This topic is for the purpose of discussing #4079: MarbleousDave's NES Wheel of Fortune in 02:21.57

You couldn't have manipulated a better Bonus Round puzzle than BARNYARD? :[

For the curious. Link to video yeah it did it again just click here T.T

I can't speak for the optimization level, but I have to give it a No for entertainment.

Vault this.

yeah it did it again just click here T.T

I love the video ID for that one: UnmrmHrguaU.

You are using "suboptimal" words. Also you start the game much later. My WIP ends first round in 2354 versus your 2840 using the MONOCLE word. It's probably the 3rd "fastest" puzzle (in terms of frames spent on selecting the proper letters and the time spent on showing the tiles) of all but I think not all of them are manipulatable to this "no computer, single player" mode.

Here's the full puzzle list (ROM file 0x109DE-0x14617, NES address 04:89CE-05:C607). http://pastebin.com/rq2PmccX However, half of the puzzles are disabled in emulators that use 00000000FFFFFFFF RAM initialization, such as FCEUX and BizHawk 1.1.0 and 1.5.0. The reason is that the game uses address range 0x780-0x7FF as a bitfield to keep track of puzzles already done (0=not done, 1=done), but never initializes the range. Therefore, the 00000000FFFFFFFF initialization disables those puzzles whose "number" is greater than or equal to 32 mod 64. Note that BizHawk 1.3.0 and 1.4.0 (maybe others) use 55AA RAM initialization. So it also disables half of the puzzles, but a different kind of half. (I don't really see this as proper design in any case.) Also, I didn't mark it in the text dump, but the last 24 puzzles not starting with ! are disabled anyway (and regardless of any RAM initialization) because of the way the game selects the puzzles. The selection procedure is around 04:8006 in NES memory, if anyone wants to check it out. General RNG procedure starts at 00:8C4C. The RNG is manipulated by any controller input at any time while it runs the procedure, which is most of the time. Edit: Technically, the above puzzles can be eventually enabled, but only after answering all the "not-done" questions. That's a lot of them.

I've actually didn't debugged anything just used http://bmf.rustedlogic.net/oldboard/thread.php?id=88 to dump the word list. Later I found that Round 1, Round 2 and final round shares the same set of puzzles (around ~100 unique) while Round 3 uses 50~70 other ones too. Currently I'm using the fastest available puzzles (dumped the list, calculated the frames spent to select letters, however I think I miscalculated some), in order of rounds: - Round 1 - Monocle (1544) - Round 2 - Chestnut (1589) - Round 3 - Aaron Burr (1686 but I select a letter) - Round 4 - Barnyard (1589 but I select a letter) I need to remanipulate a 3 frame save on the wheel spin and I will upload it. Probably ends in 7875.

8-bit Vanna White is surprisingly well animated (but no Pat Sajak?). I think it's fine for the Vault. From what I've gathered, it seems like these were the fastest available puzzles that could be generated. Even so, I couldn't imagine this run being too much quicker, even if shorter puzzles were able to be found. A 'yes' vote for the vault!

FractalFusion: So is Wheel of Fortune badly emulated? Or is it just badly coded and has similar errors on real NESes?

I feel sad that I don't have my NES hooked up with which to end this debate forever by leaving it on for a long long time and getting many random puzzles. However, there are definitely game show freaks out there who play video game adaptations of game shows and develop a "run" where they try to beat the CPU a lot of times in a row. Here's the first episode of one such run: Link to video The first puzzle is row 1001 from the Pastebin, marked with a ! as inaccessible to the emulator. The second is 973, not marked. The third puzzle is 343, not so marked. The fourth puzzle is 812 (!).

Patashu, mathgrant: Now that I think about it, I think that this is just a matter of uninitialized RAM having no standard (with emulators using a custom of 00000000FFFFFFFF RAM initialization) combined with Wheel of Fortune using uninitialized RAM. We can't prove this is a bug; maybe they couldn't care less, or maybe they even intended it that way. Whether this counts as "badly coded" is one's opinion. It isn't, however, "badly emulated". There were actually a few threads regarding this some time ago. One of them involved "Blades of Steel", a game known around here for theoretical discussion and nothing else. Anybody who was around since March 2010 knows what happened next. Thread #8874: Blades of steel proves that TASs are impossible on hardware There is also this surprisingly recent thread: Thread #13888: Defense of initial console state for TAS playback So, if any initialization of RAM is just as valid as any other, maybe we are free to choose how it is initialized, even if we have to start from savestate in FCEUX?

Here's the lengths of picking a letter from a previous letter: (yellow notes starting position) https://docs.google.com/spreadsheet/ccc?key=0AqMjARkWFsmedGNUVExYaEdJLTFCSzJtTGE5VktYbFE&usp=sharing It's basically 8 + distance between the letters and + 44 when going through Z. However I didn't proof checked it, but it's correct enough. I've also uploaded the WIP (unfortunately, I couldn't manipulate the 3 frames save on the spin wheel, maybe because it's time is puzzle related?) http://tasvideos.org/userfiles/info/9624287711530343 If we would make a lua script that solves all the puzzles and use it as a savestate, I think all the puzzles (except WheelOfFortune for sure and maybe some of from the last ones) could be manipulated and would save much more time than my WIP.

I so want to see a bot do this now.

om, nom, nom... blech, salty!

I so want to see a bot do this now.

Hmmm...

Well it still can be improved.

Provided that we are permitted any starting initialization for 0x780-0x7FF that we choose, then I have confirmed that sub-1:56 is possible. I still need to check a few things.

Provided that we are permitted any starting initialization for 0x780-0x7FF that we choose, then I have confirmed that sub-1:56 is possible. I still need to check a few things.

Would this choice of RAM initialization become part of the input file? (Would a NESbot be able to follow up on this?)

Provided that we are permitted any starting initialization for 0x780-0x7FF that we choose, then I have confirmed that sub-1:56 is possible. I still need to check a few things.

Would this choice of RAM initialization become part of the input file? (Would a NESbot be able to follow up on this?)

Unrelated, but would this also allow Cheetahmen II to play back using the original rom?

Provided that we are permitted any starting initialization for 0x780-0x7FF that we choose, then I have confirmed that sub-1:56 is possible. I still need to check a few things.

Would this choice of RAM initialization become part of the input file? (Would a NESbot be able to follow up on this?)

Unrelated, but would this also allow Cheetahmen II to play back using the original rom?

Doubt it. The way the bug in Cheetahmen II works is that the game can sometimes boot into the the last 2 levels if the RAM is dirty in the exactly right way. The TAS doesn't use a reset to achieve this, but a hack that fixes the bug.

Well, for a high score run, I might go for the N64 version. Like round 1, keep landing on the $1,000 space. Round 2, land on the $10,000 wedge for a single consonant, and keep landing on the $2,500 space. There's also a prize for round 2. Round 3, keep landing on the $3,500 space to build up the jackpot (i.e. $40,000+) and land on the Jackpot space for the final consonant. Rounds 4 and later, keep landing on the $5,000 space, make sure it lands on $5,000 for the final spin. There's another prize for round 4. Or maybe the PS2 version: Round 1, keep landing on the $2,500 space. Also get the $10,000 wedge. Round 2, keep landing on the $3,500 space to build up the jackpot (i.e. $40,000+) and land on the Jackpot space for the final consonant. There's also a prize for round 2. Round 3, keep landing on the $3,500 space, also land on one of the mystery wedges and get a prize. Rounds 4 and later, keep landing on the $5,000 space, make sure it lands on $5,000 for the final spin for $6,000 a consonant. And last but not least in the bonus round, land on the envelope that has $100,000. The Wii version is different. In that version, the Jackpot round is first, keep landing on $2,500 to build up the jackpot (i.e. $30,000+) and land on the Jackpot space for the final consonant. Rounds 2 and 3, keep landing on the $3,500 space, also land on one of the mystery wedges and get an extra $10,000. Don't forget the Million Dollar Wedge, and round 3 is a prize puzzle. Don't forget the featured prize. There are also toss-up puzzles. Rounds 4 and later, keep landing on the $5,000 space, make sure it lands on $5,000 for the final spin for $6,000 a consonant. And last but not least in the bonus round, land on the envelope that has $1 Million. To FractalFusion: Show me how the NES version can be beaten in under 1:56.