I agree that the shorter window and much less points for the starter puzzles is the answer.
Newbies get to keep big scores from all 3 versions of the same puzzles so they jet up through the ranks unfairly, then they complain that they plummet when all those easy puzzle scores fall off at the 4 month window.
It appears that Tealight doesn't understand the scoring system at all.
As far as I understand the scoring it goes like this:
There are a certain number of puzzles in any time window, so newer players are at a disadvantage until the four months is calculated, ie when they have competed in the same number of puzzles as everyone else, at that point if they have been playing well (and usually you also need to play every puzzle) their global score will go up. Whis is why it helps to have the restricted beginner puzzles.
The image shows how it works - the four players start a month after each other
Maximum number of points available at each 4 month recalculation is shown
That way if you have played the whole four months window you can be rank one.
As rav says if points were cumulative since 2008 no new people would be in the higher ranks.
Ranks are also calculated for each puzzle within the window, which means you can go up and down within the four months which keeps it interesting, your total at the end of four months is I think what you start the next four months with.
imho if <15 and <150 players are playing the same puzzles they should get the same points, it doesn't make the puzzle any easier.
actually B2 I do understand it..
You don't start the next four months with the previous score - blonde moment :)
The reality is, it isnt working like you say.
The whole issue of sudden drops can also be avoided by exponential scoring. Everyday everybody loses 1% of their global points. Any given puzzle loses half its value in 10 weeks. If you're consistent, your score levels out. There's still a max score for all players, regardless of how long they've been playing, that new players can approach in a reasonable time. There's just no sharp cutoff. If people are stressed by seeing their points disappear, it could be recalculated only when puzzles close.
Nononono, this is terrible.
It is much easier to avoid: do NOT play all three (<15, <150 and normal one) puzzles when you can :)
it is not just about sudden drops it is also about sudden rises for the players who did well.
Since we're in the mood for brainstorming, here's another idea I would like to suggest:
Why not consider tweaking the "Points = Max(1, RoundUp( 1 - (Rank - 1)/(NumPlayers - 1) )^7 ) * X" formula instead? To see why, let's plug in a few numbers:
==> Obviously, the distribution of scores are right-skewed rather than Gaussian, but the trend is clear: unless a player can consistently do very well in every puzzle, it is very likely that having a handful of top performances mixed with average performances would score better than a consistent performance that's only slightly above average. For example, a 97.8 percentile finish in one puzzle combined with a 50 percentile finish in ten puzzles would give you more global points than a 69.1 percentile finish (slightly above average) in the same eleven puzzles.
.
Why is this relevant? Because newbies are much more likely to score high in a <15 or <150 puzzle than in a "standard" puzzle available to everyone. Take Puzzle 506 as an example:
So imagine the shock for the new player who graduates from the <15 and <150 puzzles:
80+31 = 111, so the new player continues playing in both the <150 and regular versions of a third puzzle for another 31 points (subtotal: 142). Let's say that several more CASP ROLL puzzles show up, and the new player gets 2 points in a few of those to get to 150.
==> But let's see what happens if we lower the exponent in the equation used for calculating global points. For example, what happens if we use 5 instead of 7?
Notice that the player would be earning more than the minimum for an average performance, and a +2.0 std. dev. performance is only worth 5.7 rather than 10.8 puzzles at the +0.5 std. dev. level. Essentially, the system is still rewarding outstanding results while the slightly above-average ones are now worth a little more. To revisit the earlier example, a 63th percentile finish would be worth 10 points per puzzle under this new system, and 10x40 would mean keeping 400 points when the <15 and <150 scores expire. That's good for #208, which looks twice as better than before.