You are here
Cheating potential
Forums:
After checking Preliminary results WPF GP Sudoku 3, I bet two competitors have cheated during the round. The 37th in the result earned 656 points in only 21 minutes, which is faster than 30 pts/min, and the 457th in the result earned 208 points in only 3 minutes, which is equivalent to nearly 70 pts/min! Is this solving speed reachable? It is not possible even for top solvers like Tantan Dai and Ken Endo!
There should be more unusual types
Because usual types are easier for computer solve, if a round mainly consists of usual types, then it will be easy for some competitors to cheat with computers. Therefore, to reduce the possibility of cheating, a round should contain a portion of unusual (not necessarily innovative) types, like Emitters Sudoku and Borderland Sudoku in Sudoku GP3, and usually they should be high pointers.
The most reliable cheater
The most reliable cheater detection is simply name recognition - generally the top solvers are already well known and in the vast majority of cases have had their general level verified via performances at offline national or international competitions, where cheating is harder to do (though not totally unheard of).
I feel competitions that start skewing to more unusual types will be on average less enjoyable for the average sudoku solver, and I would be sorry to see this.
There are some other basic things you can do (like checking for repeat IP address) to detect cheating, alongside benchmarking performance to known good solvers, but ultimately for an online competition there's not that much you can do. Maybe you could do more with a computer interface where you can record the order of which number gets placed into which cell - but again this takes something away from the competition. Many solvers still prefer to solve on paper rather than with computer interfaces.