Showdown: Nightbird Vs Jayz Vs Cluster Fox - A Psr Comparison
#41
Posted 13 August 2020 - 02:44 PM
Its great work, time and effort.
Is P.G.I going to change on the basis of this ?
Will if it's adopted make people return, stop leaving at a slower rate ?
Wargaming couldn't make a match maker work with vastly more players and money.
They gave up on warships with far less customisation options, and went to BVP
Its ******** than what PGI have now.
#43
Posted 16 August 2020 - 07:40 AM
Standard deviation of games played is 1217. Aka: there are still a lot of players in the 10-20 games played range.
Results (600k):
JayZ with stability has only improved by 10%.
Both CFox and NB (no MA) show a 30% improvement.
Nightbird's WLR MM (no moving average, using raw WLR as the PSR) still gives 30% better results.
Discussion:
NB system with moving average.
In the same start conditions, once a moving average is introduced in Nightbird's system, CFox and NB become equivalent ways to calculate PSR within a 2% spread.
This is due to the fact that (W-L)/Games and WLR are two different ways of expressing the same variable. The spread comes from the addition of AvgMS in CFox's system.
What works for seeding P and A in CFox's system also works in Nightbird's system. Seeding A would alter the value of a win/loss event, seeding P would change the period leading to faster response.
In other words, once a moving average is used, it becomes the bottleneck.
Nightbird :
Raw WLR is 30% better for the same amount of total games simulated, with the largest benefits for players with a low amount of games. It gives a very quick movement of PSR with few games (desirable), then becomes almost a fixed value as more games are played (undesirable).
With your WLR system do you recommend using raw WLR - with the effect of reduced PSR movement over time - or a moving average solution with possible seeding?
Edited by Cluster Fox, 16 August 2020 - 07:58 AM.
#44
Posted 16 August 2020 - 10:35 AM
Still a bad system but will adjust for improving skill and will drop players into T5 almost instantly.
#45
Posted 23 August 2020 - 03:27 AM
#48
Posted 23 August 2020 - 01:41 PM
#49
Posted 25 August 2020 - 07:17 AM
VonBruinwald, on 23 August 2020 - 12:06 PM, said:
Unfortunately the ones who lobbied for the "community" decision have a direct line with PGI and are in a position to lobby against stickying a thread proving they were wrong.
That I really doubt, I actually don't think there are any real back doors involved here. imo it's more likely that pgi jumped on their alternative because it required very little work to implement but still gave the impression of "listening to the community", the buddy politics involved here was more on the reddit/discord side.
The question is can we move on to implementing a functioning system now that we know this didn't work?
#51
Posted 25 August 2020 - 01:10 PM
Sjorpha, on 25 August 2020 - 07:17 AM, said:
So PGI are trolling:
"So we have two suggestions, this one works, this one is fundamentally broken and requires no real work".
"These guys are so dumb, who cares if it works. MINIMUM EFFORT!".
"What do we tell them when they realise it's broken?"
"Who cares, they picked it "
I can see that happening...
#52
Posted 25 August 2020 - 02:30 PM
VonBruinwald, on 25 August 2020 - 01:10 PM, said:
So PGI are trolling:
"So we have two suggestions, this one works, this one is fundamentally broken and requires no real work".
"These guys are so dumb, who cares if it works. MINIMUM EFFORT!".
"What do we tell them when they realise it's broken?"
"Who cares, they picked it "
I can see that happening...
Do not forget that Paul's revision only had movement for top 8 and bottom 8 and no movement for the middle 8 on one of his versions.
And the reason PGI even thought about doing a PSR change was only after the outcry once we had the soup queue. Then PGI was going to do the reset with static PSR thresholds and the thresholds would have been mirrored images of each other, and without even breaking up the 101-200 and 201-400 thresholds. And in the PSR Update and Changes 6-03-20 thread, Paul's discussion was that Players were moving up if their max score met or exceeded 401 MS on a loss, without even tieing in that lower end performers were moving up faster than moving down. That thread and others, as well as Twitter, is where PGI decided to change the static PSR threshold to a more dynamic, based on what happens in each drop, USING Matchscore.
Again, PGI was looking for triggers and setups that dealt with MS, Paul was not looking for anything that would require a complete rewrite, and doing it by players voting with the LIKE THIS button in the new thread. Jay-Z was on the first page, Nightbird's was like page 13? and did not dealt with MS at all, then stepped back and could use MS for the bars of the tier levels but handle the match maker using W/L.
Nightbird's post from earlier in this thread does bring to question how a W/L setup would work though? Why would the MM select all of the worse players selected to team up with 4 "elite" players? If that side wins, the 4 has a higher W/L and brings the low end up by one match.. Mind you, I see nothing about how if the worse players are the worse due to primarily pumping out low damage for low MS while using light/meds, horrible in the mechlab/aiming, etc?
Nightbird, on 10 August 2020 - 08:56 PM, said:
Speak for your self... WLR will make it as hard as possible to stat pad by grouping, since after a point, your group of 4 will always be teamed with the 8 worst players out of the remaining 20. If you can still rack up wins with that, great, you deserve it. Most 4 mans can't and therefore will reach some equilibrium. We'd see very few people with over 3WLR under my system. Under Jay Z's, 30WLR, 50WLR, the sky's the limit. The same will be true under your system because you cap PSR and so under-estimate the impact of grouping and high WLR players.
We have PSR because players wanted to see where they stood, and when they won they would be rewarded appropriately based on what they put into that drop. Elo did not provide that info, that satisfaction. No right or wrong there, except for PGI decided to implement it. They knew how it would go through, posting eventually everyone could end up in Tier 1, but makes one wonder how they thought that was okay.. but why?
Then ask, did Nightbird's version even had a chance of being viable in PGI's/Paul's eyes? Even their thoughts that, in a game that is ONLY zero-sum for the players in that specific drop, that zero-sum meant that new players should be dropped in the middle of the pack.......PGI's implementation of combing the group and solo queue with its current valves would have been a disaster for any MM for MWO, since there are so many variables involved with this IP, then followed by PGI's minimum viable approach to too many aspects of the game.
The one thing they did get out of it though.. they were able to complete their PvE MW5. Good or bad, eventually when MWO does close up shop, MW5 will still be out there.
#53
Posted 25 August 2020 - 03:34 PM
#54
Posted 26 August 2020 - 04:28 AM
They picked another option
That was not laid out!
#55
Posted 01 September 2020 - 08:26 AM
OZHomerOZ, given what PGI had originally suggested, I'm actually glad they went with a community suggestion. I just think they should have put out more updates to lead the discussion rather than "surprise pick".
Nightbird, I really don't believe there was a vendetta against your solution. What Carl Tabot said is much more likely. IMHO, WLR just -seemed- like more work at the time to PGI, and the whole solution was a little more alien to what had been suggested thus far. Hence unfamiliar territory and perception of risk. WLR was also more controversial even though logic prove it could work, some wouldn't see past the end of match screen.
Bottom line, both Nightbird and I got valid but different options PGI can choose from and hopefully PGI takes the steps to make that PSR stable. Until then, we can only emphasize the results and benefits to the game to make it clear it's worth the change.
Edited by Cluster Fox, 01 September 2020 - 08:32 AM.
#56
Posted 02 September 2020 - 01:45 PM
Cluster Fox, on 01 September 2020 - 08:26 AM, said:
Nightbird, on 23 August 2020 - 01:41 PM, said:
Edited by Nightbird, 02 September 2020 - 01:46 PM.
#57
Posted 04 September 2020 - 02:18 AM
#59
Posted 04 September 2020 - 11:04 AM
I O O percent KongLord, on 04 September 2020 - 02:18 AM, said:
Your perception is off, the further one lies towards an extreme the narrower their perception of the larger spectrum. Blinded by arrogance is the common term.
#60
Posted 08 September 2020 - 04:58 AM
I O O percent KongLord, on 04 September 2020 - 02:18 AM, said:
When your a solo in a room full of groups it can seem that way
Chain fire FTW
8 user(s) are reading this topic
0 members, 8 guests, 0 anonymous users