On-line engine blitz tourney October

[Henk]

Why not split into a two monthly European tournament and an American tournament.

Of course American participants could participate in European tournament etc.

For each tournament you take time that fits best for local participants and of course for tournament manager.

I personally think 9 o'clock is one hour or perhaps to hours too late. But I probably won't join for engine plays too bad or is not interesting enough to watch it play games.

I mean split into three monthly European, American and North Korean tournaments.

[hgm]

One more thing: please have your interface save all the games, and then post them in a thread I made for this purpose in the tournaments section:

http://www.talkchess.com/forum/viewtopic.php?t=65565

[Sven]

I just patched mamer to also support McMahon pairing. This is a form of Swiss where the participants already start with a number of points. By giving strong participants more initial points, you get more interesting pairings; in normal Swiss the first two rounds are wasted on dead certain wins of top players against tail players.

The disadvantage is of course that if your initial seeding is low, you don't stand a chance to win the tourney, even if you would win all your games, because players that got more initial free points would still end above you with fewer wins. So the system is only suitable for tourneys where the players are spread out over a huge Elo range. Like in this case.

I have implemented it such that for 9 rounds the top half (ratingwise) of the participants start with 5 points. Below that the number of points will decrease linearly to 0 for number last. For 40 participants that would mean number 21 and 22 get 4.5 free points, number 23 ad 24 4.0, etc. This will not distort the winning chances for the top half, but keeps their games concentrated mostly to within that group, and avoid them playing against the bottom 25%.

Shall we use this system tonight? I have tested it on a tourney with 14 participants and 6 rounds, and it seems to work fine there now.

I think you can give it a try so that we can see whether that McMahon system works in practice.

However, in theory I can imagine that giving 5 initial points to the top half and then a linearly decreasing number of points to the bottom half might somehow lead to an unrealistic result. Also I don't know if the goal of avoiding those "boring first two rounds" would be achieved that way without getting other disadvantages. If all games would be decided according to the rating difference, with some threshold for drawing, then after 5 rounds you would have at most 1/32 of all participants with 5 points. Some more would have 4.5 (if noone has 5) or 4 points, even a few more 3.5, then 3 and so on. Players with 50% of all possible points (2.5/5) are always right in the middle of the standings. Your system is kind of asymmetric under that aspect. It leads to pairing the top quarter against the second quarter in the first round but then, with 40 participants, for the bottom half it results in pairings like 21-22, 23-24, 25-26 and so on which you would normally expect during the last rounds of the tourney.

I think that pairing players of similar strength against each other too early might sometimes also lead to running out of appropriate pairings during the last rounds, so the effect might be that those pairings with a bigger rating difference that you tried to avoid in the first rounds will now happen in the last rounds if some players were already matched against all players that are close to them in rating.

Wouldn't it be sufficient to simulate that the first two (or in general, K) rounds are perfectly decided based on rating differences, with a drawing threshold of, say, 200 Elo points? After the first virtual round of (for instance) 40 players we would have 20 players with 1 point and 20 with 0, and if the rating distance between quarters 1 and 2 and between 3 and 4 are still above the drawing threshold then we would have 10 players with 2 points, 20 (10+10) with 1 point and 10 with 0 points after the second virtual round. A third virtual round would probably have some draws already. All in all the standings would be almost symmetric, following some kind of binomial or trinomial distribution.

[hgm]

Well, first note it is not 'my system'. I also had never heard of it until I saw it at the open shogi world championship. Considering that the way I implemented it now gives on average 3.75 free points to every participant, I suppose it is a simulation of 7.5 rounds. I am not sure how much different a real distribution after a normal Swiss tourney would be. Of course we should know pretty well how it would look after 9 rounds...

I guess a simulation of a real distribution would give the number 1, 2 and 3 seeds different points. That seems an unacceptable distortion of the winning chances.

[hgm]

The server just died on me! I happened to be logged on to it through ssh; just before I was disconnected there appeared a message:

The system is going down for power off NOW!

without any prior waring. I have no idea why that is; I hope it will be back up before 9pm.

This is a rented VPS in a data center, so there isn't anything I can do about it.

[Edit]

OK, it seems we are back up. I restarted the ICS. The website of the provider had also disappeared, and rerouted me to the website of a provider I had never heard of.

[Daniel Anulliero]

The server just died on me! I happened to be logged on to it through ssh; just before I was disconnected there appeared a message:

The system is going down for power off NOW!

without any prior waring. I have no idea why that is; I hope it will be back up before 9pm.

This is a rented VPS in a data center, so there isn't anything I can do about it.

Now it works Hg

[hgm]

Good thing we did not start at 8pm!

They must have had a power glitch.

[hgm]

Code: Select all

:Tourney Players: Round 9 of 9 
:
:     Name              Rating Score Perfrm Upset  Results 
:     ----------------- ------ ----- ------ ------ ------- 
:  1 -Fizbo             [2192]  12.5  [2240] [   0] +05w +07w +11b +03w =02b =04w =06b +09w +08b 
:  2 +WaspX             [2054]  11.5  [2139] [  87] +14b =12w +04w +13b =01w -03w =09b +11w +06b 
:  3 -nemorino          [2039]  11.5  [2154] [  72] +15w +09w +12b -01b =05w +02b =04b =06w +10b 
:  4 -ArasanX           [2072]  11.0  [2069] [  60] +19w +08w -02b =09w +07b =01b =03w =05w +13b 
:  5 -Pedone            [1869]  11.0  [2078] [ 533] -01b +20w +22w +11b =03b -06w +10w =04b +09b 
:  6 +WaDuuttie         [2038]  10.5  [2016] [  77] +16w =13w -10b +19w +12b +05b =01w =03b -02w 
:  7 +Goldbar           [1930]  10.0  [1838] [   0] +17w -01b -13b +22w -04w +16b =08b =18w +19w 
:  8 -Baron             [1888]  10.0  [1878] [ 117] +20w -04b +16w =10w -14b +18b =07w +15b -01w
:  9 +Laser             [1883]  10.0  [1969] [ 376] +29w -03b +17w =04b +10b +13w =02w -01b -05w 
: 10 -NightmareX        [2080]  9.5  [1855] [   0] -13b +21w +06w =08b -09w +17w -05b +12b -03w 
: 11 +Amoeba            [1991]  9.5  [1811] [   0] +21w +22b -01w -05w +16b =14w =13b -02b =15w 
: 12 +Arminius          [1884]  9.5  [1835] [  85] +24w =02b -03w =15b -06w +21b +14b -10w =17b 
: 13 +Zurichess         [1834]  9.5  [1945] [ 522] +10w =06b +07w -02w +15b -09b =11w =19b -04w 
: 14 +Phalanx           [1726]  9.5  [1750] [ 294] -02w =23b =18w +20b +08w =11b -12w -16b +24w 
: 15 +rpiArminius       [1722]  9.5  [1784] [ 223] -03b +31w =19b =12w -13w +20b +23b -08w =11b 
: 16 +Schooner          [1655]  9.5  [1751] [  71] -06b +29w -08b +30b -11w -07w +31b +14w =18b
: 17 +Nemeton           [1626]  9.5  [1792] [ 461] -07b +47w -09b +23w =18w -10b =19w +21b =12w 
: 18 -tomitankChess     [1532]  9.5  [1782] [ 515] +34w +27b =14b +21w =17b -08w +33b =07b =16w 
: 19 +rpiStockfish      [1738]  9.0  [1740] [  48] -04b +24w =15w -06b =21b +22w =17b =13w -07b 
: 20 +Bliep             [1598]  9.0  [1604] [   6] -08b -05b +31w -14w =33b -15w +29w =22b +23w 
: 21 +PuppetMaster      [1637]  8.5  [1633] [  50] -11b -10b +23w -18b =19w -12w +30b -17w +31b 
: 22 +Myrddin           [1610]  8.5  [1635] [ 281] +47b -11w -05b -07b =34w -19b =24w =20w +33b 
: 23 +Nameless          [1559]  8.5  [1589] [  85] +30b =14w -21b -17b +25w +34b -15w +33w -20b 
: 24 +Floyd             [1595]  8.0  [1452] [   7] -12b -19b -30w =34b +28w -33w =22b +31w -14b
: 25 +Jumbo             [1532]  8.0  [1471] [  35] -31b =35w +27w -33b -23b =28w =32b +34b +30w 
: 26 -Winter            [1485]  8.0  [1377] [   0] -27w -33b +37w -28b +38b -32w +39b +40w +35w 
: 27 +Barbarossa        [1477]  8.0  [1514] [  69] +26b -18w -25b +38w +32b -31b +34w -29b +41w 
: 28 +RookieMonster     [1319]  8.0  [1615] [ 867] =36w +38b +32w +26w -24b =25b +35w =30b =29w 
: 29 +Spartacus         [1562]  7.5  [1426] [   0] -09b -16b -34w =31b +35w -30w -20b +27w =28b 
: 30 +Lozza             [1561]  7.5  [1414] [  35] -23w =34b +24b -16w -31w +29b -21w =28w -25b 
: 31 +Joker             [1533]  7.5  [1504] [  42] +25w -15b -20b =29w +30b +27w -16w -24b -21w 
: 32 +rpiFruit          [1413]  7.5  [1401] [ 193] -35b +39w -28b +37b -27w +26b =25w +42b =34w
: 33 +Isa               [1367]  7.5  [1623] [ 789] +38b +26w +35b +25w =20w +24b -18w -23b -22w 
: 34 -RomiChess         [1538]  7.0  [1404] [  99] -18b =30w +29b =24w =22b -23w -27b -25w =32b 
: 35 +KingSlayer        [1473]  7.0  [1402] [  29] +32w =25b -33w +36b -29b +39w -28b +38w -26b 
: 36 +microMax          [1291]  7.0  [1248] [  14] =28b -37w +39b -35w +40b -38w +44b +43w +46b 
: 37 +rpiEmbla          [1082]  7.0  [1339] [ 802] +43w +36b -26b -32w +44w -40b +45w +46w +42b 
: 38 +Eichhoernchen     [1424]  6.5  [1262] [   0] -33w -28w +41b -27b -26w +36b +42w -35b +43b 
: 39 -Abbess            [1228]  6.5  [1341] [ 447] +40w -32b -36w +41w +42b -35b -26w +45b +44w 
: 40 +Embla             [1217]  6.5  [1274] [ 458] -39b =41w +43w +44b -36w +37w +46b -26b +45w
: 41 +Skiull            [1209]  6.5  [1288] [ 470] +42w =40b -38w -39b +46w +45b +43w +44w -27b 
: 42 -Skipper           [1075]  4.5  [1147] [ 600] -41b +44w +45w +43b -39w +46b -38b -32w -37w 
: 43 +NEG               [1067]  3.5  [1044] [ 608] -37b +46w -40b -42w +45b +44w -41b -36b -38w 
: 44 +POS               [ 839]  2.0  [ 959] [ 836] +45w -42b +46w -40w -37b -43b -36w -41b -39b 
: 45 +TekiRandom        [   0]  2.0  [ 763] [ 790] -44b +48w -42b +46b -43w -41w -37b -39w -40b 
: 46 +ParisHilton       [ 790]  1.0  [ 781] [   0] +48w -43b -44b -45w -41b -42w -40w -37b -36w 
: 47 -gogobello         [forf]  5.0  [1218] [   0] -22w -17b 
: 48 +_BYE_             [forf]  0.0  [ 832] [   0] -46b -45b
:
:     Average Rating    1589.7 
:

A record number of players, this time, even if you don't count the one that had to withdraw after two rounds because of engine trouble!

We have to evaluate how we liked the McMahon pairing system. I think that in any case it has to be tweeked so that there are an even number of top engines that get maximum initial points, as it would be unfair to the one that gets 0.5 pt less and gets added to the group to play a top engine anyway. I also think that with such a large field perhaps only the top 1/3 should get equal points, and it should start to tail off from there. I guess it really depends on the rating distribution; in an advanced system you would calculate where the chance that a player can win gets below a given threshold (say 1%), and start decreasing the initial points linearly from there.

Another conclusion is that the gap between searching and non-searching engines is too large, so that the non-searching engines are essentially BYEs for the others. Making the luck factor for how many of those you get paired with for the tail of the searching engines substantial.

[Henk]

Maybe better play one minute games.

[Sven]

Maybe better play one minute games.

Didn't Skipper play one minute games?