Compiling a list of openings

Tafl strategy
User avatar
Hagbard
Posts: 688
Joined: Sat Mar 14, 2015 6:07 pm
Location: Copenhagen
Contact:

Re: Compiling a list of openings

Post by Hagbard » Thu Apr 20, 2017 10:10 am

Hagbard wrote:In a few weeks I'll set up a test tournament, asking players to start with the cluster 17 two moves.
A test tournament of the Copenhagen Brench cluster 17 was done with 5 players, 20 games.
Brench predicted 83% white wins, and the test came very close to that: 75% white wins!

nath
Posts: 65
Joined: Mon Oct 09, 2017 9:34 pm

Re: Compiling a list of openings

Post by nath » Mon Oct 09, 2017 10:19 pm

I think the prefixes are actually to short to say things definitely, but I am intrigued that there's something happening here. I'd love some top level player discoursing about openings. I personally think (even though I pushed an promoted that move when there where little to none composed openings) that the h3 is overrated. I think the reason that move shows the best rating is because nearly all educated players play it - not because it's extraordinary strong (even though it isn't weak - otherwise I wouldn't have adopted it in the first place).

I'd love to see discussions to understand WHY a certian opening is weak or strong. I can easily see that the results of the cluster 17 tournament are tainted because the average player was to weak and white get more wins (statistically) if the black player isn't strong.
I won't say that this move is weak (I myself play that from time to time), but I don't understand yet enough to say whether it's better or worse than h3 c8 (even though it felt more natural to me back in the day).

Brench
Posts: 19
Joined: Sat Feb 11, 2017 7:14 am

Re: Compiling a list of openings

Post by Brench » Thu Nov 09, 2017 6:17 pm

Hagbard wrote:
Thu Apr 20, 2017 10:10 am
Hagbard wrote:In a few weeks I'll set up a test tournament, asking players to start with the cluster 17 two moves.
A test tournament of the Copenhagen Brench cluster 17 was done with 5 players, 20 games.
Brench predicted 83% white wins, and the test came very close to that: 75% white wins!
It's been a while and I apologize, but real life data analysis has gotten in the way.

I'd like to continue following up Cluster 17. Although 75% is close to 83%, it's smaller by enough that I'm wondering if there is a good black countermove that can increase black win probabilities.

Not sure how long this will take, but thanks for the feedback so far. It's been very instructive.

nath
Posts: 65
Joined: Mon Oct 09, 2017 9:34 pm

Re: Compiling a list of openings

Post by nath » Sat Nov 11, 2017 3:58 am

Are you sure that judgment of 75 % (or higher) is accurate? I think if you let people like Plantagenet and me face off each other the result would be that drastic.

What are you trying to measure? What do these probability even mean? Looking at the players there where a few players in there, who are better handling the white stones than the black ones (as opposed to Plantagenet, who has a better win rate with black). I also firmly believe that if I really go serious I can give any player a very hard time with black, opposed to my white play, which probably doesn't even pose the slightest thread to Plantagenet. Please look at the players when you post these numbers. 5 players isn't enough to create reliable numbers.

The opening is interesting, but I don't think it's to scary. If somebody wants to take my word for it, I don't fear a challenge using that opening having the black pieces.

Besides that I don't think any of this is forced, black still has to openings left, I proposed back in 2014, that are still mainly unexplored.

Brench
Posts: 19
Joined: Sat Feb 11, 2017 7:14 am

Re: Compiling a list of openings

Post by Brench » Mon Nov 13, 2017 5:19 pm

I dont' know that I'm trying to measure anything. I am trying to identify themes amongst openings and the relative success or failure of those themes in practice. The approach is identical to that employed in books of chess openings, where lines and variations are catalogued by their win frequencies (https://en.wikipedia.org/wiki/Chess_ope ... eory_table). Ultimately a book of openings would be a laudable goal, but I think at this point a few tables is a good enough start.

I'm sure that accounting for individual differences in playstyle could impact these numbers to a greater or lesser extent, but I am interested in broad trends over large gaming populations if possible, where individual styles should ultimately wash out. The database of logged games on this site is probably as close to a "large population" as we are likely to get, given the relatively small number of Tafl players in the world.

A small test tournament of 5 players, is not enough to generate probabilities, but can be a part of testing established win probabilities and can be a way to assess the impact of previously unexplored lines that follow a given opening.

Obviously, the mathematical complexity of Tafl games is lower than that of chess based simply on piece differentiation and board symmetry, but the game is still very complex and establishing a solid theory of openings is critical to an overall development of hnefatafl theory.

nath
Posts: 65
Joined: Mon Oct 09, 2017 9:34 pm

Re: Compiling a list of openings

Post by nath » Wed Nov 15, 2017 3:17 am

My point is that we aren't getting anywhere near that "large population". You need a few geniuses to get closer to the perfect play, and despite our community stagnated a couple of years, there is some promise that players pick up the struggle to progress the game again.
A lot of these tables only account for players with more than 2 k elo. And 2 k elo in chess means much more than, 2 k in our current rating system. Even though I think few players are underrated I don't think we have a player comparable to a 2 k elo chess player in our community at the current stage.
I don't mind you trying, I just wonder what these numbers are supposed to tell me.

I'd like to argue that the complexity of chess is higher than the one of chess. If you just look at the first move, you are right that the symmetry plays a great role, later in the game, the importance shrinks quite quickly. In the middle game in chess there are on average a little more than 40 possible moves, leading to a little bit more than 1.6 k variations for a full move. Hnefatafl starts out with nearly 7k variations for the first more (if I don't account for symmetry, which is a nearly safe assumption midgame), while the amount of white moves drastically increase during the opening phase, quickly ramping up the amount of full moves to more than 12 k. I admit that this comparison isn't completely fair, since chess games appears to be slightly longer than Hnefatfl games, but unless you apply the ludicrous low move counts we are seeing in Valhöll currently (that would indeed suggest that the movement tree of chess bigger on average), there are more considered moves in a Hnefatafl game...not to mention the task to calculate a task to a certain depth.
I highly suspect based on my own experience: if you play a strong game, the battle is still (I said still - not again) raging quite fiercely after 35 full moves.

Draganov
Posts: 55
Joined: Fri Oct 30, 2020 2:59 pm

Re: Compiling a list of openings

Post by Draganov » Mon Jun 26, 2023 6:56 pm

Hagbard wrote:
Wed Feb 15, 2017 10:42 am
I extracted the first four moves from all games between strong players for the most used variants:
http://aagenielsen.dk/tafl_openings.txt

Hope it can be of use.
That's really great and could be helpful for analyzing the openings. Hagbard, can you also update the data with games until today since the data from the link was cut in 2017.
If you could give access to the whole Copenhagen game moves (not only the first 4 moves), it will be amazing.

Post Reply