I did a 3 second think early this AM came up with 10 wins -2 losses , a plus 8 and I can win . If you has asked me in the very beginning - I probably would have said 10-8 in favor of K - but it's obvious , after twenty games, that Komodo has made more progress than Stockfish and Stockfish is at least 50 points stronger than SF 6 if there are no bugs - but right now I'm not 100% confident that there are no bugs. Remains to be seen. Komodo looks very very solid here. These are not humans, there will be no 3 straight wins by SF . I am hoping that SF scores at least of couple of wins, K is on pace to win every 10 games or so. I just can't believe SF will get shutout and I hope it doesn't.Ajedrecista wrote:Hello Jesse:
Once the deadline was reached, I write my particular forecast after 24 games are played (+2 =22 -0 for Komodo).
I assigned probabilities in the following way: from Komodo POV, if there are w wins, d draws and l loses, the probabilities of win, draw or lose in the next game are prob_X = (X + 1)/(w + d + l + 3), where X = {w, d, l}. I do not remember the name of this well-known kind of assignation of probabilities, so someone write the name, please. Here is the code:
After 1e+7 simulations:Code: Select all
program TCEC_forecast implicit none integer,parameter :: simul = 10000000 ! 1e+7 simulations. integer :: i, j, K, SF, d, n(0:200) real(KIND=2) :: probK, probSF, PRN, t0, t1 real :: points_K(1:simul) t0 = cpu_clock@() ! Start. do i = 1,simul K = 2; SF = 0; d = 22 ! The result after the first 24 games. do j = 25, 100 probK = (K + 1d0)/(K + SF + d + 3d0) ! +3 due to {win, draw, lose}. probSF = (SF + 1d0)/(K + SF+ d + 3d0) ! +3 due to {win, draw, lose}. PRN = random@() ! PRN: pseudo-random number. if (PRN < probK) then K = K + 1 ! Komodo win. else if (PRN > (1d0 - probSF)) then SF = SF + 1 ! SF win. else d = d + 1 ! Draw. end if end do points_K(i) = K + 0.5*d end do n = 0 do i = 1, simul do j = 26, 178 ! Minimum of Komodo: 13 points = 26 half points; maximum of Komodo: 89 points = 178 half points. if (j == 2.0*points_K(i)) then n(j) =n(j) + 1 exit end if end do end do write(*,'(I8,A)') simul, ' simulations.' write(*,*) write(*,'(A)') ' K - SF Simulations' write(*,*) do i = 26, 178 write(*,'(F4.1,A,F4.1,A,I8,A,F6.2,A)') 0.5*i, ' - ', 100.0 - 0.5*i, ' ', n(i), ' (', 1d2*n(i)/simul, ' %)' end do write(*,*) t1 = cpu_clock@() ! Finish. write(*,'(A,F7.2,A)') 'Elapsed time: ',(t1-t0)/3d9, ' seconds.' ! 3 GHz in my PC. end program TCEC_forecast
The mode of that discrete distribution is 53-47 in favour of Komodo, so this is my bet with no possible reward given the fact that I did not answer before the deadline.Code: Select all
10000000 simulations. K - SF Simulations 13.0 - 87.0 0 ( 0.00 %) 13.5 - 86.5 0 ( 0.00 %) 14.0 - 86.0 0 ( 0.00 %) 14.5 - 85.5 0 ( 0.00 %) 15.0 - 85.0 0 ( 0.00 %) 15.5 - 84.5 0 ( 0.00 %) 16.0 - 84.0 0 ( 0.00 %) 16.5 - 83.5 0 ( 0.00 %) 17.0 - 83.0 0 ( 0.00 %) 17.5 - 82.5 0 ( 0.00 %) 18.0 - 82.0 0 ( 0.00 %) 18.5 - 81.5 0 ( 0.00 %) 19.0 - 81.0 0 ( 0.00 %) 19.5 - 80.5 0 ( 0.00 %) 20.0 - 80.0 0 ( 0.00 %) 20.5 - 79.5 0 ( 0.00 %) 21.0 - 79.0 0 ( 0.00 %) 21.5 - 78.5 0 ( 0.00 %) 22.0 - 78.0 0 ( 0.00 %) 22.5 - 77.5 0 ( 0.00 %) 23.0 - 77.0 0 ( 0.00 %) 23.5 - 76.5 0 ( 0.00 %) 24.0 - 76.0 0 ( 0.00 %) 24.5 - 75.5 0 ( 0.00 %) 25.0 - 75.0 0 ( 0.00 %) 25.5 - 74.5 0 ( 0.00 %) 26.0 - 74.0 0 ( 0.00 %) 26.5 - 73.5 0 ( 0.00 %) 27.0 - 73.0 0 ( 0.00 %) 27.5 - 72.5 0 ( 0.00 %) 28.0 - 72.0 0 ( 0.00 %) 28.5 - 71.5 0 ( 0.00 %) 29.0 - 71.0 0 ( 0.00 %) 29.5 - 70.5 0 ( 0.00 %) 30.0 - 70.0 0 ( 0.00 %) 30.5 - 69.5 0 ( 0.00 %) 31.0 - 69.0 0 ( 0.00 %) 31.5 - 68.5 0 ( 0.00 %) 32.0 - 68.0 1 ( 0.00 %) 32.5 - 67.5 1 ( 0.00 %) 33.0 - 67.0 1 ( 0.00 %) 33.5 - 66.5 0 ( 0.00 %) 34.0 - 66.0 4 ( 0.00 %) 34.5 - 65.5 4 ( 0.00 %) 35.0 - 65.0 7 ( 0.00 %) 35.5 - 64.5 13 ( 0.00 %) 36.0 - 64.0 12 ( 0.00 %) 36.5 - 63.5 19 ( 0.00 %) 37.0 - 63.0 45 ( 0.00 %) 37.5 - 62.5 49 ( 0.00 %) 38.0 - 62.0 101 ( 0.00 %) 38.5 - 61.5 120 ( 0.00 %) 39.0 - 61.0 188 ( 0.00 %) 39.5 - 60.5 295 ( 0.00 %) 40.0 - 60.0 397 ( 0.00 %) 40.5 - 59.5 586 ( 0.01 %) 41.0 - 59.0 915 ( 0.01 %) 41.5 - 58.5 1152 ( 0.01 %) 42.0 - 58.0 1754 ( 0.02 %) 42.5 - 57.5 2516 ( 0.03 %) 43.0 - 57.0 3458 ( 0.03 %) 43.5 - 56.5 4808 ( 0.05 %) 44.0 - 56.0 6869 ( 0.07 %) 44.5 - 55.5 9416 ( 0.09 %) 45.0 - 55.0 13020 ( 0.13 %) 45.5 - 54.5 17832 ( 0.18 %) 46.0 - 54.0 24565 ( 0.25 %) 46.5 - 53.5 33482 ( 0.33 %) 47.0 - 53.0 45844 ( 0.46 %) 47.5 - 52.5 61816 ( 0.62 %) 48.0 - 52.0 83379 ( 0.83 %) 48.5 - 51.5 113676 ( 1.14 %) 49.0 - 51.0 151159 ( 1.51 %) 49.5 - 50.5 201821 ( 2.02 %) 50.0 - 50.0 269817 ( 2.70 %) 50.5 - 49.5 356723 ( 3.57 %) 51.0 - 49.0 471932 ( 4.72 %) 51.5 - 48.5 577405 ( 5.77 %) 52.0 - 48.0 655303 ( 6.55 %) 52.5 - 47.5 699745 ( 7.00 %) 53.0 - 47.0 713783 ( 7.14 %) 53.5 - 46.5 699681 ( 7.00 %) 54.0 - 46.0 665239 ( 6.65 %) 54.5 - 45.5 615738 ( 6.16 %) 55.0 - 45.0 557825 ( 5.58 %) 55.5 - 44.5 494528 ( 4.95 %) 56.0 - 44.0 432385 ( 4.32 %) 56.5 - 43.5 371393 ( 3.71 %) 57.0 - 43.0 315378 ( 3.15 %) 57.5 - 42.5 264370 ( 2.64 %) 58.0 - 42.0 219208 ( 2.19 %) 58.5 - 41.5 179163 ( 1.79 %) 59.0 - 41.0 145748 ( 1.46 %) 59.5 - 40.5 116902 ( 1.17 %) 60.0 - 40.0 92393 ( 0.92 %) 60.5 - 39.5 72826 ( 0.73 %) 61.0 - 39.0 56881 ( 0.57 %) 61.5 - 38.5 44185 ( 0.44 %) 62.0 - 38.0 33776 ( 0.34 %) 62.5 - 37.5 25683 ( 0.26 %) 63.0 - 37.0 19404 ( 0.19 %) 63.5 - 36.5 14617 ( 0.15 %) 64.0 - 36.0 10925 ( 0.11 %) 64.5 - 35.5 7942 ( 0.08 %) 65.0 - 35.0 5813 ( 0.06 %) 65.5 - 34.5 4185 ( 0.04 %) 66.0 - 34.0 2966 ( 0.03 %) 66.5 - 33.5 2057 ( 0.02 %) 67.0 - 33.0 1518 ( 0.02 %) 67.5 - 32.5 1038 ( 0.01 %) 68.0 - 32.0 721 ( 0.01 %) 68.5 - 31.5 479 ( 0.00 %) 69.0 - 31.0 348 ( 0.00 %) 69.5 - 30.5 245 ( 0.00 %) 70.0 - 30.0 155 ( 0.00 %) 70.5 - 29.5 90 ( 0.00 %) 71.0 - 29.0 60 ( 0.00 %) 71.5 - 28.5 33 ( 0.00 %) 72.0 - 28.0 34 ( 0.00 %) 72.5 - 27.5 8 ( 0.00 %) 73.0 - 27.0 10 ( 0.00 %) 73.5 - 26.5 9 ( 0.00 %) 74.0 - 26.0 2 ( 0.00 %) 74.5 - 25.5 4 ( 0.00 %) 75.0 - 25.0 1 ( 0.00 %) 75.5 - 24.5 1 ( 0.00 %) 76.0 - 24.0 0 ( 0.00 %) 76.5 - 23.5 0 ( 0.00 %) 77.0 - 23.0 0 ( 0.00 %) 77.5 - 22.5 0 ( 0.00 %) 78.0 - 22.0 0 ( 0.00 %) 78.5 - 21.5 0 ( 0.00 %) 79.0 - 21.0 0 ( 0.00 %) 79.5 - 20.5 0 ( 0.00 %) 80.0 - 20.0 0 ( 0.00 %) 80.5 - 19.5 0 ( 0.00 %) 81.0 - 19.0 0 ( 0.00 %) 81.5 - 18.5 0 ( 0.00 %) 82.0 - 18.0 0 ( 0.00 %) 82.5 - 17.5 0 ( 0.00 %) 83.0 - 17.0 0 ( 0.00 %) 83.5 - 16.5 0 ( 0.00 %) 84.0 - 16.0 0 ( 0.00 %) 84.5 - 15.5 0 ( 0.00 %) 85.0 - 15.0 0 ( 0.00 %) 85.5 - 14.5 0 ( 0.00 %) 86.0 - 14.0 0 ( 0.00 %) 86.5 - 13.5 0 ( 0.00 %) 87.0 - 13.0 0 ( 0.00 %) 87.5 - 12.5 0 ( 0.00 %) 88.0 - 12.0 0 ( 0.00 %) 88.5 - 11.5 0 ( 0.00 %) 89.0 - 11.0 0 ( 0.00 %) Elapsed time: 93.49 seconds.
Regards from Spain.
Ajedrecista.
Komodo 9.3x give-away contest - ends 18:00 CET (Nov 11)
Moderators: hgm, Rebel, chrisw
-
- Posts: 4889
- Joined: Thu Mar 09, 2006 6:34 am
- Location: Pen Argyl, Pennsylvania
Re: Komodo 9.3x give-away contest, ends 18:00h CET (Nov., 11
-
- Posts: 4889
- Joined: Thu Mar 09, 2006 6:34 am
- Location: Pen Argyl, Pennsylvania
Re: Komodo 9.3x give-away contest, ends 18:00h CET (Nov., 11
I'm not the judge - but my take is that whoever gets the spread +6 +5 etc exactly right should win. But if there's a tie, well then who predicted the correct number of wins and losses , or closest to correct number of win and losses should win. Using your hypothetical example, say there are two entries, one with 50 wins and 50 losses and one with 100 draws. They are both net zero. Say the rest of the games are drawn - Komodo finishes plus 2 and the two closest are the two that predicted +0. The person who predicted 100 draws , got 98 games right and two wrong. The person who predicted 50 wins and 50 losses , got 2 games right and 98 wrong. Obviously, IMHO, the one who predicted the 100 draws is closest and should win. Just my $.02 I'm not the judge and it's not my contest. If you just gave a score and not the K wins and losses and you are with somebody who did , you can't win. You didn't follow the rules and you can't possibly be closer in predicting wins or losses. So the wins and losses were obviously in play to decide a tiebreak between those who got the score right and the last tie break is the time. The odds of somebody winning who just gave score is incredibly small - but I have to ask - the form asked for wins and losses - why would you not submit wins and losses? To me that that's like Exam 101 - you answer the question that is askedDann Corbit wrote:But if you just gave a score, how would you differentiate between 50 wins and 50 losses and 100 draws, for instance?MikeB wrote:+1000 it did - But you can compute the score from wins and losses a=everything else is draw.Marek Soszynski wrote:I thought the competition asked for number of wins rather than points.
-
- Posts: 1975
- Joined: Wed Jul 13, 2011 9:04 pm
- Location: Madrid, Spain.
Komodo 9.3x give-away contest: ends 18:00h CET (Nov., 11th).
Hello again:
The mode is more less 2.05% of the simulations and it corresponds with +8 =92 -0 (a score of 54-46 in favour of Komodo):
After a quick search, other results with more than 150000 simulations (> 1.5%) are:
IMHO, 0 wins for SF after 24 games hampers a lot SF in these forecasts. The current situation is even worse, with +2 =24 -0 for Komodo (no SF wins after 26 games).
Regards from Spain.
Ajedrecista.
I modified my source code. The starting point is +2 =22 -0 again:Jesse Gersenson wrote:Yes, it asked for number of wins and number of losses. Each entry was given microsecond timestamps.
Code: Select all
program TCEC_forecast
implicit none
integer,parameter :: simul = 10000000 ! 1e+7 simulations.
integer :: i, j, K, SF, d, n(2:78, 0:76)
real(KIND=2) :: probK, probSF, PRN, t0, t1
t0 = cpu_clock@() ! Start.
n = 0
do i = 1,simul
K = 2; SF = 0; d = 22 ! The result after the first 24 games.
do j = 25, 100
probK = (K + 1d0)/(K + SF + d + 3d0) ! +3 due to {win, draw, lose}.
probSF = (SF + 1d0)/(K + SF+ d + 3d0) ! +3 due to {win, draw, lose}.
PRN = random@() ! PRN: pseudo-random number.
if (PRN < probK) then
K = K + 1 ! Komodo win.
else if (PRN > (1d0 - probSF)) then
SF = SF + 1 ! SF win.
else
d = d + 1 ! Draw.
end if
end do
n(K, SF) = n(K, SF) + 1
end do
open(unit=11, file='Forecasts.txt', status='unknown', action='write')
do K = 2, 78
do SF = 0, 76
if (n(K, SF) > 0) then
write(11,'(A,I2,A,I2,A,I2,A,I8,A)') '+', K, ' -', SF, ' =', 100 - K - SF, ' ', n(K, SF), ' simulations.'
end if
end do
end do
close(11)
write(*,'(A,I8,A)') 'Mode: ', maxval(n), ' simulations.'
write(*,*)
t1 = cpu_clock@() ! Finish.
write(*,'(A,F7.2,A)') 'Elapsed time: ',(t1-t0)/3d9, ' seconds.' ! 3 GHz in my PC.
end program TCEC_forecast
Code: Select all
Mode: 205121 simulations.
Elapsed time: 80.45 seconds.
Code: Select all
[...]
+ 7 -37 =56 1 simulations.
+ 8 - 0 =92 205121 simulations.
+ 8 - 1 =91 155636 simulations.
[...]
Code: Select all
+ 8 - 0 =92 205121 simulations.
+ 7 - 0 =93 201189 simulations.
+ 9 - 0 =91 200319 simulations.
+10 - 0 =90 190051 simulations.
+ 6 - 0 =94 187463 simulations.
+11 - 0 =89 174658 simulations.
+ 5 - 0 =95 162589 simulations.
+12 - 0 =88 158377 simulations.
+ 8 - 1 =91 155636 simulations.
+ 7 - 1 =92 153528 simulations.
+ 9 - 1 =90 152081 simulations.
Regards from Spain.
Ajedrecista.
-
- Posts: 1536
- Joined: Sun Oct 25, 2009 2:30 am
-
- Posts: 1301
- Joined: Sun Mar 12, 2006 6:46 pm
- Location: Kelowna
- Full name: Tony Mokonen
Re: Komodo 9.3x give-away contest - ends 18:00 CET (Nov 11)
It was yours truly.
Congratulations to Larry and Mark for winning season 8 of TCEC.
Congratulations to Larry and Mark for winning season 8 of TCEC.
-
- Posts: 1536
- Joined: Sun Oct 25, 2009 2:30 am
Re: Komodo 9.3x give-away contest - ends 18:00 CET (Nov 11)
You guessed it, down to the number of draws?
-
- Posts: 1301
- Joined: Sun Mar 12, 2006 6:46 pm
- Location: Kelowna
- Full name: Tony Mokonen
Re: Komodo 9.3x give-away contest - ends 18:00 CET (Nov 11)
Yes, not sure how, but I guessed the exact result, +9 -2 =89
-
- Posts: 593
- Joined: Sat Aug 20, 2011 9:43 am
Re: Komodo 9.3x give-away contest - ends 18:00 CET (Nov 11)
It was a three-way tie for first place. Tony guessed 9-2 as did two other people. Way to go Tony, Ian and Mark B.!
-
- Posts: 1536
- Joined: Sun Oct 25, 2009 2:30 am
Re: Komodo 9.3x give-away contest - ends 18:00 CET (Nov 11)
Could you share how the votes got spread?
-
- Posts: 593
- Joined: Sat Aug 20, 2011 9:43 am
Re: Komodo 9.3x give-away contest - ends 18:00 CET (Nov 11)
Ozymandias wrote:Could you share how the votes got spread?
Code: Select all
0 0
0 0
0 0
0 0
0 0
0 0
0 100
2 5
3 0
3 0
3 0
3 2
3 3
4 0
4 1
4 1
4 2
4 2
4 2
5 0
5 0
5 0
5 0
5 1
5 -1
5 2
5 2
5 3
5 4
6 0
6 0
6 0
6 2
6 2
6 2
6 2
6 2
6 2
6 2
6 3
7 0
7 1
7 1
7 1
7 2
7 2
7 3
7 3
7 3
7 4
7 5
7 6
7 9
8 0
8 0
8 1
8 1
8 1
8 2
8 3
8 3
8 3
9 0
9 1
9 2
9 2
9 2
9 4
10 0
10 2
10 2
10 6
11 2
11 4
12 1
12 3
12 3
12 4
12 4
12 4
13 1
13 2
13 3
13 4
13 7
14 3
15 4
20 10
50 0
50 49
52 0
52 0
53 0
53 0
53 0
53 2
54 0
54 0
55 0
55 0
55 46
56 0