Dominion Strategy Forum

Please login or register.

Login with username, password and session length
Pages: [1]

Author Topic: Slow-Playing Detection?  (Read 2885 times)

0 Members and 1 Guest are viewing this topic.

Dingan

  • Saboteur
  • *****
  • Offline Offline
  • Posts: 1154
  • Shuffle iT Username: Dingan
  • Respect: +1731
    • View Profile
    • Website title
Slow-Playing Detection?
« on: January 22, 2019, 02:55:14 am »
0

Just thinking out loud here .. couldn't you have some sort of slow-playing (intentionally taking a long time on each of your decisions to annoy your opponent and/or encourage them to resign) detection? Thoughts:

- You could measure how long, on average, a player takes to make a decision. If it becomes over some threshold after x turns (x is greater than 2, as in most games I would think the first few turns of the game are relatively long), their opponent can force-resign them.
- If decisions are intrinsically long for a player (it's a high-profile league game, players are playing with unfamiliar cards, etc.), that player can explain in the chat that they are simply having difficulty coming to decisions and need time. It is up to their opponent to respect this explanation.
- Players get n seconds per decision before a force-resign, and m seconds per turn before a force resign. m would be not all that larger than n, i.e. a player cannot just take n seconds a bunch of times per turn (for many turns).
- Players can have a small amount of "slow-play" tokens. I get, say, 1 of these tokens for every 10 games I play. I can spend a token to basically say my opponent is playing slow and I want to divide their time-to-force-resign by 10, or whatever.
- If I recall correctly, one of the biggest selling points of ShIT over Making Fun was its blacklist feature. And part of that feature was the ability to retroactively remove a often-blacklisted ("often" being over some threshold) player from all ranking algorithms, thereby removing slow-played losses from a player's ranking-contributing games. But as far as I know, this "removal" decision is not public. How do I know a horrible slow-player has been flagged enough? And can I mark them as horrible, i.e. different than a player that I blacklisted just due to, say, differences in undo preferences? This extra marking could contribute to the Stef-investigates-said-player more than the normal blacklist flag.

Now that I write this all out and think about it, it's actually seems like a pretty interesting problem to solve. It makes sense that it's not solved yet. How do other online games address it? Is there anything else we can do to improve the online experience?
Logged

ipofanes

  • Saboteur
  • *****
  • Offline Offline
  • Posts: 1439
  • Shuffle iT Username: ipofanes
  • Respect: +776
    • View Profile
Re: Slow-Playing Detection?
« Reply #1 on: January 22, 2019, 03:29:18 am »
+1

In online Chess or Go, time control is normally exerted by variants of total time and time per move (normally added to total time). In Shuffle IT, time control applies to the maximum time per move. This is useful to penalize people who rage quit or got distracted  by a different browser tab or by the world out there. In pre-digital age, there were experiments on controlling time differences (by a pebble in a viscuous fluid that, by pressing the clock, would be tilted away from the player currently taking their time. I think a time difference mode (with optional force resign still in place) would solidly weed out slow-playing individuals but would also affect slow thinkers, who I confess to be one of (at least have been called out for it several times by juvenile-sounding opponents). A time-per-move profile as an additional matching criterion might be a good idea.

I think force-resign after a total time difference of 5 minutes, combined with playing speed matching, would tackle most of the problems.
Logged
Lord Rattington denies my undo requests

Jeebus

  • Margrave
  • *****
  • Offline Offline
  • Posts: 2529
  • Shuffle iT Username: jeebus
  • Respect: +1642
    • View Profile
Re: Slow-Playing Detection?
« Reply #2 on: February 20, 2019, 01:17:44 pm »
0

This got me thinking. Isn't the behavior of slowplaying that the player uses almost max time, or at least a big percentage of max time, for each and every click? I would think it would be a pretty easy thing to identify.

I don't really like a time difference between the players. This could give false positives with very different playing styles, especially in long games (although I guess the latter could be addressed with a percentage instead of minute count).

But something along the lines of Dingan's first suggestion should work. I think the allowed time per click currently is 4 minutes. For instance, let's say you get a "mark" if you take more than 3 minutes per click 3 times in a turn (or every time if there are less than 3 clicks in the turn). You can't get marks the first two turns in a game. If you get marks for two consecutive turns, you can be force-resigned. To make sure nobody games the system by slow-playing only every other turn, the total number of marks should also matter. For instance, after 4 marks you can be force-resigned. Hmm, one problem could be that slow-playing < 3 minutes still works. To solve that, taking more than 1 minute per click 4 times in a turn could earn you half a mark; this would count towards your total only.

Another great thing about this system would be that if a player is actually force-resigned based on "marks", and this happens a certain number of times, they could get auto-banned. No involvement from Stef or anybody else is needed.
« Last Edit: February 20, 2019, 01:21:21 pm by Jeebus »
Logged

Cave-o-sapien

  • Jester
  • *****
  • Offline Offline
  • Posts: 887
  • Respect: +1676
    • View Profile
Re: Slow-Playing Detection?
« Reply #3 on: February 21, 2019, 04:23:31 pm »
+1

If there is a solution, I think it needs to be invisible to the players. Otherwise you've just traded one meta-game for another.

Some kind of adaptive timer seems best.
Logged
Pages: [1]
 

Page created in 1.721 seconds with 20 queries.