Subject: AllTime scores

Date: Tue Oct 2 08:36:24 2018
User: HopDiriDiriDattiriDittiriDom
Message:
Recently I have been thinking about a combined score for 12-sum variants based on the all time rankings. AllTime pages have 1000 entries. So it would be reasonable to have something like 1000/your_standing: First one: 1000/1=1000 Second one: 1000/2=500 Third one: 1000/3=333.34 There is a big gap for the first entries there. So on a second thought it would be reasonable to scale them down to 10 say using cubic root: First one: cubic_root(1000/1)=10 Second one: cubic_root(1000/2)=7.93 Third one: cubic_root(1000/3)=6.93 Additionally I have thought about including average time as a factor such that fast players get better scores and came up with 1000 / average_time in seconds later I thought 1000 seconds is too long and made it 600 (i.e. 10 minutes) So now we should have score like cubic_root(1000/your_standing) times 600/your_average_time_in_seconds. Then I thought I need to take into account how hard it is to climb to the first position. For example to be number one in 8x4 all time table it takes years. I have become number one in 7x5 table in a couple of months. For 4x8 it would be a matter of days if you are lucky. So what I have come up with is to include a factor like natural_log(your_streak_length). So now we should have score like cubic_root(1000/your_standing) times 600/your_average_time_in_seconds times ln(your_streak_length) So how would it look like? Let's calculate it for the first 4 entries in the 6x6 AllTime table: ? sqrtn(1000/1,3)*600/497*log(290) %7 = 68.44926667582116351550821111 /* flobbadob */ ? sqrtn(1000/2,3)*600/129*log(193) %8 = 194.2790684197752962959976110 /* CubicSprock */ ? sqrtn(1000/3,3)*600/372*log(182) %9 = 58.19768884323857272320493176 /* wolper */ ? sqrtn(1000/4,3)*600/217*log(146) %10 = 86.80577263252270848628450035 /* myself */ So we can create a standings table using this scoring system for 12-sum variants. I believe sum of these scores would be reasonable. I would like to have your opinions on this.

Date: Tue Oct 2 10:51:05 2018
User: joeygray
Message:
Until Denny finds a way to measure actual time spent on the deal - and I'm being facetious here - I am not much interested in ratings based even in part on elapsed time. 1) I routinely start games with no desire or expectation of finishing them. For example, just last night I was watching Broncos vs Chiefs American football game, and playing at the same time. I might watch football for 10 minutes, then return to my screen at the commercial and finish a game I started during the last commercial. Or like yesterday when the furnace repairman came in the middle of a game, and I returned to the game hours later. I have 20 years of results like that, which I could not overcome even if I was inclined to change my style. 2) It's too easy to skew time the other way, by taking a screen shot and then saving, and studying the game offline. But more importantly, I don't like the way I play when time matters. I can't tell you the number of times when we were doing the Saturday competition that I lost games toward the end of the day, even in the easiest of variants, because I 'just gotta get ten more wins before midnight.' I also admit that I've lost plenty of level 5 thru 9 games 'just whipping thru those before 50' so THEN I 'can really start thinking.' Just lost one of those today, sadly. I've played tournaments before, of course, in which how fast you can lose is sometimes as important as how fast you can win. Its a different game, in some sense.

Date: Tue Oct 2 11:04:37 2018
User: TNmountainman
Message:
Completely agree with joey. I also essentially care not about time/game. I'm not a fast player, nor have ever tried to be. Sadly, tournament rankings went away before I could ever dip my toe into that water. (I was using a slow, ancient system until right at the time that happened.) Since we know "cheating" takes place, in all it's guises, this would just give more incentive for those who allow themselves to cheat, to do so.

Date: Tue Oct 2 11:33:08 2018
User: HopDiriDiriDattiriDittiriDom
Message:
Thank you for your input gentlemen. How about if we strip the average time factor? Does the rest make sense?

Date: Tue Oct 2 12:42:36 2018
User: jamesblackburn-lynch
Message:
Two thoughts. First, why the log? I gather you want to reward people with high rankings in flavors where it takes a long time to amass such a large score but at the same time you don't want the streak length number to dominate the product. But why the log? Why not stick with roots? Second, because this is All Time streaks these numbers won't change very much. Are you thinking of calculating these numbers once a year? Certainly weekly would be pretty uninteresting. James

Date: Tue Oct 2 19:28:06 2018
User: TitanicTony
Message:
I don't know if it is possible to copy-n-Paste a table; I'm about to find out. This originally appeared in one of Joey's 12-sum competition threads. Date: Sat May 13 11:00:27 2017 User: TitanicTony Message: Just for fun (and about 10+ hours of work), I have decided to post a Table like Joey's based on all-time best streaks (as of 12 May 2017). I had to lookup a lot of streaks and enter them manually, so there are almost certainly a few errors! Also, there may be a few players that got left out (sorry). I am including the top 74 players, with scores > 500,000: At the bottom of the table is the sum of the streaks for each variant, which is a rough measure of each variant's difficulty and/or popularity. Ra- nks 4x8 5x7 6x6 7x5 8x4 9x3 10x2 11x1 12x0 Sum Avg. Score Players: == ==== ==== ==== ==== ==== ==== ==== ==== ==== ==== ===== ========= = ================= 1 4 14 85 164 334 1004 1344 308 51 3308 367.6 20,265,544 ValpoJim 2 5 34 59 204 483 267 461 271 55 1839 204.3 19,915,354 Goosey_Goosey_Gander 3 6 37 50 234 561 297 383 128 36 1732 192.4 16,945,933 CubicSprock 4 5 24 61 110 360 329 227 130 47 1293 143.7 11,844,941 joeygray 5 9 27 67 78 269 256 132 157 33 1028 114.2 11,137,731 Kaos 6 5 23 38 113 328 247 208 111 66 1139 126.6 10,130,315 TNmountainman 7 1 4 61 111 1547 1556 1920 248 44 5492 610.2 9,009,472 bwi 8 6 13 45 96 402 179 187 150 44 1122 124.7 8,466,022 firenze 9 8 21 47 63 219 186 199 122 38 903 100.3 8,465,133 Mastermind 10 4 16 38 181 211 408 190 127 43 1218 135.3 8,453,276 Darkosi Nope, that's pretty awful. And, the table was truncated on purpose.

Date: Wed Oct 3 02:36:11 2018
User: HopDiriDiriDattiriDittiriDom
Message:
James, Why natural log? The table is a statistical data where using exp() and its reverse function ln() makes sense. Let's check it out (ln vs sqrt): 8x4 top spot (PudongPete): ? sqrtn(1000/1,3)*log(20011) %1 = 99.04037401341563512321107507 6x6 top spot (flobbadob): ? sqrtn(1000/1,3)*log(290) %2 = 56.69880922980519711201263487 same scores with sqrt would be: ? sqrtn(1000/1,3)*sqrt(20011) %3 = 1414.602417642497880938936807 ? sqrtn(1000/1,3)*sqrt(290) %4 = 170.2938636592640116613332182 There's a huge difference IMO. PudongPete probably would lead with his 8x4 score only, if we sum them. Products may be different but I am for sums because products quickly get bigger and bigger quickly. As for the other question, yes once or twice a year would be fine.

Date: Wed Oct 3 02:45:32 2018
User: HopDiriDiriDattiriDittiriDom
Message:
TT, I probably would be at top of that table now because my current score is: ? prod(i=1,9,log([12,35,146,767,1439,850,850,195,43][i]+2)/log(2)) %7 = 57058952.16102640191339996851 And j-b-r is not there. His current score seems to be: ? prod(i=1,9,log([7,15,57,181,5874,394,1070,297,49][i]+2)/log(2)) %8 = 29064421.79947869886046979349

Date: Wed Oct 3 04:10:56 2018
User: TitanicTony
Message:
Yes, that table was done more than 16 months ago! I looked last night, and couldn't find my original; it must be here somewhere!?? I agree, the table would be quite different now! If I can find my original, I might consider to redo it.