Maintenance for the week of November 10:
• PC/Mac: NA and EU megaservers for patch maintenance – November 10, 4:00AM EST (9:00 UTC) - 9:00AM EST (14:00 UTC)
• Xbox: NA and EU megaservers for patch maintenance – November 12, 4:00AM EST (9:00 UTC) - 12:00PM EST (17:00 UTC)
• PlayStation®: NA and EU megaservers for patch maintenance – November 12, 4:00AM EST (9:00 UTC) - 12:00PM EST (17:00 UTC)
• PC/Mac: NA and EU megaservers for patch maintenance – November 10, 4:00AM EST (9:00 UTC) - 9:00AM EST (14:00 UTC)
• Xbox: NA and EU megaservers for patch maintenance – November 12, 4:00AM EST (9:00 UTC) - 12:00PM EST (17:00 UTC)
• PlayStation®: NA and EU megaservers for patch maintenance – November 12, 4:00AM EST (9:00 UTC) - 12:00PM EST (17:00 UTC)
Interesting article about the hardest coding challenges
krachall
✭✭✭✭✭
✭
✭
According to an industry website on coding challenges, here are the top five hardest things known to man to code:
1. Aanderaa-Karp-Rosenberg evasiveness conjecture
2. Cryptocurrency algorithm replication
3. Space vehicle intercept trajectories
4. Integer factorization in polynomial time
5. Video game player grouping
So cut ZOS some slack, they've only been working on it for 5 and a half years!
1. Aanderaa-Karp-Rosenberg evasiveness conjecture
2. Cryptocurrency algorithm replication
3. Space vehicle intercept trajectories
4. Integer factorization in polynomial time
5. Video game player grouping
So cut ZOS some slack, they've only been working on it for 5 and a half years!
8
-
randomkeyhits✭✭✭✭✭That just opinion, not fact.
As an example NP complete solutions for various problems may exist but no one knows how to program them, not on that list.
Its all down to the criteria you use to rate difficulty and there is no agreement on a definitive list.
And yes, I think Cyro loading screens should be on the list too.
EU PS40 -
SirAndy✭✭✭✭✭
✭✭✭✭✭According to an industry website on coding challenges, here are the top five hardest things known to man to code:
1. Aanderaa-Karp-Rosenberg evasiveness conjecture
2. Cryptocurrency algorithm replication
3. Space vehicle intercept trajectories
4. Integer factorization in polynomial time
5. Video game player grouping
So cut ZOS some slack, they've only been working on it for 5 and a half years!
Actually, on top of that list should be "Hello World" using the Brain[Snip] programming language.
[Edit to remove profanity]Edited by [Deleted User] on October 27, 2019 7:34PM1 -
idk✭✭✭✭✭
✭✭✭✭✭Zos did nothing about the issue for a year and OP wants us to cut them some slack. For a year the forums were filled with threads informing Zos their fix 2 years ago did not resolve the issue then Zos follows up trying to tell us they thought that previous fix worked.
They either outright ignored our flood of feedback or chose to lie about the issue. I do not know which is worse.
Additionally, Rich stated they would not have another event that required the GF until they were 100% certain it was fixed. Considering they did not do due diligence and stress test the "fixed" GF on the servers before the event they could not be 100% certain. Again, their words fell short of what they actually did.
This is not about the GF not working properly. This is about management ignoring us. It is about poor management and possibly incompetence at that level. OP wants us to give Zos a break because they do not live up to their words.6 -
VaranisArano✭✭✭✭✭
✭✭✭✭✭Zos did nothing about the issue for a year and OP wants us to cut them some slack. For a year the forums were filled with threads informing Zos their fix 2 years ago did not resolve the issue then Zos follows up trying to tell us they thought that previous fix worked.
They either outright ignored our flood of feedback or chose to lie about the issue. I do not know which is worse.
Additionally, Rich stated they would not have another event that required the GF until they were 100% certain it was fixed. Considering they did not do due diligence and stress test the "fixed" GF on the servers before the event they could not be 100% certain. Again, their words fell short of what they actually did.
This is not about the GF not working properly. This is about management ignoring us. It is about poor management and possibly incompetence at that level. OP wants us to give Zos a break because they do not live up to their words.
I find it sad and funny that the Groupfinder improvements actually appear to be the thing that was working correctly, only for other things to start breaking under heavy load.
It's also funny and sad that ZOS' workaround for "we won't run events that require Groupfinder" has just been to rewrite the dungeon events so they don't "require" Groupfinder. With the result that the only canceled event was Midyear Mayhem because of Battlegrounds.
What's just sad is that everything else seems to have broken for more or less the same cause as earlier this year - that PC/EU has a high number of concurrent players during events. Whatever improvements ZOS made to handle that earlier don't seem to have held up for very long.2 -
idk✭✭✭✭✭
✭✭✭✭✭VaranisArano wrote: »Zos did nothing about the issue for a year and OP wants us to cut them some slack. For a year the forums were filled with threads informing Zos their fix 2 years ago did not resolve the issue then Zos follows up trying to tell us they thought that previous fix worked.
They either outright ignored our flood of feedback or chose to lie about the issue. I do not know which is worse.
Additionally, Rich stated they would not have another event that required the GF until they were 100% certain it was fixed. Considering they did not do due diligence and stress test the "fixed" GF on the servers before the event they could not be 100% certain. Again, their words fell short of what they actually did.
This is not about the GF not working properly. This is about management ignoring us. It is about poor management and possibly incompetence at that level. OP wants us to give Zos a break because they do not live up to their words.
I find it sad and funny that the Groupfinder improvements actually appear to be the thing that was working correctly, only for other things to start breaking under heavy load.
It's also funny and sad that ZOS' workaround for "we won't run events that require Groupfinder" has just been to rewrite the dungeon events so they don't "require" Groupfinder. With the result that the only canceled event was Midyear Mayhem because of Battlegrounds.
What's just sad is that everything else seems to have broken for more or less the same cause as earlier this year - that PC/EU has a high number of concurrent players during events. Whatever improvements ZOS made to handle that earlier don't seem to have held up for very long.
Yes. I do realize this is a mega sever and it makes the servers for games like WoW/FF/SWTOR look rather puny. However, after more than 5.5 years I would expect Zos getting a better handle on performance vs the degradation we have seen over the years.
Fortunately Zos is attempting some potentially significant improvements over the next few updates. However, I do not have a high degree of confidence in the changes given their track record. I am still dumbfounded how a team can create such an amazing game with amazing systems yet cannot figure out how to manage the environment or combat.1 -
SirAndy✭✭✭✭✭
✭✭✭✭✭
Actually, on top of that list should be "Hello World" using the Brain[Snip] programming language.According to an industry website on coding challenges, here are the top five hardest things known to man to code:
1. Aanderaa-Karp-Rosenberg evasiveness conjecture
2. Cryptocurrency algorithm replication
3. Space vehicle intercept trajectories
4. Integer factorization in polynomial time
5. Video game player grouping
So cut ZOS some slack, they've only been working on it for 5 and a half years!
[Edit to remove profanity]
Sigh, that is the name of an actual programming language ...
https://en.wikipedia.org/wiki/Brainfuck
Edited by SirAndy on October 28, 2019 12:00AM7 -
nafensoriel✭✭✭✭✭Space vehicle intercept trajectories can easily surge to petabytes of data to be calculated if you include annoying things like accurate gravity models.
Group finders are just one of those absurd logic equations. To anyone who thinks group finders are easy to do.. write a flowchart for how they should work. When you get to your 40th page and still havent reached an optimal solution youll understand why it's hard. If you never reach the 40th step congradulations youve conclusively proven you cant code and shouldn't code.0 -
SirAndy✭✭✭✭✭
✭✭✭✭✭nafensoriel wrote: »Space vehicle intercept trajectories can easily surge to petabytes of data to be calculated if you include annoying things like accurate gravity models.
Group finders are just one of those absurd logic equations. To anyone who thinks group finders are easy to do.. write a flowchart for how they should work. When you get to your 40th page and still havent reached an optimal solution youll understand why it's hard. If you never reach the 40th step congradulations youve conclusively proven you cant code and shouldn't code.
That is so last century ...
- Establish a decent amount of ground truth (Shouldn't be any problem for ZOS since they should have plenty of historical data)
- Train a neural network (AWS has cheap GPU instances just for this purpose, they even have them wrapped around a dedicated service now)
- Use the NN output model and implement it in your server app to make quick and accurate predictions about who should be grouped
This isn't rocket science, i've done just that and used NN models on anything from large high traffic client/server systems down to embedded ARM devices (Even some without floating point FPU)
0 -
nafensoriel✭✭✭✭✭nafensoriel wrote: »Space vehicle intercept trajectories can easily surge to petabytes of data to be calculated if you include annoying things like accurate gravity models.
Group finders are just one of those absurd logic equations. To anyone who thinks group finders are easy to do.. write a flowchart for how they should work. When you get to your 40th page and still havent reached an optimal solution youll understand why it's hard. If you never reach the 40th step congradulations youve conclusively proven you cant code and shouldn't code.
That is so last century ...
- Establish a decent amount of ground truth (Shouldn't be any problem for ZOS since they should have plenty of historical data)
- Train a neural network (AWS has cheap GPU instances just for this purpose, they even have them wrapped around a dedicated service now)
- Use the NN output model and implement it in your server app to make quick and accurate predictions about who should be grouped
This isn't rocket science, i've done just that and used NN models on anything from large high traffic client/server systems down to embedded ARM devices (Even some without floating point FPU)
So your solution is just "make a neural network and bam it works" ?
Are you daft?
Sorry, that's baiting.
Yes, you can make machine learning do this. It most likely would also require considerably more work. If you find it easy.. go make a group finder with an NN. Youll be a millionaire overnight selling the code to EA, UBI, and EPIC.
I do love when new blood comes along and pretends they've found the holy grail to replace all the old farts.. and then in every single case fall on their face when they hit reality. NNs are great. They are awesome tools. They are not magic and they don't replace all the tedious BS. Some yes. Never all.1 -
Luckylancer✭✭✭✭✭nafensoriel wrote: »Space
Group finders are just one of those absurd logic equations.
You select longest waiting 2 DDs, 1 healer and 1 tank. I cant imagine how hard it is!
If you dont want to put 810cp and 15 level guy, group people to "810-300 cp, 300-160cp, 160cp-10 level. If you have a lot of cp but sub-50, you are down by 1 group." parties can be formed from 2 adjent group types max, waiting 5 min means you can be in party with 2 upper or lower groups aswell.
This cant be rocket science.0 -
nafensoriel✭✭✭✭✭Luckylancer wrote: »nafensoriel wrote: »Space
Group finders are just one of those absurd logic equations.
You select longest waiting 2 DDs, 1 healer and 1 tank. I cant imagine how hard it is!
If you dont want to put 810cp and 15 level guy, group people to "810-300 cp, 300-160cp, 160cp-10 level. If you have a lot of cp but sub-50, you are down by 1 group." parties can be formed from 2 adjent group types max, waiting 5 min means you can be in party with 2 upper or lower groups aswell.
This cant be rocket science.
The first group finders were that simple... Then players complained that they were unfair or made bad group comps which resulted in poor user experiences.
Then we broke up the system to have multiple queues instead of just one. This meant we had situations where 5000000 DPS got cockblocked by there only being 3 healers. It resulted in "fake healers" which we can all understand causes issues since we currently face it with the dungeon queues for many players.
So how do you decide the logic queues to accept when to just slot DPS or more reality-based... when to adjust the size of healer slots so the teams remain functionally balanced. How do you choose which healer to go where? A blind slot results in bad user experience. Adding premades to this mix makes it worse.
Say you've decided that logic.. now you have to decide if you want to improve user experience by putting newbies with newbies and pros with pros... but you don't want to cause a situation where no BGs are formed because you don't have enough to populate either group.
Anyone can make a group finder. Making one is simple. Making one people enjoy and like isn't. The farther down the rabbit hole you go the more complex it gets. The more complex it gets the more resources it needs and the more fragile it becomes to sudden surges of players. Remember if you overload the CPU calculating all this your system crashes. So how do you make it lean, fair, fun, and efficient?
As I said if you have an idea on how to do this that is actually novel you will automatically get a job worth millions. That is how valuable it is to multiplayer games. Many have tried. Most have failed. Few have gotten lucky to get that right sauce for THEM and THEIR players.0 -
SirAndy✭✭✭✭✭
✭✭✭✭✭nafensoriel wrote: »nafensoriel wrote: »Space vehicle intercept trajectories can easily surge to petabytes of data to be calculated if you include annoying things like accurate gravity models.
Group finders are just one of those absurd logic equations. To anyone who thinks group finders are easy to do.. write a flowchart for how they should work. When you get to your 40th page and still havent reached an optimal solution youll understand why it's hard. If you never reach the 40th step congradulations youve conclusively proven you cant code and shouldn't code.
That is so last century ...
- Establish a decent amount of ground truth (Shouldn't be any problem for ZOS since they should have plenty of historical data)
- Train a neural network (AWS has cheap GPU instances just for this purpose, they even have them wrapped around a dedicated service now)
- Use the NN output model and implement it in your server app to make quick and accurate predictions about who should be grouped
This isn't rocket science, i've done just that and used NN models on anything from large high traffic client/server systems down to embedded ARM devices (Even some without floating point FPU)
So your solution is just "make a neural network and bam it works" ?
Are you daft?
Sorry, that's baiting.
Yes, you can make machine learning do this. It most likely would also require considerably more work. If you find it easy.. go make a group finder with an NN. Youll be a millionaire overnight selling the code to EA, UBI, and EPIC.
I do love when new blood comes along and pretends they've found the holy grail to replace all the old farts.. and then in every single case fall on their face when they hit reality. NNs are great. They are awesome tools. They are not magic and they don't replace all the tedious BS. Some yes. Never all.
Yes, i'm so daft that i do this kind of stuff on a daily basis for a living and have yet to fail to deliver on a project.
As for "new blood", i've been programming since 1978, with many years spent in the gaming industry.
Here's one thing i have learned in all those years, if you don't keep up with current technology, you will become obsolete very quickly.
PS: And yes, setting up a NN and training a model is not difficult at all anymore. With decent ground truth data, they will almost always outperform conventional algorithm based solutions, especially for complex problems like a smart group finder.
Edited by SirAndy on October 28, 2019 4:08PM0 -
nafensoriel✭✭✭✭✭nafensoriel wrote: »nafensoriel wrote: »Space vehicle intercept trajectories can easily surge to petabytes of data to be calculated if you include annoying things like accurate gravity models.
Group finders are just one of those absurd logic equations. To anyone who thinks group finders are easy to do.. write a flowchart for how they should work. When you get to your 40th page and still havent reached an optimal solution youll understand why it's hard. If you never reach the 40th step congradulations youve conclusively proven you cant code and shouldn't code.
That is so last century ...
- Establish a decent amount of ground truth (Shouldn't be any problem for ZOS since they should have plenty of historical data)
- Train a neural network (AWS has cheap GPU instances just for this purpose, they even have them wrapped around a dedicated service now)
- Use the NN output model and implement it in your server app to make quick and accurate predictions about who should be grouped
This isn't rocket science, i've done just that and used NN models on anything from large high traffic client/server systems down to embedded ARM devices (Even some without floating point FPU)
So your solution is just "make a neural network and bam it works" ?
Are you daft?
Sorry, that's baiting.
Yes, you can make machine learning do this. It most likely would also require considerably more work. If you find it easy.. go make a group finder with an NN. Youll be a millionaire overnight selling the code to EA, UBI, and EPIC.
I do love when new blood comes along and pretends they've found the holy grail to replace all the old farts.. and then in every single case fall on their face when they hit reality. NNs are great. They are awesome tools. They are not magic and they don't replace all the tedious BS. Some yes. Never all.
Yes, i'm so daft that i do this kind of stuff on a daily basis for a living and have yet to fail to deliver on a project.
As for "new blood", i've been programming since 1978, with many years spent in the gaming industry.
Here's one thing i have learned in all those years, if you don't keep up with current technology, you will become obsolete very quickly.
PS: And yes, setting up a NN and training a model is not difficult at all anymore. With decent ground truth data, they will almost always outperform conventional algorithm based solutions, especially for complex problems like a smart group finder.
The first instrument I had to write code for was mechanical and the only silicon in it was the glass. I understand evolving technology. Now that we've gotten the ***-swinging out of the way I agreed it can help. It doesn't solve the entire problem though. Group finders also have to be lean. NN doesn't define that. They help shave off steps 1-52 of a 400 step process. It's nice. It's not revolutionary.0 -
TequilaFire✭✭✭✭✭
✭✭✭✭✭This should handle the load.
Well it has glass in it. lmao
Edited by TequilaFire on October 28, 2019 4:27PM1 -
FrancisCrawford✭✭✭✭✭
✭✭✭✭I heard that the two hardest problems in coding were naming, caching and off-by-1.0 -
Luckylancer✭✭✭✭✭@nafensoriel
No one gonna have riggt to complain when it is "longwst waiting one will play first". There wont be bad comps, 1T 2DD 1H is the standart.
If there are 50000 DDs and 3 bealers, it is already game over. Nothing can fix that other than people que as healers. Fake healers will be kicked by other party members. This group will have highest priority to find new healer if they are grouped by dungeon finder in firstplace.
Healers will go to those who waited most. Adding premades wont cause any problems. There are rabbit holes for such easy dungeon finder. This is not a MOBA. Creating a functional dungeon que for this game supposed to be easy.
I wont talk about BG que. It is complicated.0 -
Donny_Vito✭✭✭✭✭
✭
Actually, on top of that list should be "Hello World" using the Brain[Snip] programming language.According to an industry website on coding challenges, here are the top five hardest things known to man to code:
1. Aanderaa-Karp-Rosenberg evasiveness conjecture
2. Cryptocurrency algorithm replication
3. Space vehicle intercept trajectories
4. Integer factorization in polynomial time
5. Video game player grouping
So cut ZOS some slack, they've only been working on it for 5 and a half years!
[Edit to remove profanity]
Sigh, that is the name of an actual programming language ...
https://en.wikipedia.org/wiki/Brainfuck
Haha! We were talking about this at work the other day. What an interesting language name1 -
SirAndy✭✭✭✭✭
✭✭✭✭✭nafensoriel wrote: »Group finders also have to be lean. NN doesn't define that.
Au contraire. The training portion is complex and time consuming, the resulting model can be lean and quick. As i stated before, i have successfully included NN models in real-time(!) data processing on 32-bit ARM processors that do not even have hardware floating point support. Think automotive head units.
That's about as "lean" as you can get these days.
If you have a 2016 or newer car from one of the major car companies, chances are you have been treated to data processing using my code.
Once a model is loaded and initialized, calling the predict function on average only adds a few milliseconds to the processing time.
Edited by SirAndy on October 28, 2019 4:48PM1 -
max_only✭✭✭✭✭
✭✭✭✭✭I don’t know about this thread. It starts off as a joke. Then people reply seriously. Then someone adds to the joke. And now there’s a discussion of who can call themselves a real coder. Now people are offering solutions... listen: it’s fake.
I despair for gamer kind.
#FiteForYourRite Bosmer = Stealth
#OppositeResourceSiphoningAttacks
|| CP 1000+ || PC/NA || GUILDS: LWH; IA; CH; XA
""All gods' creatures (you lot) are equal when covered in A1 sauce"" -- Old Bosmeri Wisdom1 -
Girl_Number8✭✭✭✭✭
✭✭✭According to an industry website on coding challenges, here are the top five hardest things known to man to code:
1. Aanderaa-Karp-Rosenberg evasiveness conjecture
2. Cryptocurrency algorithm replication
3. Space vehicle intercept trajectories
4. Integer factorization in polynomial time
5. Video game player grouping
So cut ZOS some slack, they've only been working on it for 5 and a half years!
When they respect their customers and have a game that can be played.
Did Zos give that courtesy to all the players that have put large amounts of money and time into the game....No they did not.
1 -
Bergzorn✭✭✭✭✭
✭I don’t know about this thread. It starts off as a joke. Then people reply seriously. Then someone adds to the joke. And now there’s a discussion of who can call themselves a real coder. Now people are offering solutions... listen: it’s fake.
I despair for gamer kind.
There are 10 kind of people when it comes to jokes about programming, the ones that get it and the ones that don't.no CP PvP PC/EU
EP Zergborn
DC Zerg Beacon
guild master, raid leader, janitor, and only member of Zergbored1
