United KingdomIt's that exciting time yet again where, as the speedrun develops, it proposes the same old question of what version is optimal in an RTA run.
This thread is focusing on Any% Warpless. For 101% you want a 30/39k NTSC-U PS2, for All Cups the best is 30/39k NTSC-J PS2 (or is assumed).
So how do we identify how good a console is? CTR records your course time on each track, as you'd expect from a racing game. This is what we're calling "In-Game Time" (Or IGT for short). The problem with IGT is it is not 1:1 with real time. In fact, 1s on the timer in NTSC is about 1.04s in real life. For PAL 1s in game is around 1.05s. Despite this, both versions play almost all tracks similarly, in that your course time can be close on both versions. So what does this mean for a speedrun?
We can take this IGT component out, then determine how much "Non IGT" there is per run. This will cover hub movement, load times, cutscenes and lag all of which differ depending on what console you decide to use, as well as the RTA/IGT difference mentioned above. The theory is, the console with the lowest Non-IGT component, the more optimal it is for a speedrun. Obviously if you crash in the Hub it'll skew the results, but adjustments have been made where necessary.
So, first let's cover NTSC-U. The main factors for this version are lag, cutscenes, animations and pause glitches. NTSC is a laggier version than PAL, and has lag on most tracks in consistent places (the water on Crash Cove is a good example of this). The console has to work harder to keep a consistent 30fps, and unfortunately isn't capable of doing so (thanks Sony). As for gameplay, it is considered easier than PAL as NTSC plays more consistent. Higher hops reduces the chance of froggy chains being cut short so you keep SGs for longer, and some skips the extra froggy height is beneficial (Polar Pass).
For PAL, you have the language glitch option and almost no lag, bar some exceptions (under the bridges on Tiny Arena for example). A consistent 25fps is better than a laggy 30fps. The PAL language glitch swaps cutscene lengths around, and completely removes the Komodo Jo cutscene which saves 22s. However, Oxide for example is some 10s longer (I haven't done a full analysis on this). Another main part of language glitch is the trophy award speech. On 3 tracks per hub, when you win a trophy, the game doesn't play the line "Congratulations, you won a trophy" on PAL. This saves in the region of 2-3s and each time and can get this on all 4 Hubs, so 12 separate occasions in the run. NTSC can do this from Hub 3 onwards, so has 6 chances. However, on NTSC it is a 1-frame trick, and going from most runs, 0-2 pauses per run is an average, occasionally more.
So there's some important aspects of versions, now to console models.
For consoles you have 30/39/50k (Fat consoles), and 70-90k (Slim consoles). On NTSC, slim consoles lag a lot. In some cases lag loses 20-25s or more over the run, so we can rule out these as optimal versions to play NTSC. They are still viable until 48 level where these seconds really matter.
30-50k Fat NTSC models have less lag, and loads are good too. On average, the Non-IGT of these is somewhere from 14:55 - 15:05. There has been cases of Fat consoles reaching as low as 14:45, so if you were to get a brand new Fat console then you could save even more time.
For PAL, from testing I have done all consoles run very close to each other, so 90k won't lose you much time. We don't have a good sample of PAL 90k, but from my own experience it is close to a 50k. However, 30-39k consoles seem to be in the range of 14:40-14:55 Non-IGT. My own 39k console is closer to14:40 than 14:50 and a console from another runner is also close to this.
So what does this mean? The PAL tests we've done are for runs in the region of RTA 50 - 53mins (or so). If the IGT in a PAL run is to reduce, the Non-IGT reduces at a higher rate than NTSC. So, a theory I have is if a PAL run was to match the IGT of the current WR (49:06), the RTA could be into mid-low 48:4x. Better IGT = Better Non-IGT. At optimal play, a PAL console will have better RTA than an NTSC console.
Now of course, PS2 age and from the sample we have, only one NTSC console rivalled PAL, but that's not to say more don't exist. As consoles age, the loads get worse and this can lose 10-20s, or even more when the console is really starting to deteriorate. I don't think it's a sensible idea for someone in America to import PAL console, and vice versa, as you're playing a lottery in the hopes of scoring a good console. I do think people in Europe and other PAL regions should look to PAL as an option as replacement units are much easier to find. Americans can easily keep using Fat NTSC consoles, and again have an easier time finding a good console.
This hopefully sheds light on the console discussion, but please post your thoughts or let me know if I missed anything/wasn't clear in places.
tl;dr PAL on average is somewhere in the region 10-15s quicker than NTSC, assuming you can play well and your console isn't dying. PAL is better for optimal level using current data. If you can hit 6 1-frame trick NTSC pauses every run then you're a god.
Fuck Sony.
As a much discussed topic over recent months, the idea of encouraging runners to use emulators is reaching a point where some changes to the boards may need to be made.
Before anyone jumps on my back, the proposed changes are not going to affect the console boards and PS2 will still be optimal.
Anyway, the ideas I'd like to propose are: Separate boards for Emulator runs Strict rules on banned Emulators to only allow BizHawk/Mednafen IGT tracking as discussed in a previous thread to aid comparison between emulators and consoles
This may mean a lot of current emulator runs will need to be reviewed and though there's never a "right" way to handle this I'd love to hear what others have to say, we can cross that bridge later.
Hopefully a dedicated space for Emulator and a standardised approach will encourage people who don't have consoles, or prefer Emulator over their current setup, to run the game and feel more competitive.
In the interest of competition and fairer comparison between runs, I'd like to propose the idea of adding an "IGT" column to the Warpless leaderboard.
IGT which stands for "In-Game Time" can be calculated from the sum of all course times over the run. This is the best comparison which will negate the differences between console lag, load times and region. The course times in CTR are accurate with or without lag, NTSC or PAL.
To be clear, however, this is not to replace the current RTA timing. Rankings will still be decided by RTA, and IGT will simply be for information only. This won't take Hub movement into account, or NTSC pause glitches.
At the moment the easiest way to calculate IGT is inputting your course times into LiveSplit (or your split programme of choice) which will add them up for you. A programme is currently in development to calculate IGT automatically from a video feed, and I'll edit this post once it has been completed and tested.
Initially I'd like to suggest either the top 10 or 20 times have the IGT calculated, or any time which is sub 53:00, but I think it's best to give runners the opportunity to submit their IGT runs if they want to and not have this as a mandatory requirement. This might all go to waste if SR.com doesn't allow optional IGT lol.
This has loosely been mentioned in the Crash Discord and the majority seemed to support the idea, but please post your thoughts in this thread before decision is ultimately made.
Traditionally speedruns have been done using Coco or N.Gin, or specifically "Acceleration" characters, and recently it's been discovered that "Fast" characters such as Dingo and Tiny are optimal for runs.
So it was unanimously decided that "Well Dingo is just faster for runs" for a while and no accurate comparison had been done. Anything from 5s, 10s or even 30s had been plucked from the sky to guess the actual timesave for using Fast over Accel so I ran some tests to look into this properly.
I tried to make it as accurate as I could, while keeping my own sanity in check. I took 7 completed attempts at each course without mistakes, then took an average of the 5 attempts, removing the fastest and slowest attempt. This isn't so much to demonstrate the absolute peak each character can achieve, but to instead look at the average course times.
Here's a summary of the average and fastest course times I recorded for all tracks you would see in a Warpless speedrun:
The comparison on the right column is looking at the time Acceleration loses compared to Fast. Over the whole run, I came to the conclusion that Acceleration loses around 27-28s compared to Fast. Now obviously this isn't an exact number, and getting a completely accurate sample of course times would take a very long time.
I condensed the data further to look at the time difference for Hubs:
Quite surprisingly, Fast saves over 9 seconds in Hub 4, which is almost all down to Hot Air Skyway and Oxide Station.
This does come with the argument of Fast characters being significantly more difficult, which I'd personally disagree with to an extent. The main problems come from Papu's Pyramids, Dragon Mines and Cortex Castle, which makes up 5 tracks in the run including bosses. These become hellish to drive using a Fast character.
However, on the flip side, Fast characters are far more consistent and easy to optimise on basic tracks, and a lot of tracks don't become more "difficult", instead just require a different approach, such as HAS and OS (imo).
So there you have it, feel free to post any questions about the data I used.
tl;dr Fast characters saves 27s in Warpless so you might as well use 'em!
Hello speed friends. For those interested in playing time trials and being involved with the CTR:NF TT community, there's a dedicated website, leaderboard and Discord server created by the original CTR community:
Website: https://crashteamranking.com/ Discord: https://discord.gg/hkBPZNN
There's also forums, lobbies and competitions!
When submitting a run, please take a look at the following classifications to ensure your submissions is accurate:
Fast Characters - Dingodile & Tiny
Acceleration Characters - Coco & N.Gin
Balanced Characters - Crash & Cortex
Slow Characters - Pura & Polar
Here's a handy chart detailing the exact stats for each character (the in-game stats are incorrect):
If you're interested in joining the community, we have an active Discord server:
Here you'll find lots of helpful guides and resources for (almost) all Crash Bandicoot speedruns as well as the opportunity to ask other runners for advice. Everyone with an interest in running Crash games are welcome!
There's also a dedicated server for all things NST, which you can join here: https://discord.gg/aGjpE6e
It's sort of common knowledge that PSP runs CTR faster than a PS1/PS2. However, what we weren't aware of was that PSP doesn't just load quicker than PS2 FDS, but it also runs games about 4% quicker than intended.
Here's some evidence for stats nerds (I'm using Joshimuz's and Pete's runs for comparison as they're currently the only ones on PSP):
On here, we can see that IGT runs about 14-15s faster over the three tracks that I tested. On Josh's run in particular, 60 seconds RTA is actually 61.9s IGT.
RedHot's run and my run are both done on NTSC PS2s. To compare, on my run, 60 seconds RTA is 59.4s IGT.
There's also this neat cutscene comparison made by Pull that shows the differences:
So basically, what we're going to do is separate games that run on a disc, like PS1/PS2 (and we'll keep emulators here as well as they emulate a PS1 console), and runs that use PSN such as PSP or PSVita etc.
The main reason is to keep a healthy competition between runners. It's estimated that PSP saves about 2 minutes over a PS2 Any% Warpless run due to poor emulation, and as PSP is still official hardware, we don't want to simply ban it. Currently, we could see PSP runs reaching as low as a 48 minute run.
Feel free to post your thoughts here!
This has been a topic coming up in numerous debates recently, and it hasn't been stated anywhere what emulators are actually allowed for the Crash series.
This thread serves to inform those who are unsure and is not a place for discussion. If you feel there is an emulator that should be allowed but isn't listed then please start another thread.
PS1 emulators is where most (if not all) the controversy comes from. As FDS is allowed for PS1 games ran on a PS2, it has been noted that "Emu is always going to be slower than console". However, the whole aim of the emulator is to be accurate to the console that the games were originally meant to be run on, which is the PS1.
GBA and Nintendo DS emulators have not been heavily discussed, however to keep the boards conformed, please use those listed below.
PS1:
UPDATED RULES BELOW
GBA:
mGBA higan VBA
Nintendo DS:
DeSmuME melonDS
If your run uses an emulator not listed here then it will not be accepted. No WR is currently set on a banned emulator, and therefore is not a cause for concern. Again, please open a new thread if you would like to discuss this.
Final note, PS2 and GameCube emulators are not allowed. These are very inaccurate and your run will not be allowed on the boards.
