Hnefatafl Internet Championship

[Hagbard]

A combination of Round Robin tourn., Swiss tourn. (Monrad system) and balanced Double Elimination tourn. was tested using Fetlar and Tawlbwrdd and worked.

1. All players were ordered by their separate rating for the variant used.

2. First round, all players have a match against their nearest neighbour in rating to both sides, that is two matches ("Battle of Brothers"). A match is a game with player A white player B black, and a game more with reverse color.

3. If a player lost 2 matches, he's eliminated. A tie is 0.5 lost match to each player.
Points:
A match has 1 point in play. A match is two games, each with 0.5 point in play.
If a player loses a game, he lost 0.5 point (0.5 match). If a game is a draw, each player lost 0.25 point from that game (0.25 match).

4. Second round, all players have a match against their next-nearest neighbour in rating to both sides, that is again two matches ("Battle of Family"). Again players with a total of 2 losses or more are eliminated.

5. It is all the way aimed at each player in each round having two opponents, one stronger and one weaker as close in strength as possible ("Battle of Neighbours"). And that a match is not repeated, as far as possible. After each round, players with a total of 2 losses or more are eliminated.

6. Eventually perhaps one player is left, the winner.

Or perhaps two players, who must then have another round possibly with several matches at once ("Duel").
If both players have 0 losses it would take 3 matches possibly plus one. If one player has 0 losses and the other 1 loss it would take 2 matches possibly plus one.

7. If top players are very even, they could all kill each other off. If that happens, their results against each other is counted for this round. In case of a tie, the Sonneborn-Berger score is calculated for this round. If still a tie, the player who won with the shortest game length, is the winner. (If a game was not played to the end, the number of moves it would take to finish the game, is added.) In the unlikely case of still a tie, another round among the tied players to find the winner.

8. Short timing was used all the way.

- The two first rounds each of 2 opponents can be one round against 4 opponents and double timing.

[Hagbard]

This is my bid for an evaluation of the Championship Tournament 2021:

Changes from last year:
A new tournament format was tried this year:
A Double-Elimination format which kind of mixes Double Elimination (balanced) and Swiss-system (Monrad).
(It is like a Round Robin with doing only the selected, decisive games.)

• The players are ordered by their max. rating of the variant.
• Round 1 everyone plays matches against the four nearest rated players, two above and two below.
• Players who lost 4 games or more (i.e. 2 matches) are eliminated and don't move on to the next round.
• Next round the surviving players are again ordered by max. rating and as far as possible:
  • everyone plays a match against nearest stronger rated player. A player with only 0 or 0.5 loss (0 or 1 lost game), however, against two stronger players.
  • two opponents to every player.
  • same opponents don't meet again.

Three lines of tournament:

• Copenhagen (strong king)
• Tawlbwrdd (weak king)
• Sáhkku

What worked well:

• The new tournament format worked well. The Double Elimination gave more accurate results compared to a Single Elimination.
• To have three lines of tourn. with different game types.
• To have an umpire for each line plus a backup umpire, in all four umpires.
• That the admin set up the games but did not umpire.
• The timing worked well ( x2 with gradually reduced timing for high time use.)

Remarks.
The games were of remarkably high quality all across the board. Opponents of neighbouring strength have an opportunity to show their very best and probably have more fun, and no matter whether it's in the top or the other end of the rating scale the games are certainly interesting to watch.

This format handles any number of players, a large number as easily as a small number.
No matter where a player is on the rating scale, this format treats everybody the same.

All players meeting their 4 nearest neighbours functions like one long continuous group of 5 players all way through from top to bottom, with every player in the middle of the group.

The tournament was like one long Final. The whole rating scale was immediately activated and top players met from the very start. This made it possible to test players of the whole rating scale very thoroughly. The winner fx. played against 6 topmost players before the title was won.
It also made it possible to accomplish the whole tournament in a relatively short time.

To let equals meet all the way as far as possible, gives max. information about their relative strength and thus an accurate tournament result.

It's easy to handle no-shows in this format. An absent player is simply substituted with the next in line, and it's as if he were never there.


The winner was found after 86 days (last year 130 days), and the whole tournament was finished after 89 days (last year 186 days).
30 players joined the tournament.
In total 210 games, 11470 moves and 2829 killed pieces.

Tournament page:
http://aagenielsen.dk/wtfturnering2021.php

Congratulations to World Tafl Federation Champion 2021, Plamen Draganov ("Draganov"), Sofia, Bulgaria, winner of both Copenhagen and Tawlbwrdd!
Also congratulations to Bogdan Varbanov ("Bogd Khan"), Plovdiv, Bulgaria, winner of Sáhkku!
Many thanks to the four umpires Mike Coveny, Hamish Lawson, Mikkel Berg-Nordlie and Adam Bartley!!
And to the 30 players who took part!

[Hagbard]

Inspired from the current Single Elimination tournament test, I had another look at the similar Double Elimination used in the Copenh. World Champ. tourn. 2021.

The Copenh. World Champ. tourn. 2021 was a big tournament with 21 players.
Many players were of equal strength, and therefore there were very many tied matches where two players won one game each.

The Single Elim. tourn. test showed, that when tied matches are counted as 0.5 loss to each player, and these losses are transferred to the next rounds, then there is a big chance for essentially all players being eliminated before a Final.

The Single Elim. tourn. test also showed, that if tied matches losses are not transferred to the next rounds, or if tied matches results are ignored altogether, then the tournament works better.

As said, there were very many tied matches, and to ignore or not transfer the results of so many games would be like a waste of a lot of players' effort. It would be 24 ignored games out of 72 games in Round 1.


Instead the "shortest game length" criteria can be used to solve all those ties and get a definitive match result:
"The player who wins with the shortest game length, is the winner. (If a game was resigned, the number of moves it would take to finish the game, is added.)"

Inspecting the 12 tied Round 1 matches:
Paul7 - Rollo: Both won in 32 moves. No winner, no loser.
Fenric7 - Waegn: Fenric7 won in 15 moves, Waegn in 28 moves, winner Fenric7.
Holidayinnes - Cacreal: Holidayinnes won in 90 moves, Cacreal in 128 moves, winner Holidayinnes.
Colophonius - Paul7: Colophonius won in 28 moves, Paul7 in 39 moves, winner Colophonius.
MasterLuke - Fenric7: MasterLuke won in 57 moves, Fenric7 in 50 moves, winner Fenric7.
Rollo - Floki: Both won in 74 moves. No winner, no loser.
Casshern - Draganov: Casshern won in 38 moves, Draganov in 142 moves, winner Casshern.
SebastianMucke - Holidayinnes: SebastianMucke won in 94 moves, Holidayinnes in 108 moves, winner SebastianMucke.
Luizz - Floki: Luizz won in 45 moves, Floki in 59 moves, winner Luizz.
Sqaree - Holidayinnes: Sqaree won in 118 moves, Holidayinnes in 50 moves, winner Holidayinnes.
Sqaree - Garun19: Sqaree won in 194 moves, Garun19 in 70 moves, winner Garun19.
OdinHimself - Casshern: OdinHimself won in 103 moves, Casshern in 64 moves, winner Casshern.

Ignoring the tied matches, eliminated are: Sakar, Colophonius, SebastianMucke, Docbullen, Waegn, Red-ron.
Inspecting and including the tied matches, further eliminated are: Cacreal, Floki, Sqaree, OdinHimself, MasterLuke.

Moving on to Round 2 would be:
Casshern 0, Draganov 1, Holidayinnes 1, Garun19 0, Jrton80 1, Luizz 1, Paul7 1, Rollo 0, Fenric7 1, Potapych 0.

Round 2.
Round 2 would eliminate Holidayinnes and Rollo, and probably also Garun19, Luizz and Fenric7.
Casshern and Draganov would move on to Round 3, and probably also Potapych.
The probable Round 2 results for Jrton80 and Paul7 are unknown.

- Noone should be eliminated except when lost to a stronger player (single elim.) or when lost to two stronger players (double elim.)

Further criteria.
If game lengths are the same, next criteria: Who captured the most pieces in his winning game.
If the players are still equal: Who captured the most pieces in his losing game.
If still equal: both players advance.

These additional criteria would in the example above let Rollo advance to Round 2 with 1 loss instead of 0, no other change.

[Hagbard]

Copied from the small forum Feb. 2022:

sqAree: I just wanna mention that "shortest game length" is a terrible criteria to decide draws (or anything at all, in the first place). Why do we still have something like that in the tournament rules?

Hagbard: The shortest game length tells which of the players had the best play in his losing game
A win with the longest game length means that your opponent gave the strongest resistance and was the hardest to beat, he played the losing color best.
Fx. Casshern who is a strong player, won a game with 38 moves against 142 of his opponent, and another with 64 moves against 103 of his opponent.

sqAree: Once the whites are completely surrounded, hardly anything that happens in the game indicates how strong the resistance is, or the like. I've always found it honestly disrespectful when players keep playing for 200 moves accomplishing nothing until the game is over, if it wasn't for that tournament rule. It's a weakness in the rules, because it incentivises those super long boring dull games where nothing happens in the end. Despite, the length of a game doesn't directly corresponds to the level of resistence, at least not in a way that aligns with what I feel should be the spirit of a strategic board game; sometimes moves that lose sooner than later if the opponent responds correctly have more potential to fool or outplay the opponent and lead to a comeback than a move that just delay the inevitable.
The criteria "longest game" doesn't imply "strongest resistance", and even beside that, it leads to boring gameplay and wastes everyone's time.

sqAree: Yeah it's subjective, what I said might be just me, but I can hardly believe many people wouldn't agree here tbh.

Hagbard: With this criteria, Casshern could've resigned after 50 and 80 moves. But most often, I suppose, a long game continues because the opponent could make mistakes and turn a lost game into a winning game.

sqAree: Haven't looked into those specific games you mentioned, but very often, no, the game doesn't continue bc there could be mistakes, because usually the enemy is already completely surrounded. We already have a rule that acknowledges this, namely the rule that after completely surrounding the enemy and covering the edge too, it is automatically lost. There are so many games where the edge isn't covered, but the only remaining edge squares are not enough to build an edge fort at all, in that case the game is exactly as hopeless as the complete encirclement position. Of course there are plenty other examples where playing on also just doesn't give any chances.
When someone wants to keep fighting for a draw or a win, sure, why not, however the rules shouldn't incentivise prolonging games for nothing. Another point is that tournament rules (means rules outside of the game itself) shouldn't dictate ingame strategy too much (this can't be prevented entirely, but let's still try?).

Hagbard: I didn't inspect the two games either, but anything longer than 38 and 64 moves would be a loss by the game length criterium, so no point in continuing only for this reason.
- "... tournament rules (means rules outside of the game itself) shouldn't dictate ingame strategy too much (this can't be prevented entirely, but let's still try?)."
Wise point. But some criteria are needed. Football championships end in penalty kick competitions, that we cannot do in internet board games -

Draganov: I think when it comes to playing only 2 games, the rule for game length is not sufficient. It is just too little criteria to determine the winner and the better player from the match. For example, there was huge controversion about the FIFA away goals rule in Football. Now, FIFA decided to abolish it. I proposed players to play more than 2 games in their matches. For example, if we have double eliminaion, every player will face 2 opponents with higher rating and 2 opponents with lower rating. It will mean that a single player is playing 8 games in total in the round. If we convert this to a simple single elimination but increase the number of played games, we can have a round in which the player is facing only one opponent but plays 8 games against him. Then, in case of a tie (4:4) the game length criteria will have more sense because it is based on the shortest game based on 4 games and not only based on just one single game. When it is based on one single game, we could have a problem when player timed-out early in one of the game and won the other. In the final of 2020 world championship, I lost to Bogd Khan with the whites and won a short victory with the blacks because Bogd Khan timed out in uncertain position. If we played in an elimination match, he should have been eliminated because my timing-out victory was shorter than his victory. However, in 2020 I was clearly weaker than him. Some players might want to surprise their opponent by playing risky tactics or openings. However, if they know that their game length could have so much weight as a criteria, they will never try to play risk. Moreover, how we could decide a winner between two almost equal wins? Imagine player A wins in 100 moves and player B wins in 101 moves. Can we clearly say that player A is the better one? It is just chance and fortune. Imagine a different situation, both players A and B are in lost positions in their respective games with the whites. Player A resigned on move 100 and player

Draganov: Imagine a different situation, both players A and B are in lost positions in their respective games with the whites. Player A resigned on move 100 and player B resigned on move 101 because he tought that if he survived 1 move more than his opponent, he would have been the winner. However, the umpire said that the positions when both players resigned were not finished, so we needed to add the additional moves in order to complete them and see which game is shorter. The umpire added additional moves on his decision and decided that Player A is the winner because his game would have finished on move 140 and the game of player B would have finished on move 142. However, player B complained that the umpire hadn't played the best possible moves and the game on his estimation would have finished on move 138. It will be really confusing situation.

Each match (of two games) has one loss to split between two players.
If a player loses both games of the match, he lost 1, and if he wins both games he lost 0.

When the players win one game each and the match is a tie, it seems simple to split the loss and mete them 0.5 loss each.

However, this method of splitting the loss, which was used in the World Champ. 2021, has some drawbacks.

Example:
Four players, A strongest, B1 and B2 equals, C weakest.
A B1 B2 C
A will win against B1, B1 and B2 will tie, B2 will win against C.
Losses:
A 0
B1 1.5
B2 0.5
C 1

If B1 and B2, who are equals, switch positions
A B2 B1 C
then B2 will get the 1.5 losses and B1 the 0.5 loss.
So they are equals and both placed between A and C, but the accidental placement B1 B2 or B2 B1 heavily determines their results, results which are used to eliminate and which are transferred to the next round.

So, how to split the loss is not that simple.

----------------

- It helps a bit when you take the next round into account:
A will now play against B2, and B1 will play against C.
Losses sum of Round 1 and 2:
A 0
B1 1.5
B2 1.5
C 2

-----------------

Player A resigned on move 100 and player B resigned on move 101

It could be decided only to use the game length criteria when the lengths differ by more than 10%.

[Hagbard]

The Single-Elimination format test has till now resulted in a final between Draganov, Dimetr and Asther-Kane, which is very reasonable.

For World Championship use, however, it is rather clear that coincidence might become a too big factor by Single-Elimination.

Some time it could be interesting to test the format:

• Double-Elimination as used in the World Championship 2021
• If a match (of two games) is a tie, then
  • The player who wins with the shortest game length, is the winner. (If a game was resigned, the number of moves it would take to finish the game, is added.) If the game lengths differ by less than 10%, they're considered equal.
  • If game lengths are the same, next criteria: Who captured the most pieces in his winning game.
  • If the players are still equal: Who captured the most pieces in his losing game.
  • If still equal: both players get 0.5 loss.

[Hagbard]

Diagonal Tournament Format.
The format which we've used for the World Champ. tournaments the latest two years, was studied and described more accurately here:
http://aagenielsen.dk/diagonalformat.php
http://aagenielsen.dk/diagonalformat_race.php

[Hagbard]

Timing.

Copied from the small forum Oct. 12th 2022:

cacreal: Can we add a total time to make our moves to the tournament games? Asking for a friend...

Hagbard: Do you have a value in mind?

cacreal: The current format makes it a world record for tournament length, which in itself is noteworthy. If each player has a master timeout clock of 7 days and 1 hour per move afterwords all the round one games would be over before Sept 30th.
The game itself is a beautiful one and maybe a live event in person event could be offered 2023?

Hagbard: It's kind of a natural limitation that one can react and move in all cases only within 12-13 hours.
At night you have 8 hours of sleep and a couple of hours before and after, that's 12-13 hours.
At day you have 8 hours of work and a couple of hours before and after, that's again 12-13 hours.

xerxes: @Hagbard, Cacreal: And you have to allow for time differences too.

The current way of timing the games is very protective:
No matter an empty time buffer or what else, you never time out until after 13 hours.
This is effectively saying that you can never time out.

The positive side of this method is, that we've seen a lot of very high quality games without the threat of sudden death.
The negative side is, that a tournament can have a really long wait before the next round can begin.

A way to modify this could be:

• Your initial time buffer is 4 days, and as long as you have a positive time buffer, you have minimum 1 day till timeout (24 hours).
• If your time buffer turns negative, the minimum timeout time is slowly reduced.
• When the time buffer is -4 days, the minimum timout time becomes 8 hours.
• When the time buffer is -6 days, the minimum timout time becomes 2 hours, not less than 2 hours.

To get below 8 hours will be like hitting the rough and the bunker in golf. You could time out at night anytime.

[Draganov]

When the time buffer is -6 days, the minimum timout time becomes 2 hours, not less than 2 hours.

Does it mean that if someone's time buffer is -6 days, he/she will be timed out if can't respond with a move in 2 hours after the opponent's move. If this is the true and a player A has to respond in 2 hours, the player B just need to make a move during the night and will win the game.

I have another sugestion for the time buffer. It is a very radical idea and maybe can sounds brutal but just think about it. Imagine a player has 2 separate time buffers. The first one is a standard time buffer of 4 days and it is reduced in the same way as we have used it in the tournaments so far. The second buffer is the move buffer. It is of 10 minutes (or you can change the time but it must be a short time and not more than an hour). When a player opens and see the current game position, then the move buffer is activated and the player has a certain amount of time to make a move (for example 10 minutes to make a move after seeing the position).
I think this will eliminate the speculation with the time buffer. I think there are players who play slowly on purpose. They open a game and see the opponent's move, then arrange the position on a real board, make analyses and spend hours of thinking on the position. Some of the players are trying to time out the opponent. When they have a lost position, they start playing very slowly, making 1 or maximum 2 moves per day despite the fact that they are active in many other games. If their opponent responds to their move immediatelly, they don't make another move until 12 hours are gone even when they have just one possible move. They hope that after 12 hours their opponent will be more likely not able to respond.
Imagine a situation in which player A is in a lost position and tries to time out the opponent. When player A makes a move after 12 hours of 'thinking' and a player B responds in 1 minute, then the player A says to himsefl 'Look, my opponent is here and made a move, OK I will play after 12 hours and let's see if he will be there to respond or maybe I can play during my opponent's sleeping or working time. But I must not play now because my opponent is here'.
This is the reason why some games last too long. I don't blame such a behaviour of some of the players. I understand that it is important to win a game especially in tournaments but I do believe we need to try to reduce such 'tactics' with a better implementation of the time buffer.

If we have a second move time buffer, I think we will make the game more dynamic and will reduce the usage of dirty tactics. If such a move time buffer is too radical, we can try and test it in a separate tournament and the winner could be considered as a blitz champion. We can keep the standard time buffer and have a Classical champion as well in the same way as in chess there are classical, rapid and blitz championships.

[Hagbard]

Does it mean that if someone's time buffer is -6 days, he/she will be timed out if can't respond with a move in 2 hours after the opponent's move. If this is the true and a player A has to respond in 2 hours, the player B just need to make a move during the night and will win the game.

That's right, the player would be open for a mere execution -

I have another sugestion for the time buffer. It is a very radical idea and maybe can sounds brutal but just think about it. Imagine a player has 2 separate time buffers. The first one is a standard time buffer of 4 days and it is reduced in the same way as we have used it in the tournaments so far. The second buffer is the move buffer. It is of 10 minutes (or you can change the time but it must be a short time and not more than an hour). When a player opens and see the current game position, then the move buffer is activated and the player has a certain amount of time to make a move (for example 10 minutes to make a move after seeing the position).

In the World Tourn. boards are hidden until the player opens the board; though the opponent can see it also.
As a principle the timing method should be usable for all games, however, and not only the World Tourn.
Also, it's enjoyable to see the very high quality of the games, quality rather than blitz.

I think this will eliminate the speculation with the time buffer. I think there are players who play slowly on purpose. They open a game and see the opponent's move, then arrange the position on a real board, make analyses and spend hours of thinking on the position. Some of the players are trying to time out the opponent. When they have a lost position, they start playing very slowly, making 1 or maximum 2 moves per day despite the fact that they are active in many other games. If their opponent responds to their move immediatelly, they don't make another move until 12 hours are gone even when they have just one possible move. They hope that after 12 hours their opponent will be more likely not able to respond.
Imagine a situation in which player A is in a lost position and tries to time out the opponent. When player A makes a move after 12 hours of 'thinking' and a player B responds in 1 minute, then the player A says to himsefl 'Look, my opponent is here and made a move, OK I will play after 12 hours and let's see if he will be there to respond or maybe I can play during my opponent's sleeping or working time. But I must not play now because my opponent is here'.
This is the reason why some games last too long. I don't blame such a behaviour of some of the players. I understand that it is important to win a game especially in tournaments but I do believe we need to try to reduce such 'tactics' with a better implementation of the time buffer.

I suppose that such a strategy will not be possible with the change which I suggested -

[Hagbard]

The timing idea of Cacreal in combination with the method above was tested on the last two World Tourn. games. If using this method then:

Draganov / Cacreal. WTF World Championship Tournament 2022, Copenhagen Hnefatafl 11x11, 214 moves, ongoing 40 days
White: It's possible to time out after 3.6 hours.
Black: It's possible to time out after 24 hours.

Garun19 / Draganov. WTF World Championship Tournament 2022, Copenhagen Hnefatafl 11x11, 185 moves, ongoing 40 days
White: It's possible to time out after 3 hours.
Black: It's possible to time out after 15.6 hours.