SportWagon
I tried a variety of Miatas around Nurburgring, but nothing new popped up.
Do you have any idea what the granularity of the skill points awarded is? That is, what is the smallest increment you can get or not get? (A question would be, how likely is it that the two points are spread out in tiny fractions across many tracks?)
I would suspect the points are tracked in a bitmap, therefore as whole numbers, and scaled to give the 10000 point result. Or do you have a different hypothesis?
Your missing points are probably track battle points which influence battle skill and course skill at the same time. Also, my experience was that 2 missing points are frequently an indication that one race on one track wasn't finished quite to the standard expected for all battle points to be awarded. The most like canidates for a missing track points are;
High Speed Ring (maybe it was Super Speedway) with the 350Z - that was a stuborn combination for me, no other data.
Grand Valley East reverse with the Miata - the JP authors missing point.
Short Courses - I repeated any race less that 2 minute long as insurance against missing points.
During B-spec test games, careful records are kept of the points after each race so it is possible to deduce where the points were lost by examining certain milestones where the fractions are 0 or small enough to detect a dropped fraction.
Even if you don't have careful records, it is still useful to check the points between each race. Assuming all floating points have been collected (this is why I encourage maxing the floating points on the Ring so strongly) then;
1. On new tracks, 13 points will be collected (sometimes 12 or 14, depending on the fractions.
2. Tracks without 8 or 9 course points might still have 4 or 5 battle points because of a mismatched race.
3. Sometimes 11 points are collected because the race wasn't perfect enough for that track. But if the course points were present, then there must still be a few battle points on the track. Repeat the race to avoid problems like #4.
4. If 2 points are collected, then maybe these are points from a previous race like #3. Or maybe the course points were previously collected as in #2, and now only 2 points were earned, which still leaves 2 points on the track similar to #3.
I know that points are divided into very small parts. Based on an early analogy relating the collection of B-spec points to hippie beads on a string, I'll refer to these as B-spec Beads. B-spec beads can be no bigger than 1/29th of a B-spec point. In other words, collect 29 level beads to earn 1 level point. Fortunally, level beads are very had to detect and easy to ignore in favor of fractions.
I was very sceptical about some of the JP authors numbers and theories when I started my test. However, my data confirms almost everything he stated. The main exception is that I use fractions, and he used the decimal equivilant (I think, I don't read Japanese). Decimals lead to huge errors in predictions quickly. The JP author identified floating points (7000) and track points (3000). I counted the 4 current classes for points separately (F-Machine, F-Battle, T-Course, and T-Battle).
The best evidence of the fractions is the point progressions of my data. I'll refer you to an previous post for details.
https://www.gtplanet.net/forum/showpost.php?p=1645868&postcount=77
My first clue was when I observed a skill change without a point change. My only explanation was that fractions must be possible. The reason 29ths of a point are required is because the 3000 track points are divided evenly on 58 tracks. The points have been counted, I can find no other explanation for the data.
Math Review
58/2=29
58*4=232
Track:3000/232= 12 and 27/29
Course:2000/232= 8 and 18/29
T-Battle:1000/232= 4 and 9/29
It's hard to predict how likely these little peices are to be spread over many tracks. If points are counted systematically, then a status check can be made every 29 tracks. It's difficult to tell if tiny fractions are lost in between. Lets see..., # of people with 10000 points (2) / # of people who tried (a few dozen) = very poor statistics. I believe there is a huge selection and reporting bias in the data. I suspect the beads are bitmapped, and therefore there are at least 290,000 B-spec beads. We need a map.
