In the Thud rules, a game has three possible outcomes: win, draw, lose. Nice and clear.
In online Thud, there's also the possibility to leave a game to complete later. But what if the game is never completed?
In online Thud, we intend for there to be a ladder scoring thungus, using the stored games from the database.
That means that there are additional moves available to online Thud players, "walk away" and "refuse to resume". I would like to make it so that these moves do not have strategic significance.
I can see various possibilities: unfortunately in each situation, walking away remains of tactical significance.
# If I count walk-away games as an automatic lose for the walker until they are resumed, then refusing to resume a game that your opponent walked away from is always the best move, so people simply can't leave games and resume them later.
# If I count walk-away games as an automatic win for the walker, then walking away is the best move, always.
# If I count walk-away games as an automatic draw, the again they can be used to affect the rankings of opponents: someone who's won 100% of their games is clearly higher in the ladder than someone who's won 90% and drawn 10%.
# If I count walk-away games the same as completed games for the sake of calculating the ladder, then walking away is a suitable move where you can only see your situation getting worse, and also (by starting a game, making one move, then walking away) allows the same attacks as if they were counted as a draw.
# If I ignore walk-away games in the scoring completely, then the move of walking away is the best one to take whenever you are losing a game.
# If I prevent people who have too many active walk-away games from playing, or impose some other social penalty on them, then it gets even more complex, remains a strategic deision, but also pulls consideration of those external social elements into the game itself, muddying the waters.
Ideas, please!
One idea:
The game is assumed to be a draw until it's completed, or until a preset time period elapses; at which time the walker loses.
As you said, this is a disincentive for people to resume games.
My suggestion would be that if someone asks to resume a game, you have the option of either playing or resigning from the game and calling it a draw.
This removes the tactical advantage of refusing to resume, and removes any advantage from walking.
The only problem with this is the social one caused by forcing someone to play a game at one particular time; which may be solved by the introduction of a 'postpone' option when challenged to resume; which either moves or removes the time restriction for the resumption.
If the 'other player' challenges the walker to resume, they have the option of either playing, or forfeiting the match (and losing), or _asking_ for a postponement.
I think this removes all potential advantages of walking away. There are advantages to the other player; they can force a draw when they were in a losing position, but I personally think this is a fairly small price.
I'm sure it'd be a nightmare to code, but I think it's fair... maybe it can be simplified?
Let me know if any of that made any sense
Jeff...
SMASH GOOHULOOG HORUG HEAD!
(Roughly translates, Hae a Faceful o hied)
T'dr'duzk b'hzg t't
(Roughly translates, hae summa that bigjobs!)
On thudgame, you can walk away at any time, just by closing the browser.
That means you can walk away in the middle of a challenge. But that can't be held against you, since you might have just been disconnected by network circumstances beyond your control.
So on the net there is always, whether there's a button for it or not, a "walk away/reject challenge with no significant penalty" option.
And you can always force someone to walk away by just not making a move forever, until they walk away, so you can't penalise walkers either.
How's this sound?
1) For a certain amount of time the game's not counted at all (since counting it as a draw will have effects on winning streaks).
2) If both players agree, any game can be removed from consideration by the rankings, whether it is completed or not. People should be allowed to have silly games where they add their own arbitrary rules like "let's see what happens if the pieces can only move diagonally!" without it affecting their rankings.
3) After a reasonable amount of time (week? month?) the game is a lose for the walker, and the stayer is the winner-by-default (it's assumed the walker walked because they were losing anyway).
4) Once a winner-by-default has been established, they stay winner-by-default until the game is resumed, if ever.
5) Extend the time period indefinitely if they just haven't been online at the same time, so that "staying offline" won't be a good tactic for the stayer.
6) Walker or stayer may challenge for a resume at any time, even after that timeout.
7) When challenged to resume, the challenged person has the option to resign the game (the challenger wins), or to postpone, in which case they are considered to be the walker.
The problem with this is that it's likely it'll end up in "tag fights" within a few seconds of the timeout. Any time the walker sees the other, they'll challenge to resume the game, so that the other half will become the walker.
A different option;
Keeping the games as undecided(non-counting) unless the walker has more than a set percentage of unfinished games, once that number is reached the oldest games are forfeited.
A greater difficulty is handeling players that refuse to end a losing game, as a player can keep a game running indefinitly by simply refusing to accept the result. These cases should be rare though, as they are annoying to carry out as well as to be subjected to. But, I can think of no other way of handeling these other than human intervention.
Good point.
The game should be able to decide "there's no way short of cretinous stupidity on the part of the proposer that the proposer could lose another piece".
At that point, I may have it that the game ending is auto-accepted.
For example, with just one troll you're /usually/ not catching anything further. But I'd definitely need to check to see if there was a dwarf next to it, or a clump of two dwarfs one space away, three two spaces away, etc. One space more if it was the dwarfs turn, one space more for each extra dwarf, up to a point, at least.
Would be interessted if anyone could come up with a good heuristic for deciding when to autoaccept.
This could easily be harder than it sounds. It would need to take into account situations where there is no safe places to move a dwarf, and situations where moving a dwarf to a space would cause another dwarf to be unable to escape. And situations where ending on a space would set up the previously mentioned situation, and so on.