Guess The Price - Season 48 - Winner: Amida17!

Congratulations amida! You both were closer than I would have been.




Loss by a pickle...I will NEVER forget that one!

I have a submission for you, can I send the pics in a day or so?

Partly because I was curious and partly because I had too much time, I decided to test the latest revisions of the tiebreaker rules on any ties in the last 30 or so seasons just to see what would have happened. Unless I am missing some, I believe those were S26, S27, S36, and S48. For those who are curious, S27 would have won by capitanfwiffo(sorry Coinskelly that I couldn't get the numbers in your favor) by rule 1 with his 2 wins; S36 would have been won by coinaholic with his 2 wins; and of course the latest season by amida17 with his 2 wins.

S26 is where it gets a little more tricky and where I found a potential problem with rule 4. Both capitanfwiffo and earle42 had a win, 2 runner ups, and a coin submitted. If rule 4 is finalized as average percentage off the actual price, capitanfwiffo would have won by 32.09% to earle42's 388.80%. Earle42's percentage was highly skewed by the R8 which was low in price combined with his relatively high guess while capitanfwiffo was left unaffected because it was his submission.

I really hate having to complicate things and creating more work for jbuck since he already has enough on his plate but I kind of feel a slight modification might be needed. I was thinking, after all the percentages are calculated, that the highest and lowest numbers be thrown out. This should help to reduce the affect of outliers. It would be similar to how they do it in the olympics. With this modification, capitanfwiffo would have 28.88% to earle42's 28.46% thus giving earle42 a very narrow victory. Another possiblity, would be to use the median instead of the mean. In this scenario, capitanfwiffo would have 35.38% to earle42's 22.22% with the end result the same. A possible problem with the median could be that a few really poor or good guesses will not be reflected in the score and I kind of feel like they should to gauge the overall accuracy of each player's guesses. I am slightly in favor of the modified average mentioned earlier over the median but will support whatever jbuck decides for rule 4 even if it is something else that I have not mentioned. I wonder if anyone else has ideas for the tiebreaker rules.

I went back to about S19 because I think I remember reading something that the current rules regarding points awarded for a submission began then. I did try applying the proposed rules to S11 but I have not gotten the numbers to work out so I can't say what the results would have been. While working on S11, I did take a glance at S10 and that one seems to have also ended with a tiebreaker.

I have been pondering about how to mitigate the effects of one or more low price coin on the calculations. Earle42 was 3300% on that R8 coin. Let's say the R8 coin was $0.01 instead of $1, earle42's guess was $3 instead of $34, and his guess did not earn any extra points. His guess would balloon to 29900% off but otherwise doesn't look all that bad. Now if we say that $0.01 coin was submitted by someone other than captainfwiffo or earle42 and the guesses were $5 and $3 respectively, captain would be be heavily penalized compared to earle even though both guesses don't seem all that far apart.

Upon your suggestion, I did the calculations with both of those hosting rounds thrown out and earle would have still won 24.90% to 33.61%. I also tried throwing out the 2 lowest and highest percentages and still earle wins 24.71% to 29.01%. At least it is nice to see that the median and 3 conditional means all agree upon a winner.



That's for sure. They also reduce the probability of needing to use rule 4.

I am all for making the calculations as simple as possible so that it would be understandable to most players. There could be another method out there that could do a better job but it likely will involve more advanced math and statistics and I don't think anyone wants that. The median would probably be the simplest and would not suffer from outliers created by low priced coins. While I feel this summary statistic should take into account the numerical value, I think the mean can only be used if there are some conditions placed on which numbers will be used to calculate it.

Whatever the calculations be, it should be simple and equitable. We do have a few days to think about it before next season. If you would like, I could email the spreadsheet that I used to calculate the numbers to you or to anyone else.

Wow - I missed all the tie-breaker action and subsequent discussion while I was away on Christmas vacation.

Congratulations to amida17 - Season 48 champion!

Well played indeed by D0ubl3Eagle ... your outstanding track record of success in GTP speaks for itself.

Regarding the proposed tie-breaker rules for the Gold - I'm agreeable to the new approach assuming it does not significantly add to jbuck burden hosting this best of all coin games.

Happy holidays to all - see you in S49.

David