I hate this hobby...

[Evert]

So, I'm trying to improve SjaakII by improving its king safety evaluation in drop-games.

So, I have a specific idea I'd like to try for Shogi and mini-Shogi. Let's give it a go:

Shogi:

Code: Select all

Rank Name            Elo    +    - games score oppo. draws 
   1 Sjaak II 700M  2168   16   16  1890   65%  2032    0% 
   2 Sjaak II 700   2032   16   16  1890   35%  2168    0% 
               Sj Sj
Sjaak II 700M    100
Sjaak II 700    0   

Mini-shogi:

Code: Select all

Rank Name            Elo    +    - games score oppo. draws 
   1 Sjaak II 700   2112   13   13  2800   53%  2088    0% 
   2 Sjaak II 700M  2088   13   13  2800   47%  2112    0% 
               Sj Sj
Sjaak II 700      99
Sjaak II 700M   0   

Meh. Ok, there seems something to this idea. Let's try some tweaks to it.
Mini-shogi:

Code: Select all

Rank Name            Elo    +    - games score oppo. draws 
   1 Sjaak II 700M  2154   13   13  2800   62%  2046    0% 
   2 Sjaak II 700   2046   13   13  2800   38%  2154    0% 
               Sj Sj
Sjaak II 700M    100
Sjaak II 700    0   

Hey, not bad! Shogi:

Code: Select all

Rank Name            Elo    +    - games score oppo. draws 
   1 Sjaak II 700   2254   19   19  1806   82%  1946    0% 
   2 Sjaak II 700M  1946   19   19  1806   18%  2254    0% 
               Sj Sj
Sjaak II 700     100
Sjaak II 700M   0   

[Daniel Anulliero]

Hello
Sometimes I hate the hobby too
So , If I look correctly you did two tests with the elo result total inversed?

[Henk]

Are there any members who don't hate this hobby ?!

[Evert]

Hello
Sometimes I hate the hobby too
So , If I look correctly you did two tests with the elo result total inversed?

Something like that. I can improve play in mini-Shogi, or in regular Shogi, but only at significant cost to the other.
Of course, what this means is that I need to find the correct way to tell these apart and scale the evaluation terms appropriately. Easier said than done, of course.

[tuxedocat]

Are there any members who don't hate this hobby ?!

I'm still naive enough that I don't hate it. In fact, I'm here working on it at 22:30 Friday night.

[Angrim]

I just hate that I don't have enough time currently to really take part in this hobby.

[Evert]

Something like that. I can improve play in mini-Shogi, or in regular Shogi, but only at significant cost to the other.
Of course, what this means is that I need to find the correct way to tell these apart and scale the evaluation terms appropriately. Easier said than done, of course.

Ok, I've found something that works: if I scale the king-safety term related to threatened squares around the enemy king by board size, this seems to improve play for both mini-Shogi and regular Shogi, while it's perfectly neutral for Crazyhouse. A test for Tori Shogi is still running.

What makes me uneasy is that I don't see why that type of scaling is reasonable: it gives more weight to having safe squares near the king on a smaller board, which seems more like a general mobility term (controlling squares on the enemy side of the board) than a king-safety term...