Hopefully ratingupdate will be up and running this weekend. Since PC and console players now play together and there's been such a long time since the site went down, I'll be clearing old ratings and starting fresh.
I made a new website for Glicko ratings for Strive:
- More data (up to 2.4M games now)
- Live updates (ish)
- Color coded matchup charts
- Doesn't load 500MB of JSON to start
#GGST
Having a lot of trouble trying to figure out the new patch. Traffic seems to be (differently?) encrypted now so just listening in with fiddler isn't as useful for figuring out the new API. If anyone's knowledgeable about this please reach out either here or on github!
Thanks a ton who ever who helped figure it out and get it working, I definitely couldn't have done it without you.
The DB is fresh so you won't show up until you've played a game (and I still need to manually do all the VIP status stuff)
Probably have some bugs to discover too
Won't be able to update ratingupdate until this evening. Might take a while depending on how the crossplay works - ideally there'll just be one website now.
DNS might be causing issues, I'm on a new server now. Old games and other info are missing but will be around tomorrow when I can start manually migrating stuff.
RU is going to go down for a bit while I pull in the old games from before the last patch. Ratings afterwards will be a bit different, with Celestial players overall being a bit higher and F10 lower.
The website is now available for console players too!
Only spun it up earlier today so it's going to take some time for it to collect games and statistics to start to make any amount of sense as peoples' ratings settle down.
ratingupdate is up on a new server now, hopefully it's relatively stable. People's ratings have decayed in the time it was down so you'll experience slightly larger swings but not by much.
Bit of downtime on the site, updated the player floor distribution chart to only take into account the past two weeks of games since I saw some people get the wrong idea of just how dead the lower floors are. Important thing for finding a match is "games played" of course
Rating update! Everyone's ratings have been reset along with the new patch. Old ratings are still viewable in history, and I'm working on a way of showing old "top ratings" for patches and things like that. Unfortunately I'll be away this weekend so no updates soon.
Experimenting with a change for grappler, the shot "charges up" the longer you swing. (Still autoreleases if you hold it for too long)
#gamedev
#rustlang
Been working on a bunch of changes to how wizard works. He can now "level up" his walls to hit the ball harder and last longer.
#screenshotsaturday
#gamedev
ratingupdate should be fully back online now! Thanks for everyone who reported the issue - it's funny how these things always seem to happen when I'm away from computers.
I've considered just not counting matches if the difference in rating is too high, but this would result in a lot fewer matches counting for the top rated players. Is that something people would be interested in?
updated with a feature I hope won't get much use: cheaters are now labelled and don't have a visible rating. They also won't affect your rating if you play against them.
People fighting other challengers and/or getting boosted definitely increases the variability in celestial. It's actually higher than F10, even! Makes it very hard to find decent matches, especially when you're stuck in a random lobby of just a couple dozen other players.
There'd also be massive variability in celestial skill depending on who gets there first. If a really skilled few players make it in, than they'd have the opportunity to shut everyone out unless you specifically only choose to fight other celestial challenges.
Haven't posted about paddlepunks in a while - just added a new parry variation to homura that "pauses" the ball for a bit before reflecting it back harder than the regular parry.
(The very hard AI is very hard, please understand)
Now that I've got a decent amount of stats at , I've started remaking some graphs I made early on show skill distribution by floor (and overall - for comparison. Skill is normally distributed, like most things, and the average player is, in fact, average)
Buffed the active time on witch's doritos a bit to make them easier to hit with - there were some occasions where they felt unreasonably short before.
#screenshotsaturday
Update: I accidentally rolled things back to may 10th instead of june 10th so I had to restore from a backup and things will take a bit longer to be online again.
Please understand, small indie website
Going to take down the website briefly to apply some changes, including both skipping over one-sided matchups as well as pagination for games (so you can scroll through and see all your past games)
Ratings will stay the same, the changes don't affect them much from testing.
Updated to keep track of ratings all the way up past 2200, since there are quite a few players up there now. Lots of Sol, Nagoriyuki, and Happy Chaos up near the top.
#GGST
RATING UPDATE
has been far more popular than I thought it would be, which has been great, but the server's definitely taking a beating. For people who want to help with that I'm opening
#GGST
Did a quickish writeup on how the new rating system works. , with example code.
I also add an online calculator for rating changes at , this might be a nice tool for getting a sense of how it works.
@tangentarine
I've sadly been pretty busy this month. Basically, the issues are with the server, not the actual rating update code itself. Some updates broke the steam runtime on it and I need to figure out how to restore it to working order in order for it to pull data.
I've been traveling and only recently been made aware of the cheater that's topping the leaderboards, unfortunate that this happened. Hopefully ASW does something about it, but in the meantime I'm just going to nuke their rating so the website isn't affected at least.
I think I've figured out the issue, during the migration I accidentally caused deviation decay to occur twice. I'll be able to deploy a fix this evening that'll bring people's deviations back to pre update levels
Noticed a crazy thing while going through the data on . Normally, you'd expect there to be a P1 advantage since people generally play that in training mode, but in
#GGST
there's actually a minor P2 advantage in tower matches: about 50.5%
Still missing some features from the old site, but I'll update it over time. I'm not webdev, so apologies for the rudimentary CSS and the lack of TLS. There's zero javascript, at least.
While I'm sure some people are excited about the whole algorithm change and being able to see ups and downs in real time, here's the big new feature upgrade: Fraud Index is now calculated in a more accurate way by just taking your mains into account.
RATING UPDATE
Added character popularity stats to , sorted by rating, with colors to show if they're more or less popular at that skill range compared to baseline. Interestingly,
#GGST_HA
seems to be popular at both ends of the spectrum
#GGST
And here's some data I haven't yet added to the site dynamically yet - player ratings split by floor. The rating gaps within floors are pretty huge and there's a lot of overlap between them.
Someone claimed that maybe the AI set to VeryHard is "bullshit" and "should not be allowed in a videogame."
Maybe the reflexes could be detuned a bit but on the other hand is it called VeryHard
#screenshotsaturday
Slight tweak to the site regarding what games count, now if either your or your opponent's deviation is greater than 300 it'll automatically count, regardless of expected outcome. This is to help prevent people getting "stuck" in the wrong rating.
@GideonTheGreat5
Negative rating isn't a problem, that's just how the system works. In practice pretty much nobody gets down there because you'd need to be truly garbage against the initial rating, but you could just as well based the system around 0 or even like -1000 or something
Weird thing happened, looks like someone's tried to actively manipulate their stats by just playing hundreds of games against themselves on F10. I've made some changes to make this harder to do but mostly it's strange to me that someone would even take the time to do it.
Grappler didn't really need buffs but I gave him new moves and some speed increases anyway. I'll try and bring other characters up to his level
#screenshotsaturday
#rust
@GriffyBones
The guy's an idiot but he's right that if you're matchmaking in the lowest rank in the game you should be placed against people equally bad, SFV's ladder is horrendously bottom-heavy. I'm not away of game game outside of FGs that have matchmaking this bad.
Experimenting with a super for the witch character. By using the super you buff all your blocks, giving them faster startup times, harder hits, and bigger hitboxes
#screenshotsaturday
Server was taken temporarily offline for maintenance by the VPN provider and I'm not well versed enough to have startup stuff that'll automatically launch the website.
(Personally I'm not looking forward to figuring out how to merge the two databases after they've both have tons of games already, but I'll cross that bridge when I get to it)
As part of the rating reset, I've updated the algorithm to give better predictions for outcomes. Previously, it would overestimate your win chances at high rating differences and consequentially you'd lose more than you should for dropped games against lower rated players.
I already knew the tower was bad but when browsing around and testing out the feature I found people with ~1600 rating that had sub 40% overall winrates on their mains, just because Celestial is pretty much a free-for-all. Skill based matchmaking in fighting games when?
I had a request for someone to add overall win rates (including adjusted) to ratingupdate, so those are now included at the bottom of your matchup chart. Turns out some people aren't having a great time!
Other changes are deviation decaying much slower than before, so you're not forced off of leaderboards just from a few days of not playing. It also keeps track of the highest rating you've achieved, as well as the top rated opponent that you've beat.
Of course there's also the risk that they change how the API works entirely and break everything I've been doing, but I'm going to hope that doesn't happen.
@Wool_Mane
@jiyunaJP
Sadly this, the reason you have younger players doing well in other games is because there's just so many players coming into the scene on a regular basis.
Strive has 50k active players total, across platforms/regions.
Valorant has 20k players in just the top 0.5% of EU.