Fisher Price: Perfect Fit - December 20, 2016

Edit history:

SDAVerification: 2016-12-20 08:54:57 am

2016-12-07 12:30:39 am

Game Page: Doesn't exist yet

Fisher Price: Perfect Fit (ntscus) (nes) [Any %] [Single Segment]

Decision: Reject

Reason: The RNG should be pushed a bit further until better board setups are had.

https://queue.speeddemosarchive.com/verificationfiles/1970/

This run will be available for a month. After that these link(s) will no longer work.

Thread title:

SDAVerification

2016-12-07 12:30:40 am

Run Information

Fisher Price: Perfect Fit (ntscus) (nes) [Any %] [Single Segment]

Verification Files

http://v.speeddemosarchive.com/fisherpriceperfectfit-20161025/FisherPricePerfectFit-WR.mp4

Please refer to the Verification Guidelines before posting.

Please post your opinions about the run and be certain to conclude your post with a verdict (Accept/Reject). If you wish to remain anonymous, you can also send a pm with your reply to 'sdaverification' (please state clearly in that case which run you have verified). This is not a contest where the majority wins - Each verification will be judged on its content.

Uilnslcoap

2016-12-07 08:10:22 am

F*ckin' sanity effects...

Haha, entire video less than a minute long. Without knowing anything about the game, it seems as though the point at which the object can scroll horizontally is predetermined? If so, it probably couldn't be much faster, and the video/audio are fine. Accept unless someone has experience with this game.

mrprmiller

2016-12-07 08:54:32 am

Everyday is puppies and sunshine...

Lol. If it is possible to move up directly on puzzles 2 and 3 rather than up and then over into the objects, that might be faster. Otherwise this would be a perfect candidate for a TAS frame war. Hehehe

Just silly. Accept, although I doubt I would ever watch it again. Lol

ktwo

Edit history:

ktwo: 2016-12-15 02:10:02 pm

2016-12-08 10:37:29 am

There is an in-game feature for resetting the game, which makes resetting very fast. I used this feature to quickly generate 20 consecutive boards. This is the number of times each location had a picture in it:
9 6 7
9 7 9
8 5 0
Based on the limited amount of attempts, I'm just going to assume the following:
- Other than the lower right location, each other position has the same probability of appearing on a board.
- The lower right location will never have a picture.
(- I'm also gonna make the assumption that boards 2 and 3 are generated similarly and follow the same distribution.)

I then timed in frames each appearance in the submission (start frame = the frame before the picture appears; end frame = the frame before the next picture appears/the frame before the end of board text appears):
335, 331 383, 384 -
266, 263 - -
265, 265 318 -
The execution is not frame-perfect, but it also has little to do with the end time, compared with the picture spawns.

We can also make a 'position ranking' based on these times:
5 6 X
1 4 X
1 3
(I timed myself the middle position to be ~10 frames faster than 'top left')

Ideally, you would get 'middle left', 'lower left' and 'lower middle' on three consecutive boards, which would have saved ~270 frames. So the next question is how hard it would be to go for a better setup? Based on the 'position ranking' above, the probability of getting all pictures of the three boards among the most favorable positions is the following:
6 (like in this submission): 4.5% (so 1/20)
5: 0.57% (so ~1/200)
4: 0.04% (so 1/2500)
3: 0.0006% (so ~1/150000)

If this submission had pushed the envolope a little further and only accepted any of the 5 most favorable positions, ~2 seconds are likely to have been saved (obviously still very much depending on the distribution of positions in such a run). Considering that the odds for such luck look manageable and how quick it is to reset, it doesn't strike me as unreasonable. Even if one assumes that the execution is not entirely trivial and that one or two of those lucky attempts could go down the drain because over-shooting the position or something. I do however understand that going for the best 3 or 4 positions would be more than what one could ask for.

In summary, it's not a bad submission. However, unless my assumptions above are wrong, I think the luck should be brute-forced a bit more before considering the run for posting here.

Reject

There is an in-game feature for resetting the game, which makes resetting very fast. I used this feature to quickly generate 20 consecutive boards. This is the number of times each location had a picture in it:
9 6 7
9 7 9
8 5 0
Based on the limited amount of attempts, I'm just going to assume the following:
- Other than the lower right location, each other position has the same probability of appearing on a board.
- The lower right location will never have a picture.
(- I'm also gonna make the assumption that boards 2 and 3 are generated similarly and follow the same distribution.)

I then timed in frames each appearance in the submission (start frame = the frame before the picture appears; end frame = the frame before the next picture appears/the frame before the end of board text appears):
335, 331    383, 384    -
266, 263    -                -
265, 265    318            -
The execution is not frame-perfect, but it also has little to do with the end time, compared with the picture spawns.

We can also make a 'position ranking' based on these times:
5 6 X
1 4 X
1 3
(I timed myself the middle position to be ~10 frames faster than 'top left')

Ideally, you would get 'middle left', 'lower left' and 'lower middle' on three consecutive boards, which would have saved ~270 frames. So the next question is how hard it would be to go for a better setup? Based on the 'position ranking' above, the probability of getting all pictures of the three boards among the most favorable positions is the following:
6 (like in this submission): 4.5% (so 1/20)
5: 0.57% (so ~1/200)
4: 0.04% (so 1/2500)
3: 0.0006% (so ~1/150000)

If this submission had pushed the envolope a little further and only accepted any of the 5 most favorable positions, ~2 seconds are likely to have been saved (obviously still very much depending on the distribution of positions in such a run). Considering that the odds for such luck look manageable and how quick it is to reset, it doesn't strike me as unreasonable. Even if one assumes that the execution is not entirely trivial and that one or two of those lucky attempts could go down the drain because over-shooting the position or something. I do however understand that going for the best 3 or 4 positions would be more than what one could ask for.

In summary, it's not a bad submission. However, unless my assumptions above are wrong, I think the luck should be brute-forced a bit more before considering the run for posting here.

Reject

mrprmiller

2016-12-08 11:24:41 am

Everyday is puppies and sunshine...

Your analysis is so indepth ktwo; I feel obligated to change my vote.

8BitSteve

Edit history:

8BitSteve: 2016-12-15 02:58:20 pm

8BitSteve: 2016-12-15 02:58:17 pm

2016-12-15 02:56:59 pm

www.8BitSteve.com

That is some very in depth analysis. I have a couple of things to add that hopefully can put it over the top. The first thing is that I have well over 500 puzzles so it is not a matter of just brute force. It seems that each set of puzzles are somehow linked to each other as you will often see the same patterns appear in all three puzzles. If this is true that means it may not even be possible to have pieces in the most favorable positions on all three puzzles.

As far as the pattern ranking listed above is concerned I have a few thoughts on that. Since the picture falls to a static position and I have to move horizontally to the right for every piece then the distance to each position should be measured from that spot. The spot would be between positions 4 and 7 on the following grid.
1 2 3
4 5 6
7 8 9

Assuming that the images themselves do not introduce any difference in speed due to shape and size then I propose the following board assignment.

2 3 4
1 2 3
1 2 3

The positions labeled 1's I am moving half a square to reach from the universal position. The squares labeled 2 I am moving 1.5 squares to reach. The squares labeled 3 I am moving 2.5 squares to reach and the top right square is 3.5 squares to reach.

As far as the reset is concerned it only applies after you complete all three puzzles. If you actually hit reset on the cart then you have to go through the entire intro again. I suppose I could attempt another 500+ times but It seems like a 1/2500 chance is in my opinion a poor way to reject a run.

That is some very in depth analysis. I have a couple of things to add that hopefully can put it over the top. The first thing is that I have well over 500 puzzles so it is not a matter of just brute force. It seems that each set of puzzles are somehow linked to each other as you will often see the same patterns appear in all three puzzles.  If this is true that means it may not even be possible to have pieces in the most favorable positions on all three puzzles.

As far as the pattern ranking listed above is concerned I have a few thoughts on that. Since the picture falls to a static position and I have to move horizontally to the right for every piece then the distance to each position should be measured from that spot. The spot would be between positions 4 and 7 on the following grid. 
1 2 3
4 5 6
7 8 9

Assuming that the images themselves do not introduce any difference in speed due to shape and size then I propose the following board assignment.

2 3 4
1 2 3
1 2 3

The positions labeled 1's I am moving half a square to reach from the universal position. The squares labeled 2 I am moving 1.5 squares to reach. The squares labeled 3 I am moving 2.5 squares to reach and the top right square is 3.5 squares to reach.

As far as the reset is concerned it only applies after you complete all three puzzles. If you actually hit reset on the cart then you have to go through the entire intro again. I suppose I could attempt another 500+ times but It seems like a 1/2500 chance is in my opinion a poor way to reject a run.

ktwo

2016-12-16 07:54:48 am

8BitSteve, thanks for the reply. I just played for 15 minutes in "attempt mode" on emulator, while recording (see attachment). Basically, if I didn't get the right locations immediately, I reset. The average time for an attempt including reset was ~3 seconds (I'm not sure I understand your reset description, but it sounds like you're resetting differently?)... So the number of attempts obviously racked up very quickly. I'm guessing I was well past the 100 mark (if anyone wants to count, be my guest). During those attempts, I had maybe 10 or so good locations in puzzle one (as expected), but then mostly got screwed over in the other puzzles (again as expected). I did however finish with an attempt with the following locations (fast forward to ~frame 58000):
0 2 0
1 2 0
2 2 -

Now, I didn't make much of an effort to execute correctly, but I think we can agree that the execution isn't overly complicated if one actually tries to do it well. The important thing is how hard it is to get good locations in a run. Now, the locations I got were ~1 second slower than your run. But this was after playing for 15 minutes. If I had spent just a few hours, which is virtually nothing in terms of putting in attempts for getting a good speedrun in most games, that would quite obviously eventually had to result in some very good location spawns.

As you can see in my previous post, I timed the different picture locations. As far as I can see, the question about the most favorable board assignments is therefore pretty much clarified already.

8BitSteve, thanks for the reply. I just played for 15 minutes in "attempt mode" on emulator, while recording (see attachment). Basically, if I didn't get the right locations immediately, I reset. The average time for an attempt including reset was ~3 seconds (I'm not sure I understand your reset description, but it sounds like you're resetting differently?)... So the number of attempts obviously racked up very quickly. I'm guessing I was well past the 100 mark (if anyone wants to count, be my guest). During those attempts, I had maybe 10 or so good locations in puzzle one (as expected), but then mostly got screwed over in the other puzzles (again as expected). I did however finish with an attempt with the following locations (fast forward to ~frame 58000):
0 2 0
1 2 0
2 2 -

Now, I didn't make much of an effort to execute correctly, but I think we can agree that the execution isn't overly complicated if one actually tries to do it well. The important thing is how hard it is to get good locations in a run. Now, the locations I got were ~1 second slower than your run. But this was after playing for 15 minutes. If I had spent just a few hours, which is virtually nothing in terms of putting in attempts for getting a good speedrun in most games, that would quite obviously eventually had to result in some very good location spawns.

As you can see in my previous post, I timed the different picture locations. As far as I can see, the question about the most favorable board assignments is therefore pretty much clarified already.

Attachment:

Perfect Fit (U)-0.fm2

8BitSteve

2016-12-17 12:38:07 pm

www.8BitSteve.com

I am not in disagreement that a faster puzzle combination is possible so lets get that out of the way. I just watched your movie and I had no idea about the in game reset feature and I must have glanced over when you mentioned it in an early post. So I am sorry for that. I will experiment more later with it when I get back to the studio. I am sure I can pummel out a lot more puzzles with that method if you feel that is necessary to accept my run. Like I said though I have over 8 hours of footage with approximately 550+ puzzles.

A few things to address:

1. Does the shape of the images actually introduce any delay when entering the same position?

2. The frame breakdown is a bit confusing to me. Are those the frames it actually takes to complete the image or are those the frame counts from my run?

Thanks

5 6 x
1 4 x
1 3 x

ktwo

2016-12-18 06:47:56 am

Sorry to hear that you spent a lot more time than necessary on this since you didn't know about the fast-reset feature. However, if you would like a run of this game to be accepted here, you should take advantage of this feature. I think as things stand, your run is definitely respectable, but could also be beaten in a fairly short time if you or anyone else were interested in investing more effort into this game.

1. I haven't really looked at this, but I'm almost certain they're all equivalent. And if you find evidence of any images being faster/slower than the others, I'm sure it's negligible compared to the time difference of getting different locations on the board.

2. What I did was to time how long it took to complete an image based on the location, in order to find the board locations that resulted in the fastest completions. I tried to describe the start and end points, but they're actually irrelevant, as long as they're the same when comparing between the different positions. For example, I wrote that the bottom left position took 265 frames, while the bottom middle took 318 frames. Those numbers are only important when compared with each other (bottom left is 53 frames faster than bottom middle). I could have chosen lots of other different start and end points, but the difference would always have been 53 frames.

Sorry to hear that you spent a lot more time than necessary on this since you didn't know about the fast-reset feature. However, if you would like a run of this game to be accepted here, you should take advantage of this feature. I think as things stand, your run is definitely respectable, but could also be beaten in a fairly short time if you or anyone else were interested in investing more effort into this game.

1. I haven't really looked at this, but I'm almost certain they're all equivalent. And if you find evidence of any images being faster/slower than the others, I'm sure it's negligible compared to the time difference of getting different locations on the board.

2. What I did was to time how long it took to complete an image based on the location, in order to find the board locations that resulted in the fastest completions. I tried to describe the start and end points, but they're actually irrelevant, as long as they're the same when comparing between the different positions. For example, I wrote that the bottom left position took 265 frames, while the bottom middle took 318 frames. Those numbers are only important when compared with each other (bottom left is 53 frames faster than bottom middle). I could have chosen lots of other different start and end points, but the difference would always have been 53 frames.

SDAVerification

2016-12-20 08:54:59 am

Decision posted.