Therefore, it seems more likely that you timed not from console power-on as we do, but from the moment you selected Sonic, which would more likely explain the 11:02.45 you're getting.
This is correct, as I explained in the comments of the submission. I failed to realize, that tasvideo.org uses the same timing method for all games. I tried to remove the unnecessary frames in the title and character select, but I it corrupts the rest of the TAS in the same way editing any level corrupts it :(
And since BizHawk has lag counter, probably lag reduction could be possible.
I am aware that BizHawk shows me lagging frames, but how can I go about reducing them? Do you have any pointers for me? Is this game specific? Sonic Chaos (NTSC) just happens to lag a lot...
I noted that some of the stages is 1 second in-game timer slower than the published run
I watched both videos (the relevant levels) side-by-side and have to admit, that calling it "_far_ less optimized" probably was a little bit exagerated. He has partially worse strategies, but does other things better. All in all he does some levels faster than me and I do some levels faster than him. Don't get distracted by the ingame time though, as a single frame (16ms) can make a whole second difference and 59 frames can be in the same 1-second-window. Ultimately I would say my run is still the faster one, since I execute the first three bosses better (on those bosses, I think I gain 3-5 seconds on mike89's run).
nd the glitch that enables you to ignore the running off the screen at the end of each act
I noted just now and didn't know about that glitch. I will research...
Rewind skip/frame? Never heard of this.
BizHawk -> Config -> Hotkeys... -> Frame Advance/Rewind (the first two options)
Also in the TAStudio tool of BizHawk you can select every single frame and jump to different frames.
The way my workflow was: I have rewind and frame advance on my shoulder buttons. If I want to record, I would hold frame advance (would make the game run at around 30 fps or so) and record stuff. If I needed frame perfect inputs or wanted to rerecord inputs, I would use the rewind shoulder button until I am on the right frame, push buttons and tap frame advance to advance single frames. While doing so, I had TAStudio open, which showed me which frame I was at, if there were frame lags, etc.... I had zero need for loading states and actually can't remember loading any states/branches at all (maybe I was trying stuff out in the beginning when I didn't establish a workflow yet).
I noted that you have a RAM Watch. I'd like to see if there are any more addresses I haven't found yet
Well, the whole RAM is filled with stuff... but the RAM watch in my videos is all I have + 12C2 contains a frame counter for the ingame timer, which is also relevant for the special stage glitch (I explained it on the speedrun.com forum as well)
Hi,
thanks for your feedback. If my submission ends up not beeing accepted, it's okay, but please let me address your points:
Very weak rerecord count: I think 64 RR
I don't see how the rerecord count is a metric of how good the quality of the TAS is. Maybe it is because my process was different from other TASers? I basically didn't rerecord anything, because I did all critical inputs on a frame-to-frame basis and used rewind/skip frame instead of loading states. As I wrote in the comments... this took me between 3 to 5 hours.
optimize the TAS as much you can
This is what I also addressed in my comments: I tried to do the best TAS possible, without checking any timings in between. After I finished the TAS, I realized that I couldn't redo any segments (even though I tried at least for 2 to 3 additional hours) because of the lag issues. If someone can explain to me how to get around these issues, I would be very glad! Otherwise, I wouldn't redo the complete run though (and more, because I would have to always additionally check that I didn't lost any frames in comparision to this TAS) to only save a single second, when this TAS _is_ the fastest known run by a far stretch.
Also, I noted mistakes
I said it myself, that my run isn't perfect, but I also don't think that anyone could optimize it by more than 5 seconds. When you say you noted mistakes, were there any crucial ones? Also note, that in order to trigger the special stage glitch, one has to loose frames/play imperfect in order to set it up (so maybe it is that)?
I strongly recommend to see the published run (and for comparision): [1504] SMS Sonic Chaos (Europe) "best ending" in 14:28.54 by mike89
I saw it before my submission and this run (no offense to mike89) is _far_ less optimized than mine. Just some examples from that run:
TQZ2: not going for the rocket boots is the worse strategy, because the rocket boots (even though you have to accelerate a couple of frames) gives you permanent max speed (7), while his route gives him an estimated avg. speed of 5 (so even with a conservative estimate he is 16,6% slower that route)
TQZ3: the execution on the boss is quiet bad (compare that to mine). The frames that could be saved on this boss alone in mike89's run is I believe 50% of the frame savings you could get in my entire run.
GZ1: After jumping over the big gap with the spikes, he misses the horizontal spring, which is a slowdown.
GZ3: The route to the boss is absolutely unoptimized. Here probably the other 50% of frame to be saved of my run are lost. (Compare that to mine)
(and no good ending)
You are aware, that good ending=100% run? I have problems understanding the feedback in that case. You are saying any% runs are uninteresting?
If my TAS doesn't end up being accepted I am fine with it, but if you are assuming it is not optimized and tell me to look at mike89's TAS for comparision, I ask you to take a closer look, as I believe the optimization of my run far surpasses that of mike89's even if the rerecord count would be zero.
PS: A quote from the 3rd best speed runner on this game in regards to my TAS (and I don't necessary agree... just take this into consideration if you critize the quality of the run):
From my understanding of the game, there isn't a single way to make this faster. 11:02 appears to be the absolute endgame.