The Player Skill Rating (Psr) System Explained... (As Best I Can)
#81
Posted 24 October 2015 - 09:28 PM
#82
Posted 24 October 2015 - 09:32 PM
I didn't screen the match, as I never knew that it would become relevant. As an average general estimation, I think your numbers are good and solid, with a clause that you may occasionally have a match that doesn't match up.
Edit: Seen as I'm talking 2 matches out of all the matches I can recall since you could see PSR changes... I'd say it's very accurate so far.
Edited by Tesunie, 24 October 2015 - 09:33 PM.
#83
Posted 25 October 2015 - 01:55 PM
#84
Posted 25 October 2015 - 08:32 PM
Titannium, on 25 October 2015 - 01:55 PM, said:
Thanks Titanium for your update! I hope the info I shared helped you make positive progress! Good work! I am about 40% through T2 moving toward T1 myself.
I agree with the Devs at PGI, this is NOT a ranking system. This is a rating system. It does NOT rank players. Rather, it rates players according to their contributions to matches. Wins make it easy to climb higher, on losses, one must outperform the rest of the team significantly to go up and still perform well to not go down in rating.
I have noticed the magnitudes of the decreases have lessened from the first week the PSR system was introduced. Ultimately everyone can make it to tier 1, theoretically.
I discount that these watermarks are wrong until there is proof that they are wrong. I suspect without a screenshot that those making claim may have recalled their damage and not their match score, an honest mistake. I have had hundreds of matches and none of the match scores/PSR movement pairs fall outside the findings I have presented, the only exception might be on a boundary case of one match score point, and I think I made those corrections correctly, but leave it open. Ofc, PGI is perfectly capable to change these watermarks at any time since they have not published the values. I suspect we, the community will notice if and when that occurs.
Cheers
Lynx
#85
Posted 25 October 2015 - 09:12 PM
7ynx, on 24 October 2015 - 09:28 PM, said:
So far, YES. None that buck the trend. For review, the following seem to be some of the rules for PSR change:
W, <100: =
W, >= 100: +
L, <250: -
L, >= 250 & <400: =
L, >= 400: +
Things we don't seem to know yet:
What score threshhold in a WIN leads to a drop in PSR?
What are the break points on either count for the difference between a SMALL and LARGE change in PSR?
SO, that established, so far ALL of my matches have met the criteria.
So far, over 16 recorded matches since the 13th of October, I'm noticing a couple of potentially interesting things:
Three of my four best match scores had one or more SOLO KILL points.
The two highest TEAM DAMAGE scores were on two of my top four match scores.
Protected [weight_class] score is higher on matches where I finished with a higher score, generally.
Not sure, again, what in-game actions constitute "Protected [weight_class]" or "Protection Proximity". Could use some clarification, too, on "Brawling".
I'm trying to gain some understanding of how the various components are weighted into the match score, though it's fairly clear so far that the match score IS the PRIMARY driver of PSR tier movement. It's also possible that a match's score alone may not be the SOLE factor in deciding PSR tier movement, owing to the anomalous cases mentioned in previous posts. I'd ESPECIALLY love to see one of those outlier cases broken down point-by-point.
Sure, it'd be nice if we just HAD the equation/formula for summing up our performances for purpose of match score determination. We don't. We CAN observe, though.
Also, I think I should start tracking the tech base and weight class of the mech I pilot in each of those matches.
AND ONCE AGAIN, I offer that I will provide you a link to the document, stored in a public DropBox account, to access and add your own data, that we'll have a bigger sample size to work with. For instance, more matches with 0 TEAM DAMAGE means more we can infer from the remaining match score elements, and basically a control group for comparing matches WITH TEAM DAMAGE. And so on.
It's been suggested before that the match score includes 1/2 point per point of damage. Is that true, do you know?
Edited by TheRAbbi, 25 October 2015 - 09:45 PM.
#86
Posted 26 October 2015 - 12:59 PM
TheRAbbi, on 25 October 2015 - 09:12 PM, said:
AND ONCE AGAIN, I offer that I will provide you a link to the document, stored in a public DropBox account, to access and add your own data, that we'll have a bigger sample size to work with.
Sorry Rabbi, I thought you were speaking to others on this thread with your offer. I feel my findings are complete, that is why I posted this PSR guide if you will. You are welcome to add to it. Everyone is welcome to add to it. But I ask that before you claim my findings are wrong, prove it, or go post in your own thread your own findings. Sorry, but I have no respect for other folks who cannot back up their challenges. I thank you for confirming my findings that falls in line with most folks I've been working with on this. Those few who still disagree and argue offer no proof in their challenge. They are entitled to their opinion.
Edited by 7ynx, 26 October 2015 - 01:01 PM.
#87
Posted 26 October 2015 - 01:07 PM
TheRAbbi, on 25 October 2015 - 09:12 PM, said:
I said it seems as if that is what is happening. Yes that is true, it seems to be about there, but really kind of hard to make that a strong assertion when it is not easily measurable. It is more an empirical thing, a pattern I have noticed after hundreds of matches. IT might not be 50% it might be 40% or 45%... really difficult to discern the exact relationship, but what is certain is that Damage is the biggest contributing pie slice to the pie that makes up match score. (exceptions made for exceptionally terrible shots;)
#88
Posted 26 October 2015 - 01:09 PM
#89
Posted 26 October 2015 - 02:06 PM
7ynx, on 26 October 2015 - 12:59 PM, said:
...
Those few who still disagree and argue offer no proof in their challenge.
Although I've found almost every single match (with only two exceptions out of all the matches I've played so far) to match your data, I will add, where is YOUR proof? I've heard you talk about it, but never provide anything. (And yes, I know it would be a mountain of stuff.)
I'm not asking for your proof, but I don't like how we who happen to come across a strange oddity need to have definitive proof, or we are wrong (this is how you sound). But the same is not expected out of you?
Really wishing I had screen shot that 330 on a lose PSR went down end of match score. I've never had it repeat so I don't know, but it did happen. (I also thought nothing of it at the time.) Then again, they also could have done adjustments since then, as that was one of my first few matches when you could see your PSR adjustments. (I remembered it because I thought it was really strange to do so well, and still go down in PSR on a loss.)
Edited by Tesunie, 26 October 2015 - 02:07 PM.
#90
Posted 26 October 2015 - 08:50 PM
Tesunie, on 24 September 2015 - 08:41 PM, said:
Just saying. Don't now why then, but I had a red down arrow when I lost. I only had the yellow equal symbol once when I had a rare (for me) match score in the 400s. (Which does match your data.)
Just want information to be as accurate as possible, even if I happen to be wrong.
I just don't feel it's right to preform better and lose and drop in PSR, and then preform worse and win and increase PSR.
I had tried to be as gentle as possible and got called names for it. If you are going to make a claim and then toss it up that you might be mistaken and I have a digital ton of work into this guide across dozens of players then I hope you'll understand why I stand by my position and still try and respect your opinion.
#91
Posted 08 December 2015 - 03:58 PM
#92
Posted 08 December 2015 - 08:05 PM
7ynx, on 24 September 2015 - 08:12 PM, said:
This article is meant to present information PL has helped me to deduce through analyzing post match scores and player stats post match.
There are 5 tiers, the lowest being 5 and the highest tier being tier 1.
We do know PGI's distribution model is just fixed PSR watermarks to meet. As the PSR acual values are hidden, PSR watermarks used are irrelevant since all one sees is a PSR Tier progress bar on one's home page. After each match, under the player stats tab, a Green chevron pointed up indicates a PSR increase, a Yellow equal sign indicates no PSR change, or a Red chevron pointed down indicates a PSR decrease for that given match performance as measured by match score.
We have through trial and error deduced some of the watermarks for match scores on wins and losses that trigger a change or no change in one's PSR.
On a Win:
PSR goes UP a lot if match score is >400
PSR goes UP a moderate amount if match score is >250 but <=400
PSR goes UP a little if match score is >= 100
PSR has NO CHANGE for a match score <100
Note: I do not have data on a zero match score if that would cause PSR to go down on a win, but I suspect PSR does not drop on a win, at the very least I have not experienced that.
On a Loss:
PSR goes UP if match score is > 400
PSR has NO CHANGE for a match score >250 but <= 400
PSR goes DOWN a little if match score is <=250
PSR goes DOWN a lot if match score is < 100
On a Tie:
Both sides, all players see a NO CHANGE to PSR regardless of match score.
As you can see, winning makes it easier to improve PSR.
However, PSR can go up steadily if you can consistently match scores above 250 and 400 on losses.
So what goes into match score?
The short answer is almost everything from Spotting, flanking, protected formations, shooting down enemy UAV's to kills, assists, damage, etc. See more on this here.
Currently, the single largest contributor seems to be damage dealt. Assists seem to be the second biggest contributor and then kills, the rest are very small contributors in the present state of PSR (September 2015), and may possibly change over time.
It appears to me that about half the damage dealt contributes straight away to match score with assist, kills and all those other things making up the difference. Sometimes a PSR, no change on a loss, can occur without a high damage (>400, meaning obtaining well under 400 damage) if many assists are made. But that assumes the losing team made many kills... which means it was a close match anyhow.
This is what we know at present, with more data it is subject to change. PGI will very likely alter it and not say how since they have not shared the above information watermarks.
Info on PSR from Paul @ PGI
Great work.
I've done a fair bit of testing myself and your numbers are pretty close I must say.
There is one point however about MM on forming teams based on previous WIN or LOSS.
Assuming a T1 player in T5. Technically speaking, this T1 player should be able to breeze through T5 to T3 with ease with the occasional loss of 1 in 3 or 5 games.
However, I have a feeling MM does 'remember' the outcome of the previous game and pair the winner with losers. That explains why this T1 player doesn't breeze through T5 to T3.
The point I want to make is there's no advantage to artificially reduce yourself to T5 and hope for easier games. MM is going to punish you one way or the other by 'remembering' your wins.
Do share your findings with regards to my point.
#93
Posted 11 December 2015 - 01:22 AM
Either way, I am not sure, and not equipped to measure it. But hope you find my opinion helpful.
Edited by 7ynx, 11 December 2015 - 01:23 AM.
#94
Posted 11 December 2015 - 03:42 AM
In early September after playing for six weeks my KDR was sitting at about 0.12. I got better, and today it's 0.51, and individual mechs I bought after about October 5 all have KDRs running from 0.80 to 2.1, and I've been in groups a lot with my team. Best mech overall, KGC with four uAC5, it's a monster. But I like mechs that aren't considered good, like the Catapult, Dragon, Victor, and Summoner, so I don't wreck face too much but I've become somewhat competent.
I get green arrows over red by a factor of about 2:1, and I've had several matches where I clocked +750 damage and got a green arrow when I'm on the losing side of a 12-3 wipeout. I hid PSR in my loading screen ages ago. So I checked last night. Still T5, and it's only gone up a tremendously small amount, maybe a couple millimeters on the bar.
Lot of Underhivers actually aren't bad pilots. Some are the best drop callers I've worked with. Something odd about how PSR is calculated, I think. You can't really put a lot of stock in it other than the consistent T1/T2 crowd that play all the competitive matches. They're uniformly monsters on the battlefield. Glad I don't see many of them outside group queue and CW.
Edited by Chados, 11 December 2015 - 03:46 AM.
#95
Posted 11 December 2015 - 05:35 AM
Chados, on 11 December 2015 - 03:42 AM, said:
That's something puzzling me as well.
#96
Posted 11 December 2015 - 06:45 AM
Jeffrey Wilder, on 11 December 2015 - 05:35 AM, said:
That's something puzzling me as well.
I may be wrong but I don't think CW counts towards PSR. So there are probably some players who spend the vast majority of their time in CW, where comms are important, but have a low PSR Tier.
#97
Posted 11 December 2015 - 08:26 AM
Moomtazz, on 11 December 2015 - 06:45 AM, said:
I may be wrong but I don't think CW counts towards PSR. So there are probably some players who spend the vast majority of their time in CW, where comms are important, but have a low PSR Tier.
You are correct. No PSR in CW. It counted in the beginning but they removed it. No idea if they adjusted individuals PSR to account for that after doing so.
#98
Posted 11 December 2015 - 01:27 PM
Now not everyone wants to increase their PSR, I understand that, but then I am not sharing this information for those players.
Last millennium, nearly a half century ago, my father explained to me about the pitfall of average. To be average he said, is to be the best of the worst or the worst of the best, take your pick. If you are going to do a thing, strive to be the best. I enjoy that journey, I am not always the best, in fact only in the confined company of average folk am I ever the best at anything, and even then I may often start out as the worst. However I always strive to be the best and be at my best! That is the fun journey!
#99
Posted 18 December 2015 - 11:05 AM
I then got lucky and was on the winning side for several matches, and bumped up to tier 4, which has markedly better skill players than tier 5. The whole point of my post is that it seems that the system is really weighted heavily on overall win/loss and damage count, which skews the ratings for lower damage/support role mechs like mediums and lights. At the higher tiers, I can see the win/loss being more important, because those players already know how to work together and the game (I'm assuming) becomes more strategic than tactical. (Sorry, I'm using chess terminology here. tactical (i.e. there's an atlas RIGHT THERE!!!) and strategic (controlling certain map sections to get overall domination of the battlefield)). At that level, the win/loss makes sense because each player is presumably contributing fairly equally to the overall victory. But at the lower tiers, it really doesn't account for anything other than being able to Mike Tyson an opponent to death, rather than take them apart like Floyd Mayweather. (maybe boxing metaphors will help)...
Anyhow, it seems to me that one way that it could be improved would be to base the PSR by comparing the player to the other players using the same weight class mechs in each match. I wouldn't think it would be too difficult to program that, as the game already segregates lances by weight class, so they are obviously being tracked. Just break the PSR down by the variable already being tracked. That way, a light mech who does an amazing job scouting and distracting with small damage hits would get a better PSR than one that just wandered around without doing much and lucked into being on a winning team.
Seems to me that would do several things: First, it would rank up good players who may have low damage scores due to mech selection, even if they are stuck on a team where every heavy goes to a different point on the compass. Second, it would be a more meaningful rating (again, at lower levels where learning is really taking place) for each player based on their preferred playstyle. Third, it would increase the overall enjoyment because players would be matched up with players of their actual skill level, not their damage output or tanking ability. I got sick of losing because everyone would just spread out and try to play COD with robots, so in my frustration I jumped way out of my depth into the factions, where I caused probably the same problem I was trying to avoid. If the tier ranking would have been comparing my performance to the other mediums, I would likely not have had that frustration, as I was doing better than the other tier 5 mediums.
Again, please remember that I am only comparing myself to the other brand-new players for the first few days I played. I am by absolutely NO means a good player yet, but I am consistently and intentionally trying to better my skills. I am just remarking on what my observations have been as a brand new player who is trying to figure out how to get better. Take it for what its worth, which isn't too much. Sorry for the encyclopedia post and random metaphors. Good hunting!
#100
Posted 30 December 2015 - 10:29 AM
7ynx, on 08 December 2015 - 03:58 PM, said:
I've been tracking this, and my findings match 7ynx's completely, with zero anomalous matches.
It's say, at this point, if you're going to argue 7ynx's findings, you must post relevant screenshots to do so. Anything else, and the assumption will simply be that you're mistaken.
If there ARE findings that are different, I'd be very interested to see them.
26 user(s) are reading this topic
0 members, 26 guests, 0 anonymous users