SMS Any% Timing Model (2023) (PAL NF BSNR)

	A	B
1	SMS Any% Timing Model (2023) (PAL, New File, BSNR Route)

2
3	The goal of the timing model is to reconstruct an RTA-timed Any% run from ILs (of Delfino movement and level segments). The fungible parts of the run are estimated using fixed quantities (e.g. an event cutscene = 91f) or averages, after being bucketed into alike groups (e.g. a Ricco world load ≈ 2.779s).
4
5	The model is specified in 4 parts:
6	ILs	The gameplay segments, which can be estimated for a player
7	Loads	The loads, which are estimated with console-dependent averages
8	Params	Constants that are universal or depend on the game version
9	Deltas	Small adjustments to fixed segments, marked ∆ (e.g. menuing timeloss)
10
11	This spreadsheet has the main model, derived from a 1:19:48 sample Any% run and supplementary data.
12	The following user-facing sheets are derived from this sheet:
13	Converter	A calculator that estimates ILs from RTA segments and vice versa
14	Load Sheets	For recording load retimes from an Any% run
15	IL Sheets	For recording IL retimes from an Any% run
16
17	This spreadsheet is comprised of the following sheets: (in loosely reverse order, bottom to top)
18		Analysis
19	Retime	The raw measurements for each segment of the sample run (the 1:19:48)
20	RetimeExtra	Supplementary raw measurements
21	text Model	An independent timing of optimal text (version-dependent), done via TAS
22	fmv Model	Analysing FMV timings using the sample run and FFMPEG frame dumps of the FMVs
23	d Model	Analysing Delfino segment outro timings from the sample run
24	li Model	Analysing Level-Intermediate segments and combined IL retiming with a QFT/SGT TAS
25	lt Model	Analysing Level(-Terminal) outro timings from the sample run
26	Loads	A Load Sheet for the sample run, and derived load distribution
27	ILs	An IL Sheet for the sample run
28	Params	A datasheet of constants derived from the preceding models
29		Synthesis
30	MapSegP	Reconstructed segment IL→RTA adjustments from the preceding models' estimates
31	MapSegA	Segment IL→RTA adjustments direct from the original data, and comparison to the model
32	MapLevelP	Reconstructed level IL→RTA adjustments from the model segment estimates + adjustments
33	MapLevelA	Level IL→RTA conversions direct from the original data, and comparison to the model
34	Converter	A Converter comparing the model predictions with the actual RTA splits for the sample run
35
36	Time formatting usage notes:
37	Times are formatted "s.000" (for small values) or "[<0.00069444]s.000;[>=0.00069444]m:ss.000" (large)
38	In both of these formats, small negative values appear as differences from 1 minute, e.g. 59.800 for -0.200
39	To type times, you must include leading zeros for hrs and mins; e.g. 0:1:2 gives 1:02.000; 0:0:1.23 gives 1.23
40	To convert decimal times in secs into time formats (as above), divide by (246060) and apply the format
41	Arithmetic between time durations will work normally since times are stored as floats (1 = 1 day)
42
43	Retiming notes:
44	The video retiming practice used on the sample run and recommended for the IL and Load Sheets is to open a local video file with Avidemux (v2.8.1 here) and copy timestamps from matched frames into the sheets. Avidemux is used because it natively handles timestamp-based video and so should give consistent readings for collaborative retiming for theoretical reasons, though it hasn't been tested much in practice.