Because I don't like ties, I decided to time frames for a run of FlandreB stage 1 I just submitted against the record at the time, since they both displayed 16 seconds for the in-game time. When adding up the time from first frame of control to last frame of boss health bar displaying for each phase, though, I was consistently getting above 17 seconds for the current record, specifically 17 seconds and 5 frames. That's a small enough amount over 17 seconds that I suppose I could have been miscounting, but I kept coming up with that time, no matter how many times I recounted it. But, if you subtract the hitstop frames that occur when attacking Reimu while her barrier is red (and the attack doesn't end the phase), which is 19 frames for Flandre's projectiles and 5 frames for Laevateinn swings, the time for the current record becomes 16 seconds and 15 frames, while also keeping my submitted run within the 16 second range as well, as I only used a projectile to break the barrier without ending the phase once in my run, while the current record did so twice, adding an extra 19 frames to it. It's certainly possible I'm wrong about the specifics and more testing is certainly needed, but I think it would be nice to be able to determine times into the milliseconds, as many individual level runs take less than a full minute already, and some don't even take 10 seconds.

This sounds like a lot of manual work to do. I wonder if such fractions carry over between stages to give it more practical use. I'd prefer it if it could be automatically fetched from the game as I find it can lag to "external" reasons sometimes, requiring more nuanced adjustments on real time.

That said, real time is good for such comparison purposes, although it favors doing some odd things such as reducing the noted in-game hitstop lag (from ensuring spawned projectiles are halted for the freeze frame?) and keeping enemies on the boat for attack transitions (is this significant?). It's optional though, and manual use of timers may have it reported ~2 seconds off, but it can always be standardized seeing as the videos are there.

The game keeps track of fractions, that's not an issue on full runs. I'm not sold on the idea of frame counting on levels. Due to the short nature of levels, ties will happen even if you do the manual labor of counting every last frame, the only difference will be that, right now people can achieve ties with different strategies, giving a good amount of diversity to strategies. If we ever decide to change this though, it'd probably better to use RT then.

I may be able to use Cheat Engine to determine the exact frame count, and potentially make a modded version of the game that displays Minutes:Seconds:Frames at the end of a stage instead of Hours:Minutes:Seconds, though I'm certainly not the most skilled at writing assembly or deconstructing games with CE. As for ties, it just feels weird to me that, in the most extreme example (Murasa stage 1), 2 runs could theoretically be considered a tie despite one run (a theoretical 7 second and 0 frame run) being nearly 15% faster than the other (a theoretical 7 second and 59 frame run). Even for the longest currently existing individual level runs, it seems the difference would still be greater than 1%. I am fine with moving to real time as the main timing method if it is deemed necessary to swap to it, and in fact I was actually in favor of it from the beginning.

Since the IGTs in this game can vary quite a bit, even for the same number, I feel that using RTs to measure them is worth considering in a category such as the Level Run, which is a fairly short segment of the competition. From the discussions I have seen so far, there seems to be both support and opposition, so as soon as we are in a position to determine that we have reached a generally acceptable conclusion, I will decide whether or not to make the transition to RT with respect to level runs.

At least, regarding the modding of the game, I don't believe that to be a solution, given that a Switch version is incoming. As for IGT vs RT, there is more to discuss about the differences, text for example becomes a huge time difference, opponent movement and lag also affect this, it is not that simple.

I don't particularly mind that ties are more likely to occur when using full in-game seconds. One could still count frames for personal satisfaction and competition, the only difference would be the exact time displayed on the leaderboards.

It may also be awkward to rely on 3rd party PC software when the Switch version is just around the corner.

My personal opinion is against measuring frames through modifications rather than external timers. The reason is that it would be a psychological and physical hurdle for speedruns, assuming that the modifications are introduced by volunteers rather than by official patches. At least at this time, I would prefer to operate on the premise that I am competing using the game as it is officially released.

I had kind of forgotten about the Switch version, oops. That would definitely get in the way of mod-based methods of frame counting, yeah. Theoretically, a piece of software could be created which would frame count the footage, rather than the game, for us, if fed the correct details, but that's quite a bit more complex than modding how a number in a game is displayed, and well beyond at least my own capabilities. I do think that the timing methods should be consistent between IL runs and full game runs, not only for the sake of logical consistency across the boards, but also because if timing methods were different, strategies would be as well. It may just be my overall inexperience with speedrunning, but it seems odd to me that, in a game in which there are no resources carried from one level into the next, full game runs would not be able to look to individual level runs as an ideal scenario. Personally, I'd prefer real time being used for both full game runs and individual level runs, but if full game runs will continue to use IGT regardless of what is decided for individual level runs, I think they should both use IGT.

Well, turns out the game doesn't even time itself properly lol. I think it's probably random chance on whether it increments any given step by 1 or 2, since I timed a recent attempt using CE and got a 234 IGT versus my count of 225 on phase 1, but got a 451 IGT versus my count of 445 on phase 3. The smaller gap on the longer section makes me think it's probably random, but it could be related to certain actions taken by a player, instead.

I don't believe that the increase is "random", I will see into this myself, thank your hard work so far.

Since no new opinions seem to be forthcoming, I think it is time to summarize my conclusions. Currently, only one person is in favor of using RT, and there are two people who clearly state that we should continue to use IGT. We have also received some comments from Japanese runners that they would like us to continue using IGT, although they are not listed on the forum. Personally, I feel that it is better to continue using IGTs, which do not depend on the performance of the PC, so unless any new circumstances arise, I will continue to use IGTs in level runs.

Blue_Wokou, I am sorry I could not meet your request this time. However, your research has helped me to understand the frame movements and when IGTs are counted, which is very significant in this game of speedruns. I would like to take this opportunity to thank you again.