57 replies to this topic

[Kalam Mehkar]

Duke Nedo, on 30 September 2015 - 05:59 AM, said:

The equation for the line is: y = -229.4x + 1037.1, so if you follow that, for a fresh player to reach the lower limit of Tier 1 it would take in the order of ~ (4.2-2.0) x 230 = 506 excess wins.

For someone starting now, this value will be off by a percentage depending on when they started to record PSR-data relative to the stat-reset. I don't remember either of these dates, but if we would just assume about 50% of the time since stat reset was used to seed PSR, then a fresh cadet would need to gather some 250 more wins than losses to reach the boundry between Tier 1 and Tier 2.

For a good solo player at a 1.2 W/L ratio, that corresponds to some 2700 games.
For a new player that learns along the way with a 1.1 W/L ratio it would take some 5250 games.
For a player playing with a good group with a 2.0 W/L ratio it could take less than 750 games.

So we find something that strongly correlates to what Paul had said regarding time to make Tier 1. Paraphrasing Paul from a twitter: "Likely take 2,500 matches to get to Tier 1, a really strong player might make it in 1,000 games."

Interesting.

[Creovex]

Lefteye Falconeer, on 30 September 2015 - 02:45 AM, said:

How can someone with:

- 5150 played matches
- 1.83 KDR
- 1.10 W/L ratio

be stuck in tier 4? Even if the system wasn't biased the way it is, I don't understand this data entry.

Either there are some serious bugs in the system or this data is false/mistyped?

Same for:

- 3764 played matches
- 2.06 KDR
- 1.57 W/L ratio

still swamped in tier 3?

What is going on here? Theories?

Another question:

Do "Archived stats" count, or only "Current", after the reset?

I am Tier 3, and then again I try to master all mechs (Good luck mastering a Commando and expecting Tier 1).
I keep over a 1.13 KDR (a 1.125 W/L) but try to keep that up when leveling lights or mediums... hahaha...

Tiers WOULD matter if included class size weighting...AKA..

• Light: 1.0 KDR = Tier 1
  
• Medium: 1.5 KDR = Tier 1
  
• Heavy: 2 KDR = Tier 1
  
• Assault: 3 KDR = Tier 1
  
• ....and then average them across the classes while leaving a default value of Tier 3 for class where the individual has not played over 100 games... (ie. Only played 5 games in a light, your Light Class Tier value is defaulted to Tier 3)

Try as they might, PGI has successfully implemented another Assault/Heavy|Skirmish metric (a general KDR wasn't enough?) and pawns it off as skill...

Example:
I ran out from cover as a distraction on Tourmaline Desert, bought my team time to finish the caps and I ended up redirecting most of the overwhelming enemy force (they had 9 left, we had 3) .... and I got a good ole decrease in Pilot Skill.... even though the tactic allowed us to win by cap...
So apparently, had I played pure skirmish rules, I would have probably had an increase but tactics take a back seat in PSR...

Edited by Creovex, 30 September 2015 - 08:58 AM.

[jay35]

Two thoughts for PGI:

#1 - They should use a rolling average of the player's last, say, 500 matches, rather than going back so far historically that players with a ton of matches can no longer move efficiently up or down the Tier spectrum. This would help players arrive at their relative skill range more quickly while also adjusting their level based on how they are playing rather than some ancient data that's not really relevant to how they're playing more recently.

#2 - Some people enjoy the journey more than the destination. Providing a longer path to get to the top means a gentler ascent and time to actually bring one's skill up along the way. Shoving people right to the top quickly means they get burned out quicker if all they're fighting is stiff competition every single match.

So if the goal is simply to label people, then sure, get them to Tier 1 as quickly as possible. But if the goal is to create a journey and an experience, then it's better that it takes some time to get there and it feels like more of an accomplishment to arrive there.

[Mister Blastman]

Alright...

Quote

201 players collected = sample size.
The average player is Tier 2 (~63% bar filled).
The average player has played 3,903 matches.
The average player has a 1.59 KDR and 1.29 WLR.

See the problem?

Average win/loss can't be greater than 1.0. But in this data set, it is. Thus, the data set is invalid for anything more than curiosity. It is good data to look at--I found some stuff interesting in it, but the >1.0 win/loss ratio prevents it from being used scientifically.

I see it as more of a, "Well, this is how I size up versus my peers." But we can't infer much about the matchmaker, PSR or anything else until we have a true set that represents a realistic pool. And a realistic pool would have a 1.0 win/loss ratio as the average because... you can't have more winners than losers. For every winner... there must be a loser (unless it is a tie and those are very, very rare).

Thanks for the efforts, regardless. It was a fun read.

Edited by Mister Blastman, 30 September 2015 - 09:06 AM.

[Death Proof]

Creovex, on 30 September 2015 - 08:57 AM, said:

Example:
I ran out from cover as a distraction on Tourmaline Desert, bought my team time to finish the caps and I ended up redirecting most of the overwhelming enemy force (they had 9 left, we had 3) .... and I got a good ole decrease in Pilot Skill.... even though the tactic allowed us to win by cap...
So apparently, had I played pure skirmish rules, I would have probably had an increase but tactics take a back seat in PSR...

I keep hearing people saying that they've gone down (as in a red "down" arrow) in PSR on a WIN, even though that is supposedly not possible according to Paul's infographic.

If someone could oblige, I'd love to see a screenshot of this actually occurring.

http://mwomercs.com/...ecrease-on-win/

[Valdherre Tor]

Riddle me this Batman-

So all tiers have a positive KDR and WLR except T5. How can this be? Shouldn't half the population have lower than 1 KDR and 1 WLR?

[Creovex]

Death Proof, on 30 September 2015 - 09:05 AM, said:

I keep hearing people saying that they've gone down (as in a red "down" arrow) in PSR on a WIN, even though that is supposedly not possible according to Paul's infographic.

If someone could oblige, I'd love to see a screenshot of this actually occurring.

http://mwomercs.com/...ecrease-on-win/

I would look and see if I have a screen tonight. Paul says alot of things....(been here since Beta...the list is too long to talk about.. again)

Edited by Creovex, 30 September 2015 - 09:11 AM.

[Duke Nedo]

Mister Blastman, on 30 September 2015 - 09:05 AM, said:

Alright...



See the problem?

Average win/loss can't be greater than 1.0. But in this data set, it is. Thus, the data set is invalid for anything more than curiosity. It is good data to look at--I found some stuff interesting in it, but the >1.0 win/loss ratio prevents it from being used scientifically.

I see it as more of a, "Well, this is how I size up versus my peers." But we can't infer much about the matchmaker, PSR or anything else until we have a true set that represents a realistic pool. And a realistic pool would have a 1.0 win/loss ratio as the average because... you can't have more winners than losers. For every winner... there must be a loser (unless it is a tie and those are very, very rare).

Thanks for the efforts, regardless. It was a fun read.

Valdherre Tor, on 30 September 2015 - 09:08 AM, said:

Riddle me this Batman-

So all tiers have a positive KDR and WLR except T5. How can this be? Shouldn't half the population have lower than 1 KDR and 1 WLR?

Can think of 2 explanations from the top of my head....

1. Forum users, and in particular people willing to share their stats are a bit above average. This makes conclusions difficult from the dataset as it is not representative.
2. This is not a closed system. People come and people go, if bad players are more likely to quit then what we see can be representative values for long-term active players.

[Mister Blastman]

Duke Nedo, on 30 September 2015 - 09:41 AM, said:

1. Forum users, and in particular people willing to share their stats are a bit above average. This makes conclusions difficult from the dataset as it is not representative.
2. This is not a closed system. People come and people go, if bad players are more likely to quit then what we see can be representative values for long-term active players.

#2 is a good point. Good luck getting those numbers--who stays and who quits. But I think it is an important number... player turnover.

Player turnover points to flaws in the game. What are those flaws? How do we address them? How may they be fixed? It is an important issue we need to explore so the game can survive and improve.

And the developer needs to be willing to implement what is needed to do this.

[THUNDERG0D]

Now that you have captured this data for these various users you need to ask them for the numbers again. If the PSR seeding was accurate and the match maker is working as designed you shouldn't see much drift. However, if the seeding was done poorly you should see massive deviations in player stats from this baseline.

[Mazzyplz]

Thundergod99, on 01 October 2015 - 11:45 AM, said:

Now that you have captured this data for these various users you need to ask them for the numbers again. If the PSR seeding was accurate and the match maker is working as designed you shouldn't see much drift. However, if the seeding was done poorly you should see massive deviations in player stats from this baseline.

my data is part of this; i can tell you it is holding pretty much the same; it seems like my psr bar went down some 20% due to a loss streak, i stopped playing due to real life duties; and now i am quickly returning it to where it was; soon i will have made a little progress; making what paul said pretty much confirmed from what i can see - it seems the surplus of matches won is the magic number - so long as mine is somewhat balanced wins-to-losses it seems my psr bar will more or less move very slowly or remain where it is

[lpmagic]

lol, well, my w/l was at 1.65 ish recently and I just ran off a string of 40-1 (w/l) and the psr bar has hardly moved despite only 1 neutral rating (the loss) yes I have a lot of games under my belt, no question, but you would think 40-1 would be able to move the needle more then a minute fraction. But, as someone else mentioned, I can either play up my mechs and prolly go 25-15 and stay neutral ish or run what I really relish and go 40-1 again, and continue to eek the bar up slowly (i know I know, just sayin )

as many mention, this seems more of a leveling experience then a PSR value..... happy hunting kids!

[Zainadin]

This is a pretty much dead thread now but I wonder how Experience Points to date factors in on the tiers.

[Sandpit]

Mister Blastman, on 30 September 2015 - 09:05 AM, said:

Alright...



See the problem?

Average win/loss can't be greater than 1.0. But in this data set, it is. Thus, the data set is invalid for anything more than curiosity. It is good data to look at--I found some stuff interesting in it, but the >1.0 win/loss ratio prevents it from being used scientifically.

I see it as more of a, "Well, this is how I size up versus my peers." But we can't infer much about the matchmaker, PSR or anything else until we have a true set that represents a realistic pool. And a realistic pool would have a 1.0 win/loss ratio as the average because... you can't have more winners than losers. For every winner... there must be a loser (unless it is a tie and those are very, very rare).

Thanks for the efforts, regardless. It was a fun read.

^

[Wintersdark]

Mister Blastman, on 30 September 2015 - 09:05 AM, said:

Alright...



See the problem?

Average win/loss can't be greater than 1.0. But in this data set, it is. Thus, the data set is invalid for anything more than curiosity. It is good data to look at--I found some stuff interesting in it, but the >1.0 win/loss ratio prevents it from being used scientifically.

I see it as more of a, "Well, this is how I size up versus my peers." But we can't infer much about the matchmaker, PSR or anything else until we have a true set that represents a realistic pool. And a realistic pool would have a 1.0 win/loss ratio as the average because... you can't have more winners than losers. For every winner... there must be a loser (unless it is a tie and those are very, very rare).

Thanks for the efforts, regardless. It was a fun read.

This is my problem with it.

It's not Tarogato's fault, but his data pool is not indicative of MWO players as a whole. People have to share their stats, and generally speaking the people doing this are more... Dedicated, I suppose, players. Of these, the majority are going to be stronger, more experienced players. Certainly with a predominance of unit-oriented players.

So, these players are going to be drawn from those spending more time in the group queue. Large group play tends to lead to a much larger kdr/wlr, given the impact team play has on winning in the group queue.

So, Tarogato's data and charts are very interesting, and very well put together, but they are NOT indicative of MWO players as a whole, and that needs to be kept in mind.

For example "the average player is tier 2" - This is extremely dubious. I wouldn't even call t2 an average here on the forums, excluding the *swarms* of low-tier steamies.

[Dawnstealer]

I know my KDR is below 1 (0.91, I think?), mostly because I'm constantly grinding new mechs and, when I'm not, I'm usually running scout mechs like my Locust or other Lights. Some of those mechs just weren't made to get a lot of kills, so while I won't die often in them, it still scuttles my rating.

Add to that I don't really care about KDR and I'll be first through the door more often than not. That doesn't exactly help my stats, either. Although lately I've been doing less of that and my stats are climbing.

[Davegt27]

I have a .89 KDR it’s been going down since the PSR came out but I find myself almost to tier 2

If you look at my sig I was at the first third of tier 3

I just looked at my stats and I died 1,319 times in my Jager DD try to top that

Man was that thing a death trap

[MrDerp]

Is there a new version of this thread?