Sync (@sync_bs) header
Sync (@sync_bs) profile picture

Beat Saber Content Creator, donations accepted @ https://teapaw.org/t.co/JpoUjq6zHJ BL Profile: https://teapaw.org/t.co/9mpAM4asEJ

Joined November 2014
Photo by Sync (@sync_bs)Photo by Sync (@sync_bs)Photo by Sync (@sync_bs)Photo by Sync (@sync_bs)Photo by Sync (@sync_bs)Photo by Sync (@sync_bs)Photo by Sync (@sync_bs)Photo by Sync (@sync_bs)Photo by Sync (@sync_bs)
PPv4 is out! We have been hard at work recently working on PPv4, which introduces dynamic per-map curves, tons of bugfixes, a retrained Acc AI, and has massively improved the worth and balance across all ranked maps. Your front page is likely to look very different now, as ratings are more balanced and more maps are worth grinding. You should see a mix of your best accomplishments, rather than high worth maps you grinded specifically for their rating. This will let you play more maps you love rather than having to compromise for what is worth PP. This also means more skillsets are now more viable on a high level, such as Acc and Tech. TLDR: The new ratings should let you play anything you want and be rewarded with a balanced amount of PP, no matter your rank or preferred style. Keep in mind that the PP scale changed. Ratings cannot be directly compared to PPv3. Dynamic curve To make this possible, we are introducing a dynamic curve with PPv4. The base curve is adjusted per map based on various factors to better match the real progression of difficulty and give out the appropriate amount of pp for the score. This means the same star rating can now give you very different PP on various maps when achieving the same accuracy. This should reflect the actual effort it took to get that score on that map, as you can imagine a linear speed map is easier to 97% than a tech map. The dynamic curve lets us take on issues we couldn't properly fix before. As this is the first introduction of a dynamic curve, we'll be keeping an eye on how this goes and tuning it in the next reweight as usual. Reweighting We have a new reweighter, which is an adjustment step done once after rating updates to nudge them towards the reality on the leaderboards. The new reweighter takes into account a new 'worth' statistic. It's a lot more reliable now and adjustments are done more carefully, several metrics are checked to ensure no systematic biases are introduced. Nevertheless, it has always been the goal to use this as little as possible, as it skews the ratings to fit with overall playerbase skill and how they performed on each map at that time. As ratings are a static measurement of difficulty, it's a major sign of improvement when the reweighter strength can be reduced. With PPv4, it's about half as strong, depending on the map. The ratings without this step look good already, but we are running this reduced strength reweight to remove the remaining outliers. Changelog: Major changes: - Introduced a dynamic per-map curve. - Rebalanced all star and pp amounts and changed the scaling. - Made a much bigger share of all maps worth playing for all players, no matter the rank or style, and improved the overall balance of all maps. - Most extreme outliers are gone. - Tech and Acc maps are worth more, bringing it more in line with other maps and competitive at a high level. - Slider spam is no longer overweighted for lower ranks. - The retrained Acc AI predicts top player accuracy more consistently (+bugfixes) - Long empty sections are now factored into the PP calculation as grinding time. Playing True acc with SF will grant less PP than before as the duration decreases without increasing acc difficulty. Shorter map length can get up to a 10% nerf. - Low note count can get up to a 5% nerf, much reduced from PPv3. - The Analyzer (pass and tech ratings) had general improvement and bugfixes on all aspects (flow, parity, dot notes, difficulty/tech estimation, etc.) Contributors: - Loloppe - Broad idea direction, implementation & tuning - Light Ai - Making & tuning the new dynamic curve, discussing changes and direction - GalaxyMaster - Discussing changes, pointing out issues and solutions - Huh - Discussing balancing iterations - edgii - Discussing balancing iterations - Undeceiver - Math explanations - And plenty more who gave feedback! Do you have a feedback? Please let us know what you think!
3
12
2
49
1,908
He really said : "😼"
40
1,701
62
19,332
153,176
The two biggest weirdos in the zoo, hahaha.
543
22,462
1,531
189,255
7,296,295
hit me some of that milk!
2
49
2
794
10,054
well boys, its time to update the game. had fun on 1.29.1 but i now have no choice
2
9
2
56
7,129
1
109
7
1,110
16,647
猫バージョンあるの初めて知った。これ無限に見てられる
20
2,264
79
17,861
1,202,952
64
542
77
11,849
1,806,216
仕事中の猫の様子が気になってペットカメラを設置した結果
33
431
20
8,248
298,211
Meowww....
20
1
153
2,719
2
51
2
1,438
12,771
23
180
9
2,119
47,716
91
5,118
292
58,523
1,087,561
26
542
44
6,127
61,130
4
247
11
3,135
49,756
mid day yawns
1
25
715
18,296
5
46
1
661
11,272
44
1
523
13,647