SPF Hall of Records

[PhineasB]

I've never gotten improvement from Glide (and actually don't use it as a result because I had glitchy issues with my mouse capture and ultimately preferred going without capture). I agree that Windowed made a noticeable difference, and creating a separate shortcut without RMW extension seemed to make an even bigger difference. And having the single character also seemed to make material difference. My sense was that there is no one "solution" but little tweaks can shave average time here and there, but there are certain times were S&E randomly rears it's ugly head. But the tweaks collectively push down the average.

I also had noticeable (~1 sec) S&E improvement from cleaning out my machine with Disk Cleanup, in particular a file extension that I think relates to graphics (can't remember name). I did that periodically and the improvement diminished a bit, but I think it made a material difference (could also be supersitition). I'm not sure if that can be done on a Mac partition, but maybe worth a try?

Also, I haven't really tested Nihl runs with my new machine (SSD and faster, etc.), so I'm hoping that makes a further difference.

 

[Albatross]

Glide settings / how I organize shortcuts. As mentioned before, never noticed any of these making a big difference. I tweak them to make the game look and feel playable.

All extensions are off (no idea what they do).

Shortcuts without RWM do not have the "start in" set to the runewords folder. Note that screenshots are saved where the shortcut points to (can get a bit messy if I switch a lot).

 

[ffs]

I now removed the -direct command as well which seemed to help. Recorded Pit runs for some 20 minutes with no issues (even had Spotify running). Windowed mode (luckily) didn't seem to make a difference.

So it generally works fine now running with glide (similar settings as @Albatross), -3dfx, -ns, and RWM disabled. As for the remaining S&E issues, I'm 99% sure that it all depends on how long the particular run is: When testing different teleport routes with another Barb (only counting bosses while not killing anything, so runs were very short) I had a significant S&E delay again. That is, every single run. But when I run normally with him (roughly 55-65 second runs) it's S&E lag seems gone. Never had it, despite recording and having Spotify/Chrome open etc. It's in line with my observations from the Berserker recordings, where I felt like those S&E lags usually occurred after fast runs with fewer boss packs (making it even more annoying to lose additional time in suboptimal runs). I guess I could try and really find out which duration of run causes it exactly, but it's not really practical anyway, it's not like I am aware of how long a run is while doing it.

Regarding those teleporting routes btw:



This is level 1 of my Singer's Pit map. Boss spawning locations are barely accurate, but I included something just to show roughly how they affect teleport path considerations. Coming back to him after a while, I got into the habit of taking the orange "shortcut" swinging by two boss spawning locations. Runs with that path initially resulted in an average of "only" 7.9 bosses per run. I was perplexed by that because I had an HoR entry with a much higher 8.7 on this same map. So I went back to watch my previous HoR entry vid with him, and noticed back then I teleported along the white line in the middle. When I did that again now, average instantly went from 7.9 to 8.4 for the next batch of runs. Significant enough for me to conclude it's not just RNG but the white line is simply better. Not sure if it's actually an example of the mechanics discussed by @Fruit in the post I mentioned earlier because routing is not that different in this case. But it seems to support the principle that if you explore more of the map, chances are density is a little better on average long term.

 

[ResTTe]

I can definitely confirm that s&e lag is connected to what happens during the run. It tends to be longer when runs are short and shorter when the runs are on point. Can still got longer s&e even with optimal runs though. I can also confirm that -ns made my runs faster. Glide isn't that much helpful to me and i have some mouse issues on top of that, so i don't use it. Btw, you can disable gamma options in glide to make the game looks the same as with ddraw or the other mode. I deleted map files on my barb, he is my only char at the moment and i did some other things, but nothing really worked. I haven't observed any connection with running OBS and other things.

 

[Albatross]

I'm 99% sure that it all depends on how long the particular run is

I'll second that with a similar % of certainty

A very compact map came up during my last AT reroll attempt. Even though it had the chest spawn tile too close, it would have been worth running if it wasn't for the S/E delay which would kick in when runs were <25s. That's probably the biggest reason running short maps doesn't work for AT.

It's like every area has a different interval when the game will lag on S&E. Sub 50ish for Pit, sub 25ish for AT?

Edit: referring to ffs' bottom paragraph, similar situation here.

I've done 2 sets of 50 runs using the first route, and 2 sets using the second. Jumping over the first segment resulted in 6.27 and 6.46, but going there first I got 6.54 and 6.67.



I wonder if there are more "triggers" like this one. Difficult to test though.

 

[ResTTe]

When does the game save your char? Wasn't it like every 30 seconds or something? Maybe it is connected with this thing? It should be way easier for the game to save your char after it already saved it a few seconds ago, right? ^^

 

[Luhkoh]

Fruit explained this to me as D2 doing something he called "mutithreading". My layman understanding is something like d2 loads all these various modules for something in the background when you enter an area or something and has to close them all out when you s/e. And I guess some of these "threads" are just temporary where the game has to process through them and then they will close or be faster to close. So when you finish the run faster, it has to close down these threads when they're less processed and it takes longer. He speculated -ns helps in a similar way. It just prevents the background threads related to sounds from opening. Oh I found the message where fruit discussed this with me. Here it is in some of his words, which probably make more sense than mine. @Fruit hope you don't mind me posting that.

those long times appear to be caused by multi-threading
the game will start up a bunch of threads when a game is created. when you exit the game, the threads have to be terminated, but one thread often has to wait for another thread to close, before it can close down itself.
i have not looked at it much myself, but i suspect that disabling sound just means that the audio-related threads do not get started at all, thus they also don't need to be terminated when exiting
well, a theory for that is that the game starts up threads to load new assets (audio, graphics, whatever), and it's possible that the thread will be terminated when the loading of those assets has been completed. which could mean that:
enter countess level -> thread is created -> thread load assets -> thread is done and is terminated -> kill countess -> exit game
in that case, the thread was already terminated, so that would be one thread less the game has to wait on when exiting
yeah, i don't know. it's possible that the mac does not even do all this threading

@ffs I'm so jealous of that map. I love small maps for simple runs. Can't believe you have one that small with such a high boss count. Also I'm so interested in how such a small route deviation could affect the bosses. I thought the "segments loaded" were more or less akin to a map tile. But that small deviation shouldn't change the tile spawn order, so it's likely more gradated than I was thinking.

 

[art_vandelay]

@ResTTe Fruit tested this once upon a time. See the Edit in this post, it's not related to character auto-saving:

Diablo 2 Resurrected Best Builds, Guides, and News

Icy Veins provides detailed builds, guides, and news for Diablo 2 Resurrected for all classes: amazon, assassin, barbaria, druid, necromancer, paladin, sorceress.

www.diabloii.net

 

[ResTTe]

So it should be worth it to wait a little bit for the game to terminate those threads and then exit? Or just find a map that has average bosspack value of 9, so you always have over 60s runs and insane efficiency xD

 

[Albatross]

We'd also notice if the autosave was as fast as 30s while using the run counter which relies on counting saves to the char file.

@ResTTe I had similar ideas on my small map, but it's just too impractical and you can't really tell when it's done so yeah, better stick to higher density/longer run.

 

[ffs]

@ffs I'm so jealous of that map. I love small maps for simple runs. Can't believe you have one that small with such a high boss count. Also I'm so interested in how such a small route deviation could affect the bosses. I thought the "segments loaded" were more or less akin to a map tile. But that small deviation shouldn't change the tile spawn order, so it's likely more gradated than I was thinking.

As mentioned above the Singer map has a crazy good L2, spawning 3-4 boss pack very often – this plays a big roll in the high overall average. L1 is nice and compact but nothing too crazy in terms of boss amount, 4-6 is rather normal I'd say... The map is kind of the opposite of my Zerker map which has a big L1 but with great density, along with a subpar L2. Wish I could swap individal map levels around around between the two.

As I mentioned above I'm not sure how such small difference in teleport route makes a difference in terms of underyling mechanics, but I'm pretty sure the difference is real and not just confirmative bias. Differences between the two routes are rather striking.

 

[Luhkoh]

@ResTTe arent your runs over 60s anyways (just some trash talk from someone with like 2.2+ adjeff). But yeah while the thing about waiting to hit some time breakpoint before s/e is something that could theoretically work, I don't think it would work well at all. So yes I'm thinking this is just another point in favor of high boss count maps. Kinda stupid, but I know @Albatross had mentioned to me he can't properly test some superfast lowish boss AT maps due to this issue. And @Gripphon had the same issue where it made his p5 trav sorc better than his p3 runs, just because of that lag. And I'm not sure how much it varies computer to computer either. I know @Nano had good success with a lower boss count one. Weird stuff for sure, and I feel like we should have an easy workaround for a 20 year old game. But fruit didn't seem optimistic about an easy fix for it. Ah well kinda makes for some interesting results and theorizing too, even if s/e delay is pretty uncool mechanism for theorycrafting imo.

 

[ResTTe]

Yeah, it was mainly theoretical. No way someone would even try to test if it's worth it or not. It's probably too random.

Maybe bigger maps aren't that bad actually. I mean like 5 tile linear maps. You teleport more, but you can get better density and better s&e times on top of that. Interesting.

 

[ffs]

Maybe bigger maps aren't that bad actually. I mean like 5 tile linear maps. You teleport more, but you can get better density and better s&e times on top of that. Interesting.

This. Apart from the S&E thing, as mentioned to others in PMs I am kinda moving away from the notion of "level 1 needs to be small". I'm rather starting to think it is not of much relevance, in fact bigger L1 maps might be better in theory, provided that they enable a teleport path which spawns a high amount of bosses on average. What seems to matter is only good average density, with any reasonably efficient teleport route. Whether that is straight, long, short, linear, or totally curved – it doesn't really matter as long as it avoids actual backtracking and spawns bosses.

 

[Luhkoh]

Super interesting stuff guys. I'm so into this discussion.

 

[ResTTe]

When i was rerolling maps, my priority was to find a map with pretty good tiles and, what's more important, to get some optimal teleporting path. I didn't want to have 6 tile map, cause those are huge and they don't get any more packs than 5 tile maps or even 4 tile ones. I also didn't want to get a map that would require backtracking, which is kinda obvious. I got some maps with like 1 tile that i had to teleport in and out to continue farming, but they were never worth it. My main indicator was always bosspack density. If the map had sub 8 average after like 20 runs, reroll. Unless i got 8 packs in lvl 1 or 4 packs in lvl 2 at least once. @Luhkoh wanted me to make some sort of algorithm for rerolling pit maps, but honestly it isn't that easy to come up with such thing.

I still have Pit map guide in my mind, don't know if i will make it though.

 

[Kinkara]

I still have Pit map guide in my mind, don't know if i will make it though.

It's not a choice

 

[Gripphon]

To think more of it, I feel like best maps are those with simplest teleporting paths while having good bosspack density, not necessarily best bosspack density possible. My pit map isn't small nor is it in top 5 bosspack density maps presented, but it has only 2 teleporting paths and I can play it almost blindfolded. I could find more dense map in theory, but will it really challenge current map in combination with it's simplicity? That is the question.

AT map I have for barb is godly, but it doesn't have simple teleporting path and it feels more taxing to play. For this thread it might provide great result, but playing some MFO on it will be harder compared to some simpler yet slightly less dense map.

So, highest density of bosspacks isn't the only thing that matters when comparing maps.

 

[ResTTe]

I can tell from my experience that if you are looking for a really good map, it will rarely have the average from like 20 runs below 8. If i saw something way below 8, so like 7,6/7,7, i was rerolling that map immediately. If the average was closer to 8 i was always doing some more runs, but pretty much all of the maps i got turned out to be trash after more testing. I was doing between 50 to 100 runs after i considered the map to be worth checking and the chance that a good map will have the average below 8 after that many runs is pretty much nonexistent. I am talking from my perspective of course, which is the dude who got 2 really good maps and did many thousands runs on multiple maps with counting packs in pretty much every run perspective XD I can always go to my excel files and calculate the average from any given number of runs and show you what are the chances of getting specific average from those runs on top of that. I probably have the data from about 15k runs.

I stopped testing some weirdly shaped maps long ago, cause they were never decent and i didn't bother with testing different routes. I wasn't even aware of the fact that different teleporting path could affect spawns at that time. Maybe it is worth testing after @ffs discoveries.

Since i was always focused on pretty much small, linear maps, bosspack density was definitely an important parameter for me. Except checking density, I was also calculating adjeff which definitely is the most important parameter after all.

 

[Kinkara]

@Gripphon do you have by any chance a 50 runs video with the AT barb or something like that? I know you have a pretty short one with just a couple of runs.