Datasets:

breadlicker45

/

discord

like 1

ResidentMadScientist#3415

03/26/2019 9:13 PM

Will move rn

chip#7643

03/26/2019 9:13 PM

no problem

Simp#8258

03/27/2019 9:40 AM

make c# bots run on .net core so it could run on linux and mac

Tangil#3830

03/27/2019 9:56 AM

this is not possible because flatbuffers runs only on .net framework

Simp#8258

03/27/2019 9:59 AM

i hope theres flat buffa for net core

Simp#8258

03/27/2019 9:59 AM

👀

Simp#8258

03/27/2019 9:59 AM

or a lib

ccman32#4022

03/27/2019 10:04 AM

oh god, pls no

ccman32#4022

03/27/2019 10:04 AM

or maybe one day when .net core becomes actually usable 😂

Martmists#3740

03/28/2019 5:04 AM

Allow bots to have a getMatchSettings method, that way they can easily detect e.g. Infinite boost, Ball speed, etc modifiers without having to do so from in-game data.

Jonas#2222

03/28/2019 5:06 AM

I think the Psyonix API will provide a lot of cool data. Shouldn't we wait and see after that?

tarehart#6423

03/28/2019 11:29 AM

You already have that available

tarehart#6423

03/28/2019 11:29 AM

Let me dig up the link

tarehart#6423

03/28/2019 11:31 AM

<https://github.com/RLBot/RLBot/pull/341>

tarehart#6423

03/28/2019 11:34 AM

tldr: implement init_match_config in your agent class

Kipje13#9878

03/28/2019 12:15 PM

something that works for all languages would be nice

dtracers#9598

03/28/2019 12:16 PM

every language has a base agent that has that in python

dtracers#9598

03/28/2019 12:17 PM

so you just need to transfer it to your language

dtracers#9598

03/28/2019 12:17 PM

same way you transfer the index and name of your bot

Kipje13#9878

03/28/2019 12:25 PM

this seems like something that the framework needs to supply rather than expecting each language to implement it individually

dtracers#9598

03/28/2019 12:26 PM

we should move this to #framework-dev

whatisaphone#8521

03/29/2019 12:02 AM

i suggest the ❕ vods command shouldn't be so long and sprawling. 2017 bots are not impressive or relevant anymore (sorry!) it should show the most recent 1v1 youtube link, and then some sort of "Go here for more" link to <rlbot.org> where all the history is listed

👍 (4)

whatisaphone#8521

03/29/2019 12:03 AM

something like

whatisaphone#8521

03/29/2019 12:04 AM

Wintertide 1v1 (March 2019) – http://www.rlbot.org/tournament/ See the entire tournament history at <http://www.rlbot.org/tournament/>

whatisaphone#8521

03/29/2019 12:04 AM

nice and short, and doesn't push the entire conversation off the screen

Skyborg#0894

03/29/2019 12:06 AM

Pinned a message.

Jonas#2222

03/31/2019 5:17 PM

Update the item ID's on the wiki page. Many items aren't even listed there.

👍 (3),pingsock (1)

Edwin.#7531

04/01/2019 9:32 AM

Okay so there's a reason why it hasn't been updated yet: The way we had grabbed the items up until now doesn't work anymore. When we tried to get the items we did get a list with all of them, but the ID's ain't right, so we tried but failed

Marivahlio#3599

04/05/2019 7:01 AM

Custom banners! It'd be awesome if we could change the banners into our own sprites. Alphaconsole already has this feature but it's subscriber only.

Marivahlio#3599

04/05/2019 11:38 AM

Something like this.

dtracers#9598

04/05/2019 11:40 AM

I agree that would be awesome. Did see you how we did logos in the last tournament? @Marivahlio

Marivahlio#3599

04/05/2019 11:53 AM

No i did not. I rejoined today :)

dtracers#9598

04/05/2019 12:08 PM

Gotcha. You might want to watch it! But I think from now on we will have logos.

MeoW#5887

04/08/2019 8:40 AM

Why isn't there a port of the framework in c++? It would be beneficial for performance reasons at least

dtracers#9598

04/08/2019 8:44 AM

No one has made it

DomNomNom#1537

04/08/2019 8:44 AM

No one has put the effort in to do it yet. Whoever does it kinda carries the maintenance burden for it. I'd recommend Rust if you're after compiled performance. Most bots are not that computationally demanding.

dtracers#9598

04/08/2019 8:44 AM

You can make it if you want :)

MeoW#5887

04/08/2019 8:48 AM

OK It just got me thinking because even scratch has one. C++ would enable using pytorch for examole. I don't think rust has support for any machine learning libraries. Correct me if I'm wrong.

Hytak#5125

04/08/2019 8:49 AM

There are pytorch bindings for rust

MeoW#5887

04/08/2019 8:50 AM

@dtracers I hoped that I could contribute in it's development. But starting it from ground up would be too much for me 😅

MeoW#5887

04/08/2019 8:50 AM

@Hytak didn't know that. Thanks

dtracers#9598

04/08/2019 8:53 AM

@MeoW yes that is what most people have said to making the c++ version. 😂 😋

Jonas#2222

04/08/2019 8:57 AM

Let's just call a c++ expert and force him to do it

MeoW#5887

04/08/2019 9:23 AM

Totally agree with you @Jonas

Simp#8258

04/08/2019 9:27 AM

i mean theoretically you can make c++ bot already its not easy tho it will never be + ml has libs for alot of languages we support

Simp#8258

04/08/2019 9:27 AM

i think only scratch doesnt but you could add it

Simp#8258

04/08/2019 9:28 AM

scratch we use is a port

Simp#8258

04/08/2019 9:28 AM

so you can add custom stuff

Jonas#2222

04/08/2019 10:18 AM

If I ever find out how, then I would love to do that

michael#0046

04/08/2019 10:25 AM

Linux support?

DomNomNom#1537

04/08/2019 10:42 AM

Yup, long standing request <https://github.com/RLBot/RLBot/issues/277>

Simp#8258

04/08/2019 10:45 AM

its comming

Simp#8258

04/08/2019 10:46 AM

patience tarehart is sweating blood

DomNomNom#1537

04/08/2019 11:08 AM

I'd like to help

whatisaphone#8521

04/08/2019 3:01 PM

if that's true, tarehart should see a doctor 😕

😓 (2)

Martmists#3740

04/09/2019 3:06 AM

As Rocket League uses Quaternions internally for angles, is it possible to provide these through the game tick packet instead? They're easier to use than Euler angles in Vector math too.

Tangil#3830

04/09/2019 5:21 AM

use the rigidbodytick for quaternions for rotations

Martmists#3740

04/09/2019 5:23 AM

last time I tried to get RBT I kept getting errors in console though :/

twobackfromtheend#1471

04/09/2019 5:25 AM

don't ask for the rbtick when the game is not running lol

twobackfromtheend#1471

04/09/2019 5:26 AM

I think checking is_round_active is good enough

twobackfromtheend#1471

04/09/2019 5:26 AM

prevented the errors I got from trying to ask for the rbtick after a training pack shot ends but before the replay starts

Marivahlio#3599

04/09/2019 5:55 AM

Why are quaternions easier than eulerangles? Seems like the opposite for me

twobackfromtheend#1471

04/09/2019 6:07 AM

I think they're entirely interchangeable and equivalent

twobackfromtheend#1471

04/09/2019 6:07 AM

it's like multiplying by 0.1 or dividing by 10.

DomNomNom#1537

04/09/2019 10:21 AM

While I don't have a strong opinion whether we should change things, I'd say Quaternions are easier than eulerangles: 1) quaternions are widely used in computer graphics 2) Euler angles always require fiddly details to be specified which are often undocumented because people assume it's as obvious as lattitude/longitude: What is the default rotation? (How does pitch/yaw/roll map onto X/Y/Z? ) Degrees/radians? Which order are the pitch/yaw/roll rotations applied in? (I'd put emphasis on this last part) 3) Euler angles can make it easy to write mostly-working code that is wrong in general. e.g.: "I want to detect if my car is mostly upside down, I'll just check roll." only works if your pitch doesn't mess you up, but even then, it requires careful thinking whether that assumption is valid in your order of rotations. (note: I don't think the gimbal-lock argument applies here as the rotations are not chained) While twoback is technically correct that they both express the same information but I find quaternions are more standard to work with. Just because we can express speed in unreal units per fortnight doesn't mean we should

⬆ (4)

Marivahlio#3599

04/09/2019 11:17 AM

Oh boy thats some solid arguments, well played.

◱ Mark#0001

04/09/2019 2:53 PM

Damn, someone has a strong argument there

◱ Mark#0001

04/09/2019 2:53 PM

I would totally agree but it might break a lot of agents (Especially ML agents as they might need to be retrained)

Tangil#3830

04/09/2019 2:53 PM

btw we can't change out euler angles for quats, we can only add in quats as an additional field

◱ Mark#0001

04/09/2019 2:54 PM

Well, that would be a nice solution

whatisaphone#8521

04/09/2019 4:11 PM

turn on slow mode in #bot-clips

⬆ (1)

michael#0046

04/09/2019 6:31 PM

A rocket league gym environment?

👌 (1)

Skyborg#0894

04/09/2019 6:32 PM

@michael like RLBotTraining?

michael#0046

04/09/2019 6:37 PM

No like an open ai gym environment

r0bbi3#0269

04/09/2019 6:39 PM

Somebody did that, I believe, try searching for it on this discord

ThatOtherNinja#6315

04/10/2019 12:44 AM

Last time I saw a discussion about it I think someone said that it wasn't possible to actually use the gym structure directly (I don't remember why), however I'm sure you could make something that functions a lot like it.

ThatOtherNinja#6315

04/10/2019 12:48 AM

A quick search and I found this https://discordapp.com/channels/348658686962696195/348658686962696196/553608597112619008

ThatOtherNinja#6315

04/10/2019 1:10 AM

Maybe add a question to the faq about what the relevant math topics are and good resources for learning them

Eastvillage#2628

04/11/2019 12:27 PM

Rename #human-rl to something like #human-rocket-league or #rocket-league-play so it doesn't suggest that it is only for finding team mates

👌 (9)

Jonas#2222

04/11/2019 12:43 PM

That's a good idea. Something that people can recognize as the place to talk about that stuff

whatisaphone#8521

04/11/2019 4:09 PM

#meat-emporium

dtracers#9598

04/11/2019 4:10 PM

you know there is a channel where you can make that a reality

whatisaphone#8521

04/11/2019 4:11 PM

it's pretty lonely there though

whatisaphone#8521

04/11/2019 4:12 PM

in fact it's basically a solo queue

whatisaphone#8521

04/11/2019 4:12 PM

i'm going to !redox myself

RLBot#1485

04/11/2019 4:12 PM

Channel police, anti-spam agent

Skyborg#0894

04/12/2019 5:16 PM

adding a "resolution" option to polyline so that it skips a percentage of the vectors

chip#7643

04/12/2019 7:02 PM

Just give it fewer segments in the first place, imo

chip#7643

04/12/2019 7:02 PM

Better for performance that way too, generating/sending less info

Skyborg#0894

04/12/2019 7:18 PM

@chip but its easier to shove the whole prediction vector packet there and select a resolution XD

dtracers#9598

04/12/2019 7:21 PM

yeah i would say that is a you problem not a framework problem 😃

😂 (1)

chip#7643

04/12/2019 7:23 PM

ditch the framework prediction, at least until it gets an update

chip#7643

04/12/2019 7:27 PM

or, just do

chip#7643

04/12/2019 7:28 PM

``` a = [i for i in range(100)] print(a[1::5]) # [1, 6, 11, 16, 21, 26, 31, 36, 41, 46, 51, 56, 61, 66, 71, 76, 81, 86, 91, 96] ```

Skyborg#0894

04/12/2019 7:28 PM

i was actually using the RLU prediction

twobackfromtheend#1471

04/12/2019 7:32 PM

range takes a step parameter I think

r0bbi3#0269

04/12/2019 7:34 PM

``` a = [i for i in range(1, 100, 5)] print(a) ```

chip#7643

04/12/2019 7:37 PM

you're missing the point

chip#7643

04/12/2019 7:37 PM

he already has the array, he's not using range to generate it

chip#7643

04/12/2019 7:38 PM

he just (probably) didn't know about the step part of the slice operator