infjamc Lv 1
- While I'm not against the intention behind all-hands puzzles, let's not forget the important observation from the Foldit developers on this type of puzzle: In practice, people tend to converge toward evolving the top-scoring solution rather than exploring.
==> So, as a compromise, I would be in favor of a two-round system where the top 5-10 unique solutions from Round 1 (where "unique" is defined as two solutions having an RMSD above an arbitrary cutoff) are automatically provided as templates for Round 2.
.
- Regarding the idea of individuals not getting credit for evolving: my concern is that this could decrease the incentive for evolver play. But the following system might work:
a) Record the score gain achieved by each player above a certain threshold (say, the media score of the puzzle).
b) The player on each team that contributed the most to evolver player will receive the same number of individual evolver points as the team global points for this puzzle. Everyone else on the team would receive a reduced amount that is scaled down proportionally.
For example: A team is placed 2nd in a puzzle and receives 75 points. The median score of the puzzle is 10000 points, and the team's top evolver solution scores 11000. Suppose that one player contributed to 400 points of the 1000-point score gain, while three other players contributed to 300, 200, and 100 each. The four players will receive 75, 57 (56.25 rounded up), 38 (37.5 rounded up), and 19 (18.75 rounded up) points, respectively, in evolver global points.
Now, this system does have the downside of giving too much credit for the lower score ranges, where evolution is easier. A possible fix might be reweighting the score gains by the range where the score gain has occurred. For a more complicated example:
(Starting configuration = 8000, median score = 10000, team evolver score = 11000)
Player A: 8000 -> 10400 (note that only the last 400 points would count)
Player B: 10400 -> 10450
Player C: 10450 -> 10500
Player D: 10500 -> 10600
Player B: 10600 -> 10850
Player C: 10850 -> 11000
Now suppose that relative contribution is reweighted based on the function (credit) = 2 * (location_of_score_gain - median_score) / (team_evolver_score - median_score). Integrating this function yields x^2/1000, which is then applied for each segment of score gain
Player A: x = 0 to x = 400, relative contribution = 160
Player B: x = 400 to x = 450 and x = 600 to x = 850: relative contribution = 405
Player C: x = 450 to x = 500 and x = 850 to x = 1000: relative contribution = 325
Player D: x = 500 to x = 600: relative contribution = 110
So Player B actually gets the most credit despite contributing to a smaller number of points from an absolute standpoint… because most of the points occurred in a higher range, which is more difficult. Now Player B gets the full 75 global evolver points for his/her evolver play, while Players A, C, and D receive 30, 61, and 21 points, respectively. (As you can see, the main downside of this system is sheer complexity, as the scoring function would have to be integrated for each team.)
