I remember that the discovery of the Japanese SMK community (or at least a representation of it) in 2004 was quite a big deal for all of us around back then. Their way of using and thinking about NBT (they actually were New Boost Techniques back then haha) was quite a bit ahead of us at the time, it seemed. I also recall that the top Japanese players at the time, Monchuck and donnki (nicknames, no doubt) created a page/map for every track where they detailed the best boost spots and trajectories. And on top of that they forecasted which times they expected to be possible when applying these strats. Of course I don't know if these times were meant to represent theoretical limits (TAS based), human limits or perhaps even goal times  , but needless to say that part of it really interested me, whatever was the case. The only times I really remember them predicting were sub 56 and sub 11 on MC1 at a time when our site records were still sitting at high 57 and 11"2x. I figured these were based on the famous 55”95 – 10”99 TAS at the time.

Anyway, I tried to dig up these pages again to see what their predictions were (their MB is still linked on the Players' Site), but sadly the track specific links are now dead. So I went the Waybackmachine route…and that archive is awesome as I succeeded in finding them that way!  

I have snapshotted all of the track pages here, some haven’t been captured properly and sadly it appears MC3 hasn’t been snapshotted at all. The pages can all be seen here: https://web.archive.org/web/20041102104419/http://f31.aaacafe.ne.jp/~smkt/mc1...
For other tracks, replace the track code accordingly.
As for the boxes above the tracks with the star rating, I believe the first one denotes difficulty of track/strat (0 to 5 stars….KB1 being 5*, jeez they defo got that right) and the second one denotes the level of the WR (at that time). They were aware of our WRs early on btw, that much is clear from reading their MB.

Personally I think that they had excellent vision for the time (for some tracks in particular: MC1, GV1, BC2, VL1, VL2), but of course at others they were way off and didn't foresee all the strategy changes that were ahead. I also remember their CI2 predictions to be much more bold (like 12”00 flap or something), but that must have been a later uncaptured update. In any case make sure to check out GV1, they were right on the money for that one and then some!   That one made me chuckle quite a bit.  

So, here we go:

MC1

Older shot with star ratings, but slower times: http://i.imgur.com/bczr83c.png

DP1


GV1

11"65 baby!  

BC1


MC2

Older shot with star ratings, but slower times: http://i.imgur.com/2kIieiu.png

CI1


GV2


DP2


BC2


MC3
Nope

KB1


CI2


VL1


BC3


MC4


DP3


KB2


GV3


VL2


RR

Many thanks for your huge work there Karel, all this is literally fascinating! Some of the predictions are so close to reality (not to mention GV1 which is just INSANE), what a mindfuck.

GV1 lap is fake anyways. 11"65 never happened

Yeah VL2 5-lap is just a more daring estimate than most others, I guess they also recognized how this one is harder to max out. Of course there is also a bias in which track they personally tested most; VL2 may have been one of their favourites, who knows. GV3 flap is defo a typo though.

I definitely notice such bias in the predictions we made in the past too, they tend to be more accurate for tracks that we already overplayed.

This is a French forum topic in which we predicted (PAL) WRs (not limits), also from 2004 onwards.  
http://www.ffsmk.org/forum/index.php?topic=707.0

MC2 is quite impressive too as they already figured out the LB and basically the current strat being used. the 1'01"50 mark had been beaten only like 1 year ago.