MWOWC didn't make top 100.

Then sort by w/l. Again, remove under 200 matches. Suddenly the MWOWC last couple of bracket teams are top 100, winners top 50.

No question group queue skews stats - for players who play a significant number of matches in group queue. That's a really tiny segment. Like approximate to FW population.

What makes good match score is potentially related to but not causal to what wins matches.

[Ghogiel]

MischiefSC, on 04 February 2018 - 05:40 PM, said:

Sort the list. By match score you get a pretty big segment of sub 2.0 w/l in the top 100 - scrub everyone under 200 matches. Most of the actual players in the finals in MWOWC didn't make top 100.

Then sort by w/l. Again, remove under 200 matches. Suddenly the MWOWC last couple of bracket teams are top 100, winners top 50.

No question group queue skews stats - for players who play a significant number of matches in group queue. That's a really tiny segment. Like approximate to FW population.

What makes good match score is potentially related to but not causal to what wins matches.

Where as jarls list is just half alts half solo queue heroes. Half of which probably shouldn't be getting a look in on the top 100 let alone top 10 lol

[MischiefSC]

Ghogiel, on 05 February 2018 - 01:13 AM, said:

Where as jarls list is just half alts half solo queue heroes. Half of which probably shouldn't be getting a look in on the top 100 let alone top 10 lol

I don't have an issue with alts as it's functionally the same player. While that will impact exact position numbers it won't really impact the general rankings. So if one player is spots 2, 6 and 10 he's bumping people down 2 spots, but on a list of 15k that's not as big an overall factor as people running LRM + ERLL Scorch almost exclusively with a 1.15 w/l but huge damage because they brag about how 'sharing armor is a myth and you should always stay in the back'. That guy is above you (significantly, almost a full %) in the rankings for Jarls list.

What I don't want to do is say that it's a bad tool though. It's amazing and a huge credit to the community. It's the go-to for discussions on stats and performance and it should be. It's miles ahead of the in-game leaderboards on their own and the best tool available to players who want to take a serious look at their own performance and improvement.

[Ghogiel]

MischiefSC, on 05 February 2018 - 08:54 PM, said:

I don't have an issue with alts as it's functionally the same player. While that will impact exact position numbers it won't really impact the general rankings.

Sigh. No that's not what's happening, people aren't getting bumped down a couple spots lol. There are dozens of spots possibly triple digit differences between accounts depending on where you play them. What's happening is alts in solo what the list is heavily biased towards.

Nothing to do with player skill or anything. I can get first page on this list if I play solo in an alt. I've topped the leaderboards for ms already so all I need to do is 200 matches so it gives those accounts the full score and they'll be on there. I don't even know where my main is on there. That's what I am saying, it's not a few spots at all.

Edited by Ghogiel, 06 February 2018 - 06:09 AM.

[Humpday]

Tarogato, on 02 February 2018 - 05:25 AM, said:

WLR absolutely should never be used as a performance metric so long as PGI's leaderboard combines solo and group queue.

For instance, in December I had only a 1.52 WLR, because I played basically 100% solo queue. But January I had a 2.72 WLR, because I played group queue instead - I didn't magically become a 179% better pilot over the course of one month. So no, WLR absolutely will not become part of the Jarl's List scoring mechanic until PGI separates their leaderboards properly. Right now match score is the best we have to go by, though it is flawed as well...

God, about damn time someone made this clear to knuckle heads.

PREACH!!!

[Mystere]

MischiefSC, on 05 February 2018 - 08:54 PM, said:

What I don't want to do is say that it's a bad tool though. It's amazing and a huge credit to the community. It's the go-to for discussions on stats and performance and it should be. It's miles ahead of the in-game leaderboards on their own and the best tool available to players who want to take a serious look at their own performance and improvement look down on others.

FTFY. <shrugs>

[MischiefSC]

Ghogiel, on 06 February 2018 - 06:09 AM, said:

Sigh. No that's not what's happening, people aren't getting bumped down a couple spots lol. There are dozens of spots possibly triple digit differences between accounts depending on where you play them. What's happening is alts in solo what the list is heavily biased towards.

Nothing to do with player skill or anything. I can get first page on this list if I play solo in an alt. I've topped the leaderboards for ms already so all I need to do is 200 matches so it gives those accounts the full score and they'll be on there. I don't even know where my main is on there. That's what I am saying, it's not a few spots at all.

You mean by playing an alt so you get that sweet stretch in T4 and T3 before you're back in the same sort of thing.

Except the T4 stretch is down to about 13 matches now and T3 is about 60. Less, if you're doing really good.

Then you're back in the same environment with a smaller stable of mechs.

With the new grind with the skill tree it's really not like it used to be and the trip to T2 (which is almost exactly like T1 anymore) is pretty short. By 200 matches you'll be about where you should be, maybe a tiny bit higher -

which isn't a problem. You playing an alt now and being higher ranked than your main is correct and accurate. Your main will reflect how well you played 3, 6, 9, 12, 18 months ago. I'm assuming you're not an absolute inflexible potato and try to improve, so 2 months of your more recent stats are better than 18 months of stats about how you were before now.

Make sense? That's not actually bad data other than multiple slots held by single players.

That's also far more corrupted by match score than w/l. If you sorted by w/l the issues that would cause are less relevant, save that your alt (only reflecting your current skill and not prior skill) would likely rank higher with a large enough sample.

[Ghogiel]

MischiefSC, on 07 February 2018 - 08:17 PM, said:

You mean by playing an alt so you get that sweet stretch in T4 and T3 before you're back in the same sort of thing.

Except the T4 stretch is down to about 13 matches now and T3 is about 60. Less, if you're doing really good.

Then you're back in the same environment with a smaller stable of mechs.

With the new grind with the skill tree it's really not like it used to be and the trip to T2 (which is almost exactly like T1 anymore) is pretty short. By 200 matches you'll be about where you should be, maybe a tiny bit higher -

which isn't a problem. You playing an alt now and being higher ranked than your main is correct and accurate. Your main will reflect how well you played 3, 6, 9, 12, 18 months ago. I'm assuming you're not an absolute inflexible potato and try to improve, so 2 months of your more recent stats are better than 18 months of stats about how you were before now.

Make sense? That's not actually bad data other than multiple slots held by single players.

That's also far more corrupted by match score than w/l. If you sorted by w/l the issues that would cause are less relevant, save that your alt (only reflecting your current skill and not prior skill) would likely rank higher with a large enough sample.

lol you are so ignorant about this game. the tier is actually secondary. All you have to do is play to farm while solo. that's the main deciding criteria.

I think you are trying to say that I wouldn't be able to do it and that my T2 alt wouldn't be on first page after 200 games or something. LOL

Edited by Ghogiel, 08 February 2018 - 07:42 AM.

[MischiefSC]

Ghogiel, on 08 February 2018 - 07:35 AM, said:

lol you are so ignorant about this game. the tier is actually secondary. All you have to do is play to farm while solo. that's the main deciding criteria.

I think you are trying to say that I wouldn't be able to do it and that my T2 alt wouldn't be on first page after 200 games or something. LOL

No. I absolutely do not think you're good enough at the game to get on page 1 with over 200 matches.

You're also missing the point - I'm specifically saying that is WHY W/L is a superior metric to match score (which is effectively damage average). You could not get to front page on W/L via pug matches unless you're one of the best players in the game. You can game your rank way up by damage farming.

Even with that though I don't think you're good enough at this game to go front page even with the match score rankings in an alt with over 200 matches. Higher than you would regularly get, sure. Just not front page.

[Ghogiel]

MischiefSC, on 08 February 2018 - 09:43 AM, said:

No. I absolutely do not think you're good enough at the game to get on page 1 with over 200 matches.

You're also missing the point - I'm specifically saying that is WHY W/L is a superior metric to match score (which is effectively damage average). You could not get to front page on W/L via pug matches unless you're one of the best players in the game. You can game your rank way up by damage farming.

Even with that though I don't think you're good enough at this game to go front page even with the match score rankings in an alt with over 200 matches. Higher than you would regularly get, sure. Just not front page.

I already topped the old mechstats website by a lot. I've already topped the leaderboards for match score. I think I might even have the highest on there. I'll top it again for this month and be rank 1 again later today just to show everyone how deluded about this game you are.

[darqsyde]

Anybody have a list of the dates of each season...Apparently I was pretty damn good in Season 6, in heavies. I wonder if that was when I was working on the Maddog and Timby or maybe Ebon Jag, or was that the season of the Grasshopper?

Season 13-14 was probably when I was working the freebie Banshee and Stalkers.

[Bombast]

darqsyde, on 08 February 2018 - 11:53 AM, said:

Anybody have a list of the dates of each season...Apparently I was pretty damn good in Season 6, in heavies. I wonder if that was when I was working on the Maddog and Timby or maybe Ebon Jag, or was that the season of the Grasshopper?

Season 13-14 was probably when I was working the freebie Banshee and Stalkers.

Each season is a month, so just count back.

[MischiefSC]

Ghogiel, on 08 February 2018 - 09:58 AM, said:

I already topped the old mechstats website by a lot. I've already topped the leaderboards for match score. I think I might even have the highest on there. I'll top it again for this month and be rank 1 again later today just to show everyone how deluded about this game you are.

Okay. You do that. 200+ matches, #1 on the leaderboard.

Then go get a book on reading comprehension because the point I've been making from day 1 is that match score, being damage based, is way less reliable than win/loss because it's easier to game.

While I do believe that it's easier to game your match score up and as such is less reliable I absolutely do not think that you are a good enough player to do so to rank 1.

[Ghogiel]

MischiefSC, on 08 February 2018 - 04:55 PM, said:

Okay. You do that. 200+ matches, #1 on the leaderboard.

Then go get a book on reading comprehension because the point I've been making from day 1 is that match score, being damage based, is way less reliable than win/loss because it's easier to game.

While I do believe that it's easier to game your match score up and as such is less reliable I absolutely do not think that you are a good enough player to do so to rank 1.

Nuh uh bro, no moving the goal posts. First page. You have paypal?