For what it's worth; I would NEVER want to see a legit run be taken down however improbable said run appears.

As a fellow TASer, nymx asked me for my opinion on the Tetris 2 run in question here. The following are some observations I've made on the run itself as well as other observations regarding Cyber_of_Memory. None of the following proves that this run isn't legitimate, but it does make me wonder.

1. I almost immediately noticed how precise the splits occur in the LiveSplit timer. Going through the video frame by frame (where necessary), showed that every split occurs within 3 frames of "Round Clear" appearing on-screen; with multiple of these occurring exactly on the frame where "Round Clear" appears and some even on the frame immediately before it appears. Out of curiosity, I checked his other two 1st place runs in the other categories for this behavior, and the nearly perfect splitting occurs with them also. Unless the runner is using an autosplitter script, this is insanely good timing for a human to accomplish at that level of consistency over 3 different runs. Given that there is no standard autosplitter available in LiveSplit for the game, this adds a degree of suspicion for me. Further, an autosplitter would have to be one that watched in-game RAM as some of these splits occur the frame BEFORE the text appears. Thus a screen-watching autosplitter wouldn't work. (More thoughts on LiveSplit later)

2. Some of the piece drops, rotations, and dangerous maneuvers in this run appear very precise. I'm no expert on this particular game, but there sure seem to be a lot of such crazy precise maneuvers made in this run. Were this a game that always presented the same layout for a given level, I can see where a player may get proficient in maneuvering blocks through the obstacles; but given that RNG affects stage layouts, such maneuvers cannot be pre-planned and have to be decided and performed in-the-moment.

3. Another thing that makes me curious is that Cyber_of_Memory only joined speedrun.com about 2 months ago (approximately when his 10-round run took its 1st place position). According to the 'Info' tab on his profile page https://www.speedrun.com/user/Cyber_Of_Memory he only has 4 total runs submitted on speedrun.com; three of these are the 1st place finishes on this game's categories. I don't know what the 4th run is, but I have a speculation (again, more on this later). I find it somewhat suspicious that a runner who shows up out of nowhere, suddenly takes first place in all categories of a specific game within a 2 month span of time, and he further appears to have either created his own autosplitter or is insanely consistent at splitting on an in-game event that has no visual or audio cue that would allow for a split that precisely when the text appears.

4. Other videos in his youtube channel show some pretty amazing and nearly TASlike gameplay though none (from what I can tell using google translate) are labeled as tool-assisted. For example, this run is a run of a player vs the computer in Dr. Mario where the player plays at level 20 and the CPU plays at level 0 with both set on high speed; yet the player still manages to win with nearly perfect pill movement and placement throughout the run.

5. After looking into Cyber_of_Memory's youtube channel, I found this run which I'm wondering if it may be his 4th submission on speedrun.com, but the run has perhaps yet to be approved by the mod of that game https://www.speedrun.com/tetris_2_snes#Vs._CPU_rounds_1-15 and is thus not yet visible on his profile. This is a speculation on my part. What I find interesting though that this run ALSO shows the behavior of near perfect time splitting even though it's a different game (and system) entirely. This again suggests that either Cyber_of_Memory is nearly a perfect manual splitter or is using an autosplitter. Given that this version of Tetris 2 is different, he would have had to create a completely new autosplitter for this version of the game because, again, there is no standard autosplitter available in LiveSplit for that version of the game. Further, as with the run this forum topic is discussing, this SNES Tetris 2 run would also be the top spot by a rather wide margin; nearly 20 minutes shorter on a game category where the only other runs posted are approximately 1 hour and 1.5 hours. If I'm correct in my assumption that this is indeed his fourth submission; then we have an individual who shows up out of nowhere and takes top leaderboard positions in multiple categories with 4 runs over two completely separate (though similar) games. On top of taking the top spots in all those categories, he does so with some quite impressive improvements when considering the ratio of time saved compared to overall length of run (both his own and those being beaten).

6. LiveSplit: From his youtube channel, it appears that Cyber_of_Memory has only begun using LiveSplit on his videos for these past 4 videos. As they are the most recent videos on the channel, it could just be that he's just now started using this method of timing. This did make me further curious about the potential use of an autosplitter that's not standardly available.

7. This one is a minor note, but it does play into my suspicion. Cyber_of_Memory doesn't have a twitch account (that I can find) where these runs were performed live. I realize that many runs on speedrun.com don't have a live recording to compare to, so this is hardly a deal breaker. But the combination of my other suspicions already noted and the fact that the player is not visible playing these runs, I would have liked to have found a live recording of the run to verify this youtube video against.

As mentioned at the top, none of this proves that the run is illegitimate. It sure does make me suspicious though. The split timing is actually what is bothering me the most; especially with the method of splitting being near identical between how the 2 different versions of the game are split. I just don't see a human being that nearly perfect in manual splittng. So I'm left with 1 of 2 possible conclusions:

1. The run is legit. And thusly, Cyber_of_Memory is an absolutely AMAZING game player showing near tool-assisted levels of gameplay over a variety of games (as seen on his youtube channel). He also is either nearly perfect at manually splitting or is using a non-public RAM-watching autosplitter.

2. The run is not-legit. And while not perfect, the run was still made using some form of tool-assistance. If so, I'd guess it was savestates and rerecording a movie. Then the livesplits were somehow added in over a playback of the recorded movie.

Regarding duplicate seeds... Some solutions are used as the starting word for multiple sequences with differing latter answers.

Here's the script: https://pastebin.com/v4K33cs8 It may not be the most efficient way of doing it, but I'm a novice when it comes to programming. And hey, it works.

As far as the best starting point....from what I've been able to discern based on my work, A theoretical fastest game (which I've created via TAS) pushes the 'Start' button to enter the team name on frame 481 and yields the word sequence starting with 'CUP'. This theoretical perfect game also always rolls 6 on the die requiring only solving 6 words to complete the game. This also limits all previous 'Start' presses to a two frame window leading up to the 481st frame as the 481 frame is only the third possible frame on which to press 'Start' for the team name.

For RTA runs: CHEESE, WINK, and NO should be desired sequence starters with some of the fewest inputs over the course of up to 10 words. Their starting frames are 492, 496, and 502 respectively.

The WINK sequence has the fewest inputs over the course of 7, 8, 9, or 10 words. CUP only beats WINK for number of inputs through 6 words. From this WINK is probably the ideal desired starting word for an RTA run. NO only trails WINK by about one letter input, but is also a later seed.

Here's a link to my current spreadsheet of word sequences. https://docs.google.com/spreadsheets/d/12_ealGeGx_VdxQLL0J2xylnTWGoEjA7itqnwEeJF7wc/edit?usp=sharing

I've included the first 10 words of every sequence I've tested through.

There are two sheets. One has the words sorted in order they appear based on what frame the 'start' button is pressed to enter your team name. The other is sorted by length of the first word in the sequence. I will update this spreadsheet as I am able.

Beware! (Thus Far) the following starting words have multiple sequences: ARM HONEYCOMB

Final note: For anyone who cares or is just wondering, the glitched puzzle's answer is DRAINPIPE.

Thanks for your perspectives.

Regarding distance traveled from letter to letter, I have found that it is often (but not always) faster to wrap around the alphabet than it is to change directions when traveling between letters due to decelleration/acceleration time for the shoe cursor (this assumes drive-bys). So calculating distance between letters may not yield the fastest choice.

I'll see if I can determine how many letters apart one has to to be to make a changing directions more worthwhile than just wrapping around the alphabet.

Also, based on my current work, "CUP" and "TEARDROP" may be early sequence starters that yield the fastest runs. Over the first 6 words in the sequence, these two starters yield only 27 total letters and 28 total letters respectively. Other seeds I've worked through average about 40 total letters over the first 6 words.

WOLF, ANTENNA, and WINK also are all on the lower 30's end of this range so should be worthwhile seeds from an RTA standpoint.

I'll keep working on this as I'm able.

While playing around with TASing this game, I've written an lua script that prints the answer to the puzzle in the upper left corner of the screen.

Would runners using lua enabled emulators be allowed to use this script? It does not do any inputs for the player, just prints the answer onscreen.

In my mind this is the same as having the list of answer sequences in front of you when playing, it's just on the game screen instead of on a paper (or different screen).

If this would be acceptable, let me know and I'll make it available.

Side Note: I'm also using this script in conjunction with TAS tools to further develop the spreadsheet with the answer sequences. I'm focusing on sequences not currently on the sheet. Hopefully this will ultimately help all runners including those using consoles instead of emulators.