abraker's Osu! Log

abraker · ^

[PLY] 02/29/16 - <MANIA> Finally passed FX - Madness P.3 [Lv.4 - Sketching You! [4K Stamina Test]] by [FX] AEM with a score of 665k (88.26%), first recorded try 04/12/15

Last edited by abraker (2016-02-29 06:03:39)

abraker · ^

Today I spent the majority of the time modeling the new skill that is going out hopefully by the end of the month. No it's not the reading skill, and I want to to keep it a surprise. What I can say regarding it is that during my research of the reading skill, there was one thing that didn't go nicely with everything else. When you read, you process what is happening, and this particular thing does the opposite. Ok, that hint was too much already.

I spent the rest of the time playing mania. I left the settings on speed 33 fixed and my native resolution. It was previously speed 32 and the lowest widescreen resolution. Why? I figured the thinner strip allowed me to respond to the notes faster since it the notes wouldn't start from my peripheral vision. I managed to make many new highscores that way, but a time came where I couldn't make new scores at the same rate. My reading skill got capped. By making the entire field thinner vertically, all of the notes get bunched closer. I can counteract that by raising the scroll speed, but there is a limit to how fast I can react to the notes. So there must be a balance between my reading skill and my reacting skill. It happened so that today I was feeling fatigue (had a good weekend sleep) and achieved one of my peak response times. I wouldn't bet that would be the case tomorrow though :(

I finished mapping the standard map based on pseudo-technical hold-jump style. Sadly, it's not rankable. I am yet to make more difficulties, so perhaps I might have those diffs follow the rank criteria. You can check out the map here. Currently there is one finished diff, WIP. The rest are placeholders for some difficulties to come.

I don't know why, but I am not feeling like modding maps in my mod queue as much anymore. Ofc, I will mod m4m maps since that is to my benefit, but idk... just don't have the motivation to. I think this is me loosing hope in the ranked maps. Most mappers, including me as of right now, can't create a good map. It has to have some jack midstream or some other bullshit to justify its difficulty. Mappers as well as the staff *cough* loctav ztrot *cough* don't understand what makes a good map, and force criteria to produce half-assed crap. It's not as bad a autoconverts, but I think it's time to see a bit more quality. We were at near edge of going the other direction recently, and thanks to community we didn't.

I proposed a couple solutions to several matters, one of which I posted on the forum thread (when I find it). The first solution which I posted deals with resolving the lack of and surplus of certain difficulties and the suggestion is not perfect. It suppresses the the surplus and does little to encourage mappers to make more of what is lacking. A bottleneck effect results, and we are left with a loose-loose situation. My other suggestion deals with how to make kudos relevant again as well by making them mapping/modding pseudo-currency. You can see that suggestion here. Tell me if you think it's good or if it can be improved. I am planning to post it, but want a second opinion first.

I guess that pretty much sums up what happened between this and last post regarding what osu! related things I'm doing. I just realized that I put another hint regarding the new skill. I guess it's not a big secret after all, now is it?

abraker · ^

[PLY] 03/01/16 - <STD> Have 100 C's remaining
[PLY] 03/01/16 - <MANIA> Achieved 3,300 S's (counting autoconverts)
[PLY] 03/01/16 - <MANIA> Finally passed sun3 - Morgenglut 2012 [Challange] by blazziken with a score of 588k (85.67%)

Last edited by abraker (2016-03-02 05:15:19)

abraker · ^

[PLY] 03/09/16 - <STD> Highest star passed: 4.21*; SID - Ranbu no Melody (TV Size) [Happy's Insane] by Kiddo-Kun with a max combo of 109/439 (81.59%)

abraker · ^

Been working some more on osu!skill, mainly trying to figure out how each skill interacts with other skills and how to measure the reading skill. I made a model which is supposed to describe the process a player undergoes when performing a note or a pattern. There are several stages: visual delay -> initial reaction -> pre focus reading -> focus -> post focus reading -> action -> hardware delay. I won't go into specifics, but that's the model I came up with. This flow applies to a single note though. I have a very rough sketch of a 2 note model and it gets exponentially complex with every note added.

So far, I should be able to simulate the a 3 note system, but going beyond that is difficult due to uncertainty of how the 4th note and beyond affect the reading ability of every other note. That's where I'm at right now. Next I hope to do one of the following things: try and see how a three note model would look like and figure out any edge cases and/or some other idea. It's just a matter of time until I figure out how to measure reading.

I've decided to base the metric of the skill points as follows: 1000 skill points will be the point where the total sum of half of the players would fall below. 500 skill points will be the average of the lower half. For example, half of the people in the world have a reaction time of 240ms or less. 240ms is equivalent to 1000 skill points. The average reaction time of the bottom half is 350ms. 350ms is 500 skill points. Anything above 1000 skill is considered superhuman as far as I am concerned. The distribution is skewed left, so there is rapid degradation in the quantity of people able to perform better. If I have learned anything from the Zipf's law, it is that there will always be that one guy that is outmatches everybody else by a huge margin. So it will be expected to have at most one individual with 3000 skill points or so (I doubt it will be more, but who knows), assuming he/she is legit.

That said, the main obstacle concerning the reading skill is not knowing the point unit where half of the players should be able to do and the average of the lower half. Hence why I need to figure out how to measure reading before I can do anything.

I took the time to redo some parts of my skin. After seeing a video on Youtube with someone having a deep blue skin for the hitcircle, I got inspired to redo my skin's theme to some kind of neon/tech deep blueviolet. I like that color and it has hypnotic feel to it. Neon blue always attracted me in a weird way. I could stare at it for sometime. When the kai time hits in my skin, I get reminded of that feel which is pretty sweet. It's not distracting and rather has better odds of putting me into a light trance. And what happens when that occurs? I start give the game to my subconsciousness while I sorta space out. The point to have less chance to choke due to panic, anxiety or some other shit when seeing a near FC. I am starting to be more proficient at the osu! because of it. I should be able to pass up to 3.8 stars now, from 3.6 stars a couple months ago. This doesn't work on maps that require reading, like heavy overlap or EZ modded maps unfortunately. Need brain cells to do those.

Now to talk about this topic. I got restricted last week. Not 100% what for, but I can only guess. I saw the restricted notice when I was going to respond to emailed notification that someone posted on some thread. As I said, already could guess what got me restricted and went to check a certain thread. It was locked. No message from a mod or anything. A silent ban it was.

I am not sure if that thread really warrants a ban. I explained the situation to that person, hoping he would change his ways. I also wanted to let the community be aware that there were still scammers out there asking to test out their game. Yes, I did leave the name uncensored for 21 players or so to see who I might be calling out, but I censored it before it got anymore exposure. I have checked if that player is banned or not at the time of writing this, and he is. I don't know if he continued his ways or it was I who exposed him for that 7 minutes or so I left the name uncensored. Knowing the staff, that might be enough justification.

So how does the restriction affect me? Well for one, I am completely cut-off from the forums. I am unable to maintain my mod queue nor shitpost in the forums. I also can't work on my maps. I had notifications of forum replies I can't answer to. There are also a couple players posting "rip abraker" or something along those lines. A couple players have my contact and were able to ask me what happened. It kinda saddening not able to respond to a call and seeing everybody going about their own way after.

Does this affect my ability to work on osu!skills? Not in the slightest - for now. All of the info and communication regarding osu!skills takes place in a Trello board. Why for now? When and if my reworked formulas go live, I don't expect Kert to understand how they work. That's where I will need to communicate with players to resolve any bugs in the algorithms.

Hopefully, though, I can get unrestricted soon. I sent an appeal request a week ago and got nothing yet. I sent an email and will be waiting for something. If I don't get a reply by Friday, I'll resort to reddit and hope Ephemeral will help me sort things out. I still have several things I want to do before leaving osu! if possible. Revising the SR system, attempting to engineer a cheat proof system, making visualization maps, and attempting to make a new gamemode are the things I hope to achieve. Will I achieve them all? I am optimistic, but not that optimistic. The higher I aim the higher I will need to shoot, and I hope to it will produce better results than if I was to shoot a bit lower.

abraker · ^

Quick update regarding my restriction status. I am currently no longer restricted. According to the email, that thread wasnt the cause of my restriction. Turns out something running on my compruter caused it. Last time I restricted, it was because I was running cheat engine along side osu!. I was doing stuff with cheat engine when I decided to take a break and play osu!. That caused it to trigger the restriction regardless whether I used cheat engine on osu! or not. This time I wasn't running cheat engine. In fact, I have no idea which program might have caused it. My best guess is that some shit is lurking on my computer. That poses a problem. Without being able to identify the source, it will take me to play one ranked map to potentially get restricted again. I emailed them back asking for more info regarding the application. Due to their policy to to divulge any info I don't expect much from it, but it's worth a try. I guess I will be performing virus scans in the meantime. What a mess this is. :\

Last edited by abraker (2016-03-10 14:25:21)

abraker · ^

Update 2:

I looked through my processes and tried to remember what I did. I am 60% confident that it was a window resizer. Not sure how it works, but it uses some direct x hacks to force a window that cant be resized to be resizable. I was using it for 02jam, and tried it on osu to see if it can resize osu like letterboxing but with no latency. That might be what did it

Last edited by abraker (2016-03-10 17:19:31)

abraker · ^

[PLY] 03/13/16 - <STD> AR limit: 8.5

abraker · ^

I got unbanned and it seems that the reason I stated in the update might have indeed been the cause. The only hint I got is that "it is unlikely it will happen again".

I made some new scores in mania today, S'ing several 4.4* maps and higher. I also found that my passing limit in standard is currently AR 8.5 and 3.8*. Nothing else notable gameplay wise.

I managed to get a couple mod requests done. I am slowly, no, VERY slowly churning through them. I more likely to accept a M4M request at this point. I would turn it into a M4M queue too, but don't want to rule out any maps in case those actually deserve my attention.

It's almost the end of the month, which means an osu!skill update is going to come soon. This update will increase the scores it captures from top 100 to top 1000 by utilizing this nifty thing I never knew about before. I still hope peppu completes my request to allow it to be top everything.

Desmos is a good tool for modeling formulas, but it has limitations. For one, it's not powerful enough to handle simulations in realtime. The 6 note simulation I set up was going at 10 FPS. So I took the week to code up a tool to allow fast realtime formula testing on any pattern. It allows to move the hitcircles, adjust AR, CS etc with sliders, and displays the results instantly. Wait... that does sound familiar... coincidentally my original username is abraker95... I think kert is onto something!

abraker · ^

[PLY] 03/23/15 - <STD> Highest pp from single play: 48 pp

[PLY] 03/23/16 - <STD> Highest star passed: 4.4*; DJ S3RL - T-T-Techno (feat. Jesskah) [wkyik's insane] by nold_1702 with a max combo of 69/804 (84.55%)

[PLY] 03/23/16 - <STD> AR limit: 9

Last edited by abraker (2016-03-23 22:08:44)

abraker · ^

[NEW ACHIEVEMENT]: Highest star 8k NM B'd

New:
[04/01/16] pic; replay
4.24*; Tou Chi Chen - Secret Kakuranger TEKINA Remix [8K Ninja] by -Kamizake-

Prev:
[12/14/15] pic; replay
4.14*; Hecq - Dstrukt [D's 8K Insane] by YunoFanatic

Last edited by abraker (2016-04-01 18:09:15)

abraker · ^

[PLY] 04/02/16 - <MANIA> Finally passed Owl City - Shooting Star [Hinagi's 7K SC] by NaxelCL with a score of 684k (90.91%), first attempt 2014?

abraker · ^

I am becoming ever more lazy with my mod queue. The overwhelming number requests look like static to me now. At this point I am likely to accept a request that stands out from the rest - That being a M4M or M4S request. To try fix this, I decided to have players pm me in-game if they want me to do the request on the spot. I set up rules and time I would available as well. No players so far pm'd me which is either good or very bad depending on how you look at it. It's good because I am more very more likely to take a request that someone pm's me. It is very bad because it doesn't seem to be effective. I guess I'll have to motivate myself somehow else.

For April fools I did two things. A week before April fool's I made a thread that is saying I am developing a new game called Osue! and have available job positions. I based the job positions on the common osu! meme culture, which include shitposter, moderator, admin, cheater, developer, modder, etc. On April fool's day I ended the thread and pm'd whoever posted on the thread and was in-game that their application was rejected. I expected no less from Railey2, who took the act to the next level.

▼Hidden text

The second thing I did for April Fool's is post a creepy pasta on the osu!forums and /r/osugame. This thing was actually in the works since September 27, 2015. I proofread it several times to make sure there were no mistakes (one slipped though). I didn't know when to post this thing actually. The idea hit me to post it on April Fool's was when seeing [Taiga]'s "Creepypasta for 1st april" thread. I didn't want to post it there, however, because it would be lost to shitposting. I felt it needed its own thread. And boy did everyone get a kick out of that thread - but not because of the creepy pasta. Zalgo invaded the OT subforums and had a several hour raid before Deadbeat killed it. While, it vandalized my thread, I kinda liked the idea that the creepy pasta became even more creepier through Zalgo's corruption. Nobody would read it ofc, but that was not the point. The point was for it to set a theme and mood. The mood was lost since Zalgo all over the thread was funny and not creepy, but the creepy/corruptive theme remained which made me relieved.

abraker wrote:

Today I spent the majority of the time modeling the new skill that is going out hopefully by the end of the month.

The skill I was referring to is reaction and I was wrong to assume it will be done by now. What seemed simple at first became a bit more complex. I am at the stage where I am testing how reaction values look on no modded maps, not taking account sliders. It gave me values higher than I expected, so I'll have to figure out what I missed. To test, I had to go back to the osu!skills code base after several months. I knew it was a mess before, but looking at it now after I did the osu!skill Formula Tool coding, it's a total utter mess. So currently I am spending some time refactoring the shit out of that code. Following the refactoring I will transfer the beatmap file reading and processing code from osu!skill to the tool so I can inspect the calculation process more closely.

I had a lengthy talk with Endaris about a week ago. The convo involved what the skills are lacking, what it is getting wrong, and some theory behind how reading works. It made me decide that the agility and precision skills need to go and have a targeting skill instead. The tenacity skill also needs to be reworked. Targeting will be sensitive to velocity, distance, hit area, and time interval. Agility currently is only sensitive to velocity and distance, and precision is just sensitive to hit area. What non of them take account is, how Endaris said, speed control. Agility sees speed, but it's just speed. Precision sees how small things are and that's pretty much it. Those two skills do not see for what time interval you need to maintain a precision given the agility. As such, having agility and precision will not work there needs to be one skill I call targeting.

Tenacity will be changed to not only focus on how long and fast a stream is, but how big the tapping imbalance in the pattern is. Let's start with this: what is harder to maintain tapping for: 180 BPM stream or a 130 BPM stream? If you say 180 BPM stream then you are wrong. 130 BPM is easier to tap but certainly not easier to maintain. Low BPM streams tend to increase room for inaccuracies with tapping while stream like 400 BPM have little room the inaccuracies, making it either hit or miss and nothing in between. Why? OD. OD pretty much controls this aspect of accuracy. A 400 BPM with some OD much larger than 10 would comparable to a 130 BPM stream on OD 10. You need finer control of the tapping. Not tapping precision, tapping control and control would be precision over a period of time.

Now, streams are not the only patterns that are hard to maintain precision control on. Your tapping precision may be rated such that you can do OD 10 no problem, but if that is assuming that the pattern has constant spacing intervals. Remember that different BPM stream under different OD require different requirements over tapping control? Well every pattern can be interpreted as a stream that changes time interval (BPM) on every note. This gives the skill requirement on maintaining precision over a series of notes with varied time intervals, aka tapping control for everything.

I am also thinking of renaming stamina to speed, but other than that, it's not going to change much. What tenacity doesn't see is how fast you need to tap, while stamina does. Endaris actually made realize there is an uncanny valley between tenacity and speed. Higher BPM streams tend to have a higher speed requirement, lower BPM streams tend to have a higher tenacity requirement and lower BPM still have tend to have a lower tenacity and speed requirement. Low -> High -> Low -> High. As such, there is only a certain range in the mid BPM and in the very low BPM where players prefer to tap to.

The hardest step would not be calculating the map's skill requirement, but the player's resultant skill for a play. The system currently calculates the skills based on how many misses the player got on a map, which is VERY inaccurate. To calculate the resultant skill do to a play, you need to examine the replay and see where the cursor is when the player taps vs where it should be when it should tap. This would be the first version of the replay skill calculator. The second version would take account what happens in between the where the player taps. It would take account what and how the player did to get to the point where and when hitobject needs to be hit. That way skills would be rewarded even for cursor dancing. Unfortunately, I do see the latter being heavy on the processing time, so I'm not sure how it would work if replays come in one per second at least. Multi threading probably, and dunno if that will even be enough.

I think everything is starting to come together and the finalized skill theory is starting to emerge. A bit more to go. While I am not sure whether I will finish the reading the skill by the time I predicted, but July still looks like a realistic deadline to me.

Last edited by abraker (2016-04-03 21:36:05)

abraker · ^

[PLY] 04/10/16 - <STD> Highest star passed: 4.5*; Cres - End Time [Insane] by Kyshiro with a max combo of 134/754 (79.78%)

[PLY] 04/11/16 - <STD> Highest PP gained: 49pp; Roberto Colombo - Tutti Beautiful [insane] by Leader with 1 try (90.00%)

abraker · ^

[PLY] 04/12/16 - <MANIA> Finally S'd Remixed by DJ Command - Mermaid girl -Akiba Koubou MIX- [LNP's Lv.27] by bilqiscutee with a score of 829k (95.40%); first attempt 11/03/14

^ #61 2016-02-29 06:02:29

Re: abraker's Osu! Log

^ #62 2016-02-29 07:29:59

Re: abraker's Osu! Log

^ #63 2016-03-02 04:00:10

Re: abraker's Osu! Log

^ #64 2016-03-09 18:25:06

Re: abraker's Osu! Log

^ #65 2016-03-10 03:49:33

Re: abraker's Osu! Log

^ #66 2016-03-10 14:24:27

Re: abraker's Osu! Log

^ #67 2016-03-10 17:19:16

Re: abraker's Osu! Log

^ #68 2016-03-15 02:24:33

Re: abraker's Osu! Log

^ #69 2016-03-23 02:50:59

Re: abraker's Osu! Log

^ #70 2016-03-23 22:04:37

Re: abraker's Osu! Log

^ #71 2016-04-01 18:07:15

Re: abraker's Osu! Log

^ #72 2016-04-03 03:49:17

Re: abraker's Osu! Log

^ #73 2016-04-03 21:27:18

Re: abraker's Osu! Log

^ #74 2016-04-11 17:44:47

Re: abraker's Osu! Log

^ #75 2016-04-13 02:53:19

Re: abraker's Osu! Log

Board footer