Messages

In

#5293041 on tokyo.

Been frozen for 15 minutes after granting an undo request from my opponent. We both tried disconnecting and reconnecting. I am still in the game, showing his turn, but I am not being prompted to force him to resign.

I agree with Ingex. It does not seem to be a resolution only issue.

I would record it as a win for each remaining player over the resigning player, and for the two remaining players, not record a result.

Yes this is very close to what I implemented on our dev server now. Only the original game was rated, the continuation is not. The original game only produces 2 results instead of the planned 3. The difference is that I count these wins for a bit more then just a regular win. Normal wins are counted as (1 0), ties counted as (0.5 0.5), and the plan now is to count these 3P-resignation-wins as (1.5 -0.5).

That will imply that in 3P resigning is worse for your ranking then losing, which is fine with me, and the two remaining players are both a bit compensated because they were deprived of the opportunity to score 2 wins in that game.

Nice. I like the idea.

And giving 1.5 points to people is effectively the same thing as option (2) in my previous post (counting the remaining result as a tie for the players who didn't resign), unless I'm misunderstanding. I won't go into the list of reasons that I think that's not a good idea.

I think it's different, but it depends on how the rating recalculation is actually done. A tie between two people of different skill would bring their rankings closer together. Here, there is no calculation done between those two people.

(Unless I am misunderstanding)

"tl;dr": We don't include bot games on the 2P leaderboard -- what are the reasons for that, and why should they be different than 3P games?

Agreed. Bots are extra-exploitable opponents, so they shouldn't be ranked, so we still shouldn't rank them in multiplayer.

For the joy of finishing out the game, the best option if someone drops is to replace them with a bot. The bot may decide the game between the other two players due to its decisions; there's no avoiding that, it's in the nature of 3-player games. You could have the player drop out completely - or have a bot that just passed - but that's not finishing a 3-player game, that's playing a 3-player game that turned into a messed-up 2-player game. I have played that game! Having a different player (in this case a bot) take over is better.

Replacing with a bot is the solution I would most prefer, as well.

Regarding ratings, I can absolutely see an argument for allowing the complete game outcome after replacement with a bot count. Currently, the bot is very likely to buy Provinces/points when able pushing the game to a conclusion. It's not unreasonable to expect far behind players in multiplayer games to do this to speed things along, too. Would ignoring the final outcome of games with resignations that are finished with a bot give better rankings than not counting it? I don't think that's an obvious "yes", at all.

Seconded. Ideally, this would be implemented along with optional background music in client.

Ah, because the Possession turn is really my opponent's turn and he did not gain the card; I gained it instead?

Nice job, markus! Very interesting.

The first chart in the middle row is the most important one to me. It's what I would look at first when evaluating whether the rating system is doing a good job. It kind of supports this idea that it's risky for a strong player to play against someone rated too much lower because the strong player is more likely to lose than predicted.

I think you have presented some pretty strong evidence that some tweaks are needed.

I was Possessing my opponent. On the Possession turn, I bought Haunted Castle. I was not forced to place two cards on the deck from my normal hand. There was no top decking effect at all.

Game #3501373.

I analyzed the full leaderboard at Scavenger (http://dominion.lauxnet.com/leaderboard/) and I noticed that the initial phi=2 seems to be chosen too high....

Therefore, I'd suggest to lower the parameter to phi=1 for new players and also to cap phi there....

After a couple of months, it would actually be possible to estimate the parameters for initial phi, sigma, and tau, that give the best results in predicting game outcomes....

My preferred way would be to use expected win probabilities and let players set a range. The advantage would be that a certain winning probability is more understandable for the layman than some level difference.

I agree wholeheartedly with this.

A new player is going to be really confused when they play Lookout and don't get to trash a Copper, then discard a Copper, then topdeck a Copper. I've been playing with this client for months and it's still jarring to me when that happens.

It's very, very clear to me that the option to have trivial decisions made for you should be off by default and I'm willing to stand behind that position if I need to.

Definitely, it's a confusing thing and new players should not experience it.

Agreed. Except, I am not a new player, and I don't think I should have to look in the log to see what happened after playing a card, either. Lookout's a great example. Did it reveal three Curses? three Coppers? three Villages? three Provinces?

I think some sort of alternative positive feedback for playing bots would be a good idea.
Some ideas:

1) Campaign progress
2) Recording your wins / losses
3) Bot only leaderboard
4) Win X games, unlock access to a promo card
5) In game achievements, Councilroom style.

I would love 1, 2, and 5 to be implemented.

I'm adding this thought to the "general discussion" of leaderboard: I would allow to have rated games against the bot, if they are drawn randomly (like in "find a game"). The bot would then also get a rating - and could be shown on the leaderboard.

I don't think I like this idea.

....

I agree with Stef. Ratings are primarily used to improve matching when seeking games. There is a danger that including the AI as a player would mess this up more than it helps right now (As Stef mentions, we have seen this in the past).

I think my list remains unchanged.

1) Make game playable without looking  at -- or at least clicking in -- the log.
2) Implement more animations, visual cues, and better graphic as to make 1) possible and the game overall more like a complete game.
3) Implement solutions like dedicated apps in order to be able to play on more mobile devices.

My ideal solution is in line with what Stef is suggesting. Sub an AI in when someone resigns.

For a two player game this would both
1) immediately award you the win and
2) let you finish your turn/keep playing as long as you like.

If you want to quit the game before finishing yourself or AI wins, instead of "resign", the resign button would read "end game" and still award you the win for rating purposes.

For three or more players, the resigning player would be put in last place, an AI would be subbed in, and the game would continue.

This is in line with how many online card games and casual games have handled this type of situation for years. Many sites (e.g. Yahoo games) would also allow real players to jump into open tables to replace AI players at any time, but I would not recommend this for Dominion.