49 replies to this topic

[Yokaiko]

stjobe, on 22 August 2015 - 02:19 AM, said:

Day-one rookies get seeded into "mid-to-high range of Tier 4", according to the patch notes.

Which is something I've never understood. I believed it was a mistake to do it with Elo (and they did have to lower the score for new players, I think they did it twice or thrice), and I still believe it is a mistake to do it with PSR.

A pilot dropping into his first match should have a PSR of 0. Not 100 or 500 or 1,000. Zero.

They started new accounts dead center when they released Elo, which was way to high, it sounds like they are putting newbies right where they did at the end of Elo.

Just above the no back-up weapons "get locks for me guyz" and people that only run stock mechs and such.

[Corrado]

Yokaiko, on 22 August 2015 - 02:41 AM, said:

They started new accounts dead center when they released Elo, which was way to high, it sounds like they are putting newbies right where they did at the end of Elo.

Just above the no back-up weapons "get locks for me guyz" and people that only run stock mechs and such.

my 4J hunchie runs without backup weapons. always in the death ball taking my share and never ask for locks. if i already took out 4 mechs with 700 damage and i'll end alone against 5 mechs how im supposed to win with say 5 backup small lasers? 1vs5? enlighten me.

[Yokaiko]

Corrado, on 22 August 2015 - 02:53 AM, said:

my 4J hunchie runs without backup weapons. always in the death ball taking my share and never ask for locks. if i already took out 4 mechs with 700 damage and i'll end alone against 5 mechs how im supposed to win with say 5 backup small lasers? 1vs5? enlighten me.

Well then I wouldn't be talking about you.

I'm talking about the LRM Atlas that never closes below 800m. You KNOW they are out there.

Fun one, I got two kills with the head laser on an out of ammo GI today, one of them a solo kill.

Edited by Yokaiko, 22 August 2015 - 03:01 AM.

[Yellonet]

This is why getting a good team is difficult.

[jss78]

If they used an even 20-20-20-20-20, T1 will definitely lose its character as the elite/comp level. Those players constitute much less than 20%.

[Yellonet]

Corrado, on 22 August 2015 - 02:53 AM, said:

my 4J hunchie runs without backup weapons. always in the death ball taking my share and never ask for locks. if i already took out 4 mechs with 700 damage and i'll end alone against 5 mechs how im supposed to win with say 5 backup small lasers? 1vs5? enlighten me.

L2P noob! Git gud!

Spoiler

[stjobe]

Yokaiko, on 22 August 2015 - 02:41 AM, said:

They started new accounts dead center when they released Elo, which was way to high, it sounds like they are putting newbies right where they did at the end of Elo.

Just above the no back-up weapons "get locks for me guyz" and people that only run stock mechs and such.

Yeah. Only problem? Unless you're tier 1, you'll play with those day-one rookies.
Problem for the rookies? They'll play you (and me, and the rest of the three-year vets in our tricked-out 'mechs)

As I said, new pilots should have PSR 0 when they drop for the first time. It would be better for them, for us, for everyone.

[Cat-in-Exile]

Tastian, on 21 August 2015 - 05:46 PM, said:

I was thinking about that today. Any reason not to have the tier boundaries dynamic so that the playerbase is always 20,20,20,20,20?

I also think this would be broad enough to allow leeway for players to feel better about levelling new chassis or pull out their "for fun" mechs without obsessing over whether they'll drop tiers.

jss78, on 22 August 2015 - 03:05 AM, said:

If they used an even 20-20-20-20-20, T1 will definitely lose its character as the elite/comp level. Those players constitute much less than 20%.

But, assuming the MM wouldn't need to reach for T2 as much, or T3 at all, to fill a match, they'll get even better match quality more consistantly.

Edited by Blitzburger, 22 August 2015 - 01:57 PM.

[Sug]

Vlad Ward, on 22 August 2015 - 01:46 AM, said:

Why would it be anything but a normal distribution centered at tier 3 with delineations at each standard deviation?

Because no one at pgi knows what any of that means.

[Deathlike]

Based on how the PSR scoring changes go..

I suspect the distribution is going to be flawed...

Tier 1: 10%-15% - There will be plenty of Tier 1 players, but it's somewhat flawed in the same breath as it has expanded in a "generous way"

Tier 2: 25%-30% - Better than average players will be here, but probably not as dedicated as the Tier 1 players.

Tier 3: 25%-30% - If you're a regular veteran of this game, but you like your bad mechs, you'll be here (or mistakenly in Tier 2, with certain damage inflating weapons). This will include selfish players that care too much about their K-D ratio, but not actually winning (some will find their way towards Tier 2 though).

Tier 4: 10%-15% - All newbies start here, and descend the "tier mountain" initially.

Tier 5: 5%-10% - Derp happens here, until they get a clue...

Done.

Edited by Deathlike, 22 August 2015 - 02:18 PM.

[Fleeb the Mad]

Sky Legacy, on 21 August 2015 - 06:04 PM, said:

That's unfair to reward the best players with the longest loading times so if tier 1 is only 2 or 5% of the population then their "pool" would be so small their loading times would be half hour per match. Unless ofcourse you open all the "valves" and allow them to play against tier 2-3 as supposedly is the case now

I'll start out by mentioning that the objective of the game is not to reach the highest tier, nor should the developers care about what sort of players they cater to except for those who support the game.

That said, I believe the tier setup is much more convoluted behind the scenes than it seems.

I'm willing to bet that there's actually a blob of outliers outside the normal player distribution that is formed from competetive players in units that form large groups, because they can and frequently do roll competition flat on account of their coordination and regular practice. The influence of effective leadership and coordination is much higher than just being able to shoot straight. I expect that group will form tier 1 because there will probably be a large gap in win/loss ratios and other statistics compared to those who play mostly in the solo queue.

Or to be more blunt, many group queue players will have inflated stats compared to solo queue players and I expect there will be a stat gap between that population and the rest. It's fairly logical to me to put the tier 1 boundary there, to encompass 'competitive' players. Unless, that is, the number of players is too small to matter.

The goal posts may be much wider than anyone expects there, since one of the objectives is to isolate new players from the hardest competition. Many forumites don't consider themselves 'top tier' but would still be grossly unfair opposition for new players in trial mechs. Tier 1 logically wouldn't be the top 5% of the players, since people would hardly see them in the solo queue anyway and new players would still be regularly exposed to veteran players in fully custom mechs that would wreck their ****.

I'll be curious where it all shakes out, but thinking from that perspective being 'top tier' may be meaningless for forum epeen waggling. It might include everyone here.

[Will HellFire]

Most characteristics in the study of a population fit this curve. One supposes that skill too.

And no, it has nothing to do with the rifle.

Edited by Will HellFire, 24 August 2015 - 02:44 AM.

[Joseph Mallan]

Sug, on 22 August 2015 - 02:11 PM, said:

Because no one at pgi knows what any of that means.

Yeah... How many of our players know what he said? I had Statistics classes for my job and I almost forgot what that sentence meant!

[Kjudoon]

It should also be hard to go up in psr. Right now if you win you generally rise even if you do bad you stay even if i understand correctly. While if you do good in a loss you can go up in psr. The general direction you go should be enhanced by win loss giving it a neutral change ratio. Currently it is weighted to rise most of the time. This will cause elo hell sooner rather than later and incentivize tanking games.

[Windschreiter]

Will HellFire, on 24 August 2015 - 02:43 AM, said:

Most characteristics in the study of a population fit this curve. One supposes that skill too.

And no, it has nothing to do with the rifle.

in fact it has:
https://en.wikipedia...Friedrich_Gauss


on topic:
New players should be locked in Tier 4 (at best) until they at least complete the Cadet Bonus - or had 100 matches or so. Otherwise there just wouldn't be enough "reliable" (quite a loose term here) data.

There is a reason PGI went back 8 months in tracking our data, that should not be lost on new players.

[Aresye]

Correct me if I'm wrong, but wasn't Elo spread based on a Bell curve?

[Kmieciu]

Gaussian Distribution means that when you hit an Arctic Cheeter with double Gauss, it will distribute the damage across at least 2 components.

[Deathlike]

Aresye Kerensky, on 24 August 2015 - 11:49 PM, said:

Correct me if I'm wrong, but wasn't Elo spread based on a Bell curve?

It was "supposed to be", but the initial graphs were nothing like that (a badly design initial algo IIRC).

[Dino Might]

Why have tiers or buckets?

Everyone has PSR, start slotting in players for matches based on combined PSR for groups and allow PSR to differ from the overall match average (average PSR for both sides) by no more than a certain amount for each group.

For solos, do the same. Ensure that each match there is a mean PSR and nobody more than part of a standard deviation away. Seems simple enough to me, but I know there will be the code-gurus that tell me something like this is impossible. Well, the math is straightforward, and I'm sure we have more than a few pre-packaged sorting algorithms to deal with this sort of problem.

Kmieciu, on 24 August 2015 - 11:56 PM, said:

Gaussian Distribution means that when you hit an Arctic Cheeter with double Gauss, it will distribute the damage across at least 2 components.

MWO quotation of the year! I spit up laughing at this.

[KodiakGW]

I think Yellownet is mostly correct. The majority of players are probably Tier 5 to Tier 3. An offset bell curve toward the lower tiers. Tier 4 being the peak of active players. We know it would be the peak if they counted all accounts, since many have been inactive prior to January 2015.

Would be nice if they at least gave us the distribution of active players. No solid numbers, just percentages.