second-based records

Post content relevant to Guild Wars. This includes, but is not limited to: screenshots, guides, tutorials, questions & answers, etc.
Flo
Moderator
Moderator
Posts: 1366
Joined: Tue Dec 22, 2015 4:23 pm
In-game name: Hi Im Undercover
Guild: Golt

Re: second-based records

Post by Flo »

Currently checking some options how to properly implement a second based record system based on ingame time. Will hopefully have some news in a couple of days. If you re bored in the meantime, check the gwscr channel on the SC discord for some juicy drama.

haskhasin
Moderator
Moderator
Posts: 208
Joined: Thu May 31, 2012 6:03 pm

Re: second-based records

Post by haskhasin »

some thoughts and practical problems:

I am overall supportive of second-based records
Minute restrictions made sense in the past, and still do for many areas and party sizes, but I agree that many (most?) other areas are unbeatable in the current system.

It will still suck that players with money (bots) and time (grind) can beat more skilled players, but I've accepted that.

Tables and tech
Ideally we would have a custom form to submit records that will go into a mod-queue to be approved, and then into an automatically sorted table of records. The custom form would allow people to enter both time, video, category, etc. The problem is that we don't have any of that, and it would require development time in php. This is not my expertise and I don't like it. If someone wants to volunteer to work on this, either in this website or on your own, please do. I'm maintaining (and paying for) this website for the community and because no-one else stepped in to do it, not for the power or ad-money (lol).

Since we don't have the above technology, the moderator has to manually add an entry to the table, which currently looks like this:

Code: Select all

[tr]
[td][TAS][/td]
[td]8 players[/td]
[td][-][/td]
[td](14min)[/td]
[td][url=http://gwscr.fbgmguild.com/viewtopic.php?f=31&t=2518]Thread[/url][/td]
[/tr]
This is to say (other than thank you Flo), that adding a new entry somewhere in the table is annoying but doable, manually maintaining a rank list is a nope.

in my opinion, the table could look exactly like the current one: one list per area sorted by time, with columns for party size and category (tas/pure/whatever). Just add the seconds to the time, and keep more entries. The question now becomes, when are runs removed from the list? Not sure.

Timing runs
We don't want to require records to show custom software like toolbox. Probably the best solution is to rely on video time (optionally reference tb timer for single-level areas, or add timer overlay afterwards). However we can and should add a compatible timer to toolbox (starts on 1st map load at 00:00, only resets in outposts).

As for multi-level areas, I think the best solution is to just use real time, avoiding map loads is just too much of a pain, and loads are relatively quick on gw anyway. If you don't have an SSD buy one.

Trust
Overall, changing to seconds will require more trust on the community. Any kind of tb-related security measure would be annoying, hard to enforce, and easy to bypass. Time comes from video, and as much as the mods can check, I'm sure people are able to cheat, especially in some of the most tight and sought-after records, like in Fronis Irontoe's Lair.

We'll just have to trust the community and hope for the best. Do I trust this community to not cheat? I'm honestly not sure. I definitely don't trust the community with regards to content, but we haven't had cheaty submissions in a long time. A few submissions that didn't adhere by the rules were generally not malicious. I believe we can try.

More mods?
There are a lots of requests for more mods. Let's be honest here, the GW community is extremely toxic, and I trust very few of you. The last time we asked for more mods we had very few submissions and at the end no-one but for Flo did anything anyway. That said, I'm not against adding more mods, and it would actually be great to have more experienced eyes for uw and dungeon records, especially as second-based records might increase the rate of submissions. I'll try to setup mod applications.

Flo
Moderator
Moderator
Posts: 1366
Joined: Tue Dec 22, 2015 4:23 pm
In-game name: Hi Im Undercover
Guild: Golt

Re: second-based records

Post by Flo »

I ll somewhat follow Has's "template".

I am generally open to second based records

Pro is of course that it might facilitate some competition in otherwise dead categories (even though you could of course also argue against this).

But I do not think second based records should be a complete replacement to minute based records, because the goal that is tried to be achieved in those different settings is pretty different (not superior, not inferior, just different). This becomes most apparent in dungeon records: So far people have been optimising for minute thresholds on certain levels while sacrificing time on others, because at the moment a 0:59+0:59+1:59 is considered superior to a 0:40+0:40+2:10. For single level area this is also relevant, because so far the goal was to beat a minute threshold while at the same time trying to achieve this with the tactics that have the lowest inherent variance.

I think both of the underlying optimization problems between minute and second based records are really interesting, and they can not simple be compared. Therefore my suggestion would be to not have a replacement, but just keeping track of both at the same time in separate tables.

Timing runs

There are 2 approaches for tracking second based records:

1. Tracking real time from start of the run to finish of the run
PRO: A little bit more precision
PRO: Does not require people to inject a 3rd party tool
CONTRA: This requires video coverage from the first person that loads into a level and the last person to leave the level for every level. This will most likely mean in practice, that EVERY SINGLE RECORD, EVEN TAS ONES WILL REQUIRE EVERY PARTY MEMBER TO RECORD, because there is variance to load times so you can not be sure who will load first in the actual record run. So if you lose that single video of the fastest loader, you can not post this record as second based AT ALL. Could probably find a solution for this like just adding +2 seconds or something by default if the video coverage provided isnt suffiecient, but then again we re not better off that the other option.
CONTRA: Human interpretation is required which leaves room for debate. I think its a reasonable goal to minimize the ambiguity. And yes, I think this even stands if you go frame by frame in a video (FPS, which frame is really the first and last frame of a run...).

2. Tracking instance time
PRO: Videos are not necessary for timing purposes only, TAS records are still allowed to only have 1 video
PRO: There is no ambiguity - instance end time is instance end time and no human interpretation is required
CONTRA: There are minor imprecisions (will probably average out to <=1 second per level, but similar for every record so the delta should be low) with tracking load times at the start of the run. The instance is created during the loading screen so the timer starts already. Some people have reported high load times when running multiple 3rd party tools.
CONTRA: Getting the instance time requires at least 1 person of a team to inject a 3rd party tool.

Both methods of meassuring can account for multi level areas, so I do not think that makes one of them better in comparison. Maybe a slight edge to instance time because there is less effort.

I personally think using instance time is way easier and there will be much less uncertainty and therefore debate with the small downside of losing a slight amount of precision. I think this is mostly an operative issue and should not have a noticeable effect on the actually achieved times (or their ranking), so not sure whether community feedback is required for this aspect.

For the implementation there would of course be quite a few questions, mostly whether this table would start from scratch or whether old records (that have video coverage) should also be migrated onto the second based table.

Tech regarding instance time

I think technical solutions to the instance timer problem are fairly easy to come by with, as it is already pretty much implemented in TB for dungeons and would only need to be extended to save the instance time on every map load and keep track of it for like the last 10 areas. Maybe this could be put into an independent tool outside of toolbox, so people have less concerns using it in a PURE record for example. But I already talked to Jon and he is generally open to incorporating a feature like this into TB, so I think that would probably be the easiest way for a lot of people.

Tech regarding tables

I think what we re doing with like 50 areas you can run in with 8 different party sizes and 2-3 modes would not work well in the type of layout speedrun.com uses. I think the tables are fine at the moment but I honestly do not have a real opinion either way. Before implementing it I think there should be some sort of UX test though, otherwise people might put a lot of effort into a inpractical solution.

Trust

Trust is a great thing, but that does not mean we should be naive. There is a reason we currently require full video coverage for PURE and standard records. So my suggestion would be to incorporate some kind of Hash validation values into toolbox that can be printed out alongside the instance time timestamps. This can of course be manipulated as well (open source and so on), but I think there is absolutely no downside and if it discourages malicious behaviour even a slight bit it's worth it.

More mods?

Regarding the mod application stuff I definitely agree that the situation has not been ideal. It should be people that a) have a spine, b) will commit to this for an extended amount of time (>1 year) and c) that do not already have a lot if influence in the community*.

*Ideally we would have moderators that do not do records themselves and that are not affiliated with any runners or guilds to ensure there is no bias whatsoever and at the same time have at least a decent understanding of tactics for most areas. Unfortunately that is impossible and of course there has to be a compromise, but I think it would be a bad idea to have guild leaders / discord admins / reddit admins / people that prodvide tools for records also as moderators here. Separation of power in general is a good idea to avoid individuals excerting too much influence over a community as a whole and not just one aspect of it.

--

I think this captures quite a bit of what people have been asking for. Please keep in mind that there are 20 people in this thread with 30 different opinions, each of them claiming their solution is the one true way (with a few of them even straight up contradicting others). Some compromise will have to be made.

Misty
Posts: 560
Joined: Fri Feb 24, 2012 7:19 am
In-game name: ____ of the Owls
Guild: Illumination Theory [IT]

Re: second-based records

Post by Misty »

Flo wrote:
Mon Nov 23, 2020 7:06 am
I think technical solutions to the instance timer problem are fairly easy to come by with, as it is already pretty much implemented in TB for dungeons and would only need to be extended to save the instance time on every map load and keep track of it for like the last 10 areas. Maybe this could be put into an independent tool outside of toolbox, so people have less concerns using it in a PURE record for example. But I already talked to Jon and he is generally open to incorporating a feature like this into TB, so I think that would probably be the easiest way for a lot of people.
Plenty of people are completely opposed to using third party timers, so unless they change their minds or they party with someone else, they would be excluded from records that require that. It also invalidates all past records, even if they provided ample video coverage, simply because the required timer didn't exist. This is not a small downside to be dismissed so quickly.
Flo wrote:
Mon Nov 23, 2020 7:06 am
More mods?

Regarding the mod application stuff I definitely agree that the situation has not been ideal. It should be people that a) have a spine, b) will commit to this for an extended amount of time (>1 year) and c) that do not already have a lot if influence in the community*.

*Ideally we would have moderators that do not do records themselves and that are not affiliated with any runners or guilds to ensure there is no bias whatsoever and at the same time have at least a decent understanding of tactics for most areas. Unfortunately that is impossible and of course there has to be a compromise, but I think it would be a bad idea to have guild leaders / discord admins / reddit admins / people that prodvide tools for records also as moderators here. Separation of power in general is a good idea to avoid individuals excerting too much influence over a community as a whole and not just one aspect of it.
I think your concerns become less important as more mods are added. If we have dozens of mods, none would have too much responsibility, and each can keep a check on the rest.
Flo wrote:
Wed Apr 20, 2016 10:56 am
Sadly, Misty is right.

Cruz
Posts: 209
Joined: Tue May 14, 2013 2:29 am
In-game name: Ashei Cruz
Guild: IT

Re: second-based records

Post by Cruz »

Flo wrote:
Mon Nov 23, 2020 7:06 am
Things
Personally I have no issue with any of the two ways of timing runs for RTA, but I think the implications for one of them is a lot more significant.
Basically the choice we have to make is: Do we want to require video recordings from every player also in TAS (no change for other categories), or do we want to force teams to run at least one third party tool on one player (even pure runs). To me it's obvious that timing from video recordings is the superior option, just based on that. I understand that you have a different opinion, but I think you put way too much weight on the risk of idiots whining because they can't record. It's not hard to record with OBS.
Flo wrote:
Mon Nov 23, 2020 7:06 am
CONTRA: Human interpretation is required which leaves room for debate. I think its a reasonable goal to minimize the ambiguity. And yes, I think this even stands if you go frame by frame in a video (FPS, which frame is really the first and last frame of a run...).
Determining which frame to start/end is just a technicality, not really an issue. Like I've suggested before, you can use the first frame with visible UI and the first frame with visible completion. (Gate open message/reputation gain/quest update, these are available for every area right?)
You already require the video to be watchable right? So the fps of the video would introduce way less variance than letting an instance timer tick while you load, if I'm not mistaken.
Do you think it likely that a time gained with this method would be disputed? Worst case scenario, get a second opinion from another mod.
Flo wrote:
Mon Nov 23, 2020 7:06 am
Both methods of meassuring can account for multi level areas, so I do not think that makes one of them better in comparison. Maybe a slight edge to instance time because there is less effort.
I don't really have a problem with including intermediate loading screens, but if mods are prepared to time each level I welcome that. The extra work is not a lot if you know how to do it properly.
Flo wrote:
Mon Nov 23, 2020 7:06 am
For the implementation there would of course be quite a few questions, mostly whether this table would start from scratch or whether old records (that have video coverage) should also be migrated onto the second based table.
I think the best way to go about this is to ask runners to request timing their past runs for RTA. This way mods don't get swamped timing runs few or no one cares about. Even so, more old runs could be timed if they are deemed significant.

Flo
Moderator
Moderator
Posts: 1366
Joined: Tue Dec 22, 2015 4:23 pm
In-game name: Hi Im Undercover
Guild: Golt

Re: second-based records

Post by Flo »

I have not met any of those people that refuse 3rd party tools completely in a speedrunning context (which might just be due to my filter-bubble), nor have they spoken up here. Also I would guess there is a huge intersection between people who will not record a video and people who do not want any of their teammembers to run a 3rd party tool at all.

For the other part of the argument I think it would be reasonable to use video time for the records from the past that do not have any sort of ingame timer and add some seconds to account for the difference, probably like 2-3 seconds. This would only work for records from the past and it would be necessary to ensure that this timing variant does not provide any advantage (so probably when in doubt just add an additional 0.5-1 seconds). And I really liked the idea from Cruz of only considering records that get send in again by the original team. All of this of course only if we even want to retroactively track second based times for records done in the past.

This could also work in general to get time in seconds from video time, if we beforehand created a list with "punishments" for not providing the correct timing method. For example (not saying it should be exactly like this or the numbers need to be what I am saying here) we could say "if video is provided by any1 else than the first loader add 3 seconds to the time", "if loading screen is not visible for the whole duration to unify with ingame time add 2 seconds", "always round up to the next full second" and so on).

I still think the downside of enforcing a live-split type of thing based on ingame times is acceptable, especially if it is not part of toolbox but as a standalone tool. I am sure it would be helpful if ideas about how to make the technical aspects of this as little invasive as possible are adressed in the issue on the TB github (https://github.com/HasKha/GWToolboxpp/issues/553).

Sylars
Posts: 194
Joined: Sun Dec 18, 2011 7:32 am
In-game name: Musical Was Dub
Guild: [DS]

Re: second-based records

Post by Sylars »

Flo wrote:
Mon Nov 23, 2020 7:06 am
Trust

Trust is a great thing, but that does not mean we should be naive. There is a reason we currently require full video coverage for PURE and standard records. So my suggestion would be to incorporate some kind of Hash validation values into toolbox that can be printed out alongside the instance time timestamps. This can of course be manipulated as well (open source and so on), but I think there is absolutely no downside and if it discourages malicious behaviour even a slight bit it's worth it.
What? It's indefinitely harder to fake a video than it is to fake instance time and some random hash values. I don't even understand how this would help at all.

Literally every game with a speedrun community works the same way: video footage + livesplit. That's it.

Literally every commonly speedran game entirely forbids the use of any tools. We don't, but keep it consistent and have different tiers for allowing all tools or allowing none.

This whole thing should be so easy, yet it's so hard because for some reason people in a dead video game feel like they *must* have their own rules that are indefinitely more retarded than the rules literally every other game uses.
Flo wrote:
Sat Nov 28, 2020 12:25 pm
I have not met any of those people that refuse 3rd party tools completely in a speedrunning context (which might just be due to my filter-bubble), nor have they spoken up here. Also I would guess there is a huge intersection between people who will not record a video and people who do not want any of their teammembers to run a 3rd party tool at all.
You mustn't have been around any speedrunner of popular games?
Last edited by Sylars on Mon Nov 30, 2020 4:35 pm, edited 1 time in total.

Sylars
Posts: 194
Joined: Sun Dec 18, 2011 7:32 am
In-game name: Musical Was Dub
Guild: [DS]

Re: second-based records

Post by Sylars »

-deleted-

Doc
Posts: 38
Joined: Mon Dec 24, 2018 8:25 pm
In-game name: Doc Is Back At It
Guild: Back ConS

Re: second-based records

Post by Doc »

Sylars wrote:
Mon Nov 30, 2020 4:30 pm
Literally every commonly speedran game entirely forbids the use of any tools.
Exactly, everything not pure is a bunch of cheetahs

Flo
Moderator
Moderator
Posts: 1366
Joined: Tue Dec 22, 2015 4:23 pm
In-game name: Hi Im Undercover
Guild: Golt

Re: second-based records

Post by Flo »

The current plan is the following: We (and with "we" I mean Has or some other good developer friend) will enhance the instance/objective timer that is already incorporated into toolbox to achieve some key objectives. The most relevant feature of this timer will be that it will measure the time from the loading screen starting until an objective has been completed, across multiple instances. This will allow for clean, mostly undebatable time measurements.

This also ensures the following things:
1-1. The time given by this instance timer will be the same (slight difference in the 2nd digit after comma) as the time that could be derived from a video, determining start and end frame by hand. This ensures that people that do not want to run any third party tools can still do records.
1-2. The time from all POVs will be identical. This allows for records that can not provide appropriate video coverage (see 2-1) to not suffer from a time penalty, even if the first loader is unable to provide video coverage at all which would have been a concern in other timing methods.

Additional notes, clarfications and suggestions:
2-1. If video time is used and there is no ingame timer, then the video must include the loading screen when zoning into the area from start to finish. If that is not the case, then there should be a flat addition to account for this, which should discourage not providing complete videos. Therefore I suggest: a) adding 3 full seconds in case the loading screen is only partially visible in the video b) adding at least 5 full seconds in case the loading screen is not visible at all in the video, depending on how far the run seems to have progessed before video coverage started.
2-2. This timing approach also means that instance times visible on old record videos will largely be in line with current measuring, allowing old records to be migrated fairly easily. I suggest that handling times for old records where no sufficient ingame time nor video coverage can be provided should be done in the following way: a) If there is an instance timer visible in either a video or a screenshot, round up to the next full second. b) If there is no instance time and no sufficient video coverage, proceed as described in 2-1.
2-3. Statements 2-1 and 2-2 combined mean that we would be able to migrate every old record that has either video coverage or that is a single level record. We could also migrate old records that provide only a ending screenshot without an ingame timer by assuming every instance ended on 59 seconds ingame time. Of couse in the latter case the records would mostly not be competitive.
2-4. Yes, this timing method punishes people with slow load times. This is only somewhat relevant when doing low man multi level records. In higher party sizes it is very likely someone will load fast anyways, and in single area records that usually last for several minutes the time loss will be negligible. Also, most people will suffer from this in a similar magnitude. Load times are usually around 2 seconds, with some people loading slightly faster and some loading slightly slower. If you are loading considerably slower than 2 seconds, I would suggest testing around with removing some third party tools. In any case, some sacrifices have to be made and this seems to be the most reasonable one. So sorry to the guy with the 5 second load times that wants to beat a solo 5 level dungeon record, but for everyone else it should be at least understandable why this approach has a lot of pros.
2-5. By the way, I currently see no reasons to stop at seconds. Both the instance and the video time will have better precision (videos are somewhat limit to 1/30 intervals due to many recordings only being 30fps), so time measurements could definitely happen in 1/10th of a second intervals. I would not go further than that because of the fps stuff and to avoid debates about milliseconds.
2-6. The tables created this way would not replace the current minute based tables, but instead be included in another pinned post over each forum section. Newly posted records can then be added to both or just one of the tables.

Now of course you once again have an opportunity to provide feedback, I am sure there will at least be some people that dislike 2-4. The actual implementation could happen within the next couple of weeks, as soon as we have a proper tool to support this (at least for most records) and time newly created records in a way that does not require human interpretation.

Post Reply

Who is online

Users browsing this forum: No registered users and 166 guests