June 26 2002 RussCon Report

News Game Results Meta Party Game Results Evening's Soundtrack Ratings Links 13 of us played 15 games. Low turnout: perhaps people were frightened by thinking they'd have to crunch their scores in Dudley's number-happy Meta Game! RussCon Mailing List Subscription Info RussCon Report Archives Previous Report Next Report RussCon Info Russ Zone Check out my webcomics. Comicollage: daily ongoing comic jam with 7 artists and no planning ahead! There's a russ strip every Tuesday. I make exquisite corpses at deepmachines.

Game Results

I had an intuition that Roadkill might suffer from what I call "Avalon Hill disease" (wherein what should be a simple elegant or quick silly game gets unfortunate rules and playing length as if it were a hardcore wargame). And based on the ongoing comments (Whendy: "This is a horrible game!"; MarkY: "I could be working now..."; JP wanting to play Cosmic Coasters concurrently with Roadkill; JP & MarkY apparently actually playing a game of Cathedral concurrently; etc.) I'm glad I sat out doing some go life and death problems till Clayton showed up... The irony was that they were playing the "short" scenario.

Carcassonne is still a reasonable game 2-player, though I think it's better with more, just because shared points are common in Carcassonne and thus become less interesting in 2-player.

Babel on the other hand is of course designed as a 2-player game, and it had been a while since we'd played. I really like it. And I'm convinced a crucial bit of strategy is to try to get ahead in number of cards early on, ideally by stealing cards from your opponent. That seems to always work for me. There's one minor annoying rules thing, that they don't make it real clear that on the first turn the first player can't just grab both of the "1" temples. Buried somewhere is a mention that each player starts with a "1" which they must use on their first turn (which implies the first player can't grab them both, else the second player couldn't use the "1" they started with...)

Drakon 2-player is quite fun and different from multiplayer. Of course there's no messy kingmaking issues. Dudley probably should have won the first game, but he brainfarted and didn't mind-control me when I was on a teleporter. Since he had 3 gold to my 1, I decided to be a stickler. :) In the second game I managed to get Dudley down to 0 tiles (and thus I had 8), a wonderful advantage, and then at one point we were both on a teleporter. I should have teleported to a destructor tile and killed the last tile-stealing tile, but foolishly didn't, so Dudley teleported to the tile-stealer and got a tile back... having even one tile available makes a huge difference. We later got into a funny potential infinite loop where we kept stealing a gold back and forth from each other. It was not clear that doing something else would be better, so I was worried we needed a ko rule to break the cycle, or maybe we'd have to agree to a draw. So I played a shift tile and left the loop just because, and managed to end up winning. Also, Drakon doesn't take as long with only 2 players (not surprisingly), though the 2nd game wasn't short.

I have received a special request from MarkY to report his glorious victory over William in Can't Stop, which was apparently partly enabled by JP, depending on whom you listen to (if you can listen to anyone over William's loud anguished wails and caterwaulings!)

Evening's Soundtrack

April's Motel Room (5)
The Moody Blues, Days of Future Passed (7)
Louis Armstrong, What a Wonderful World (11)
Iggy Pop, Brick By Brick (13)
Alanis Morissette, Supposed Former Infatuation Junkie (17)

Prime numbers of tracks on these CDs! As far as I know, no one detected this theme.

Ratings

Rank ratings:
 0.8000 Daniel (2 games played)
 0.6000 Steve (5)
 0.5556 Dan (4)
 0.4286 RussW (8)
-0.1538 JP (4)
-0.2000 Brady (2)
-0.2353 MarkY (6)
-0.2353 William (6)
-0.4000 Dudley (6)
-0.8000 Fred (2)
-1.0000 Marty (1)
-1.0000 Whendy (1)
-1.0000 Clayton (2)

New Win ratings:
 0.7333 Steve (5)
 0.5000 RussW (8)
 0.3000 Daniel (2)
 0.2222 Dan (4)
-0.1538 JP (4)
-0.1765 MarkY (6)
-0.2000 Brady (2)
-0.2353 William (6)
-0.2500 Whendy (1)
-0.4000 Fred (2)
-0.5000 Marty (1)
-0.6000 Dudley (6)
-1.0000 Clayton (2)

Steve is the Devil. Daniel is Vice-Devil. I am Vice-Vice-Devil. Dan is Vice-Vice-Vice-Devil. Marty and Whendy pulled Reverse Brady Maneuvers, and Clayton is the Generous Angel. I win the Dedicated Award.

Exciting news! I have implemented a new ratings system, motivated by the fact that I won 7 of 8 games tonight yet did not top either list. :) Recall that back in March I had some ideas about trying to capture the multiplicative flavor of probabilities of independent events. The existing ratings systems adds your devil points, so if you win a lot of 2 player games (as I did this evening) you only get 1 devil point each time. My idea was to think in terms of the probability that you'd do as well as you did: there's a 0.5 chance you'd win a 2-player game, and a 0.25 chance you'd win 2 2-player games. I.e. winning 2 2-player games should have the same effect as winning 1 4-player game. (In my opinion of course.) Thus the probability that you would do as well as you did could literally be your rating, i.e. the lower your rating the better: it expresses the probability that you'd do as well as you did. But doing no worse than last place is probability 1, which wouldn't hurt your rating to multiply it in. I didn't want you to be able to improve your rating simply by playing lots of games and merely not coming in dead last! I futzed with various complicated schemes to attempt to normalize it based on the expected values of devil points, but that got into very murky territory (e.g. ties are impossible in some games and not in others, which affects the distribution of possible devil points earned).

Eventually I hit on the simple notion of scrapping the idea that the rating would end up literally being a probability. After all, I don't really care about performing arithmetic on the final ratings; all I care about is getting a linear order. Instead it would be centered at 1.0, and for any game you earned negative devil points in, we will divide by the probability that you'd do no better than that (instead of multiplying by the probability that you'd do no worse than that). E.g. in a 2 player game, the winner is multiplied by 0.5 and the loser is multiplied by 2. In a 3 player game, the winner is multiplied by 1/3 and the middle guy is unaffected and the loser is multiplied by 3. Any game that you earn 0 devil points in has no direct affect on your rating. This preserves a pleasingly elegant sort of symmetry (albeit not as simple as the old scheme). Generalizing to teams etc. isn't hard and is left as an exercise for the reader.

Since we're used to higher ratings being good, I finally take the reciprocal of the final result. Here's what this week's ratings would be:

New Multiplicative Rank Ratings: 82.2857 RussW (8 games played) 35.0000 Steve (5) 17.5000 Daniel (2) 10.0000 Dan (4) 0.6000 Brady (2) 0.5333 JP (4) 0.4286 William (6) 0.3333 Marty (1) 0.2500 Clayton (2) 0.2009 MarkY (6) 0.2000 Whendy (1) 0.1500 Fred (2) 0.0625 Dudley (6)

New Multiplicative Win Ratings: 96.0000 RussW (8) 70.0000 Steve (5) 2.5000 Daniel (2) 2.0000 Dan (4) 0.5000 Marty (1) 0.5000 Whendy (1) 0.2500 Brady (2) 0.2500 Clayton (2) 0.2500 Fred (2) 0.2500 JP (4) 0.0938 MarkY (6) 0.0625 William (6) 0.0156 Dudley (6)

This seems more in keeping with my intuition (I won 7 of 8 games; Steve won 4 of 5 games; Daniel won 1 of 2 games). Note also that this system will frequently automatically handle situations like ties when several players win all their games, or when someone pulls a Brady Maneuver, since in the old scheme, all such players earned a rating of 1 (the maximum), whereas in this system, they will earn points equal to the product of the number of players in all the games they played and won, which typical will end up being different numbers. The player who beat the odds by winning more games will have more points. E.g. if Biff wins a single 3-player game he'll have a rating of 3. If Eugene wins 2 2-player games, he'll have a rating of 4. Likewise total losers will be differentiated by the combined probability of doing as poorly as they did, rather than all having equal ratings of -1 as in the old system; in the new system their ratings will more clearly reflect how lame they are relative to each other. Woohoo!

To put it another way, this ratings distribution is asymptotic between 0 and infinity rather than being in a closed interval from -1 to 1.

Note that in the old system 0 is the center. In the new center, 1 is the center. I.e. if your rating is higher than 1 then you can feel like you "beat the odds" that evening.

Feedback encouraged! Unless someone points out some nonobvious flaw, I plan to start using this system next week.

Links

1000 Blank White Cards is a game we should try at RussCon!

Ludum Dare's 48-Hour Game Programming Competition

Cute games

Villain supply

Museum exhibits Russia's cold war secrets

A Brief History of Negative Space (boardgame-related fiction by Chris Nakashima-Brown)

Jerry Falwell worries the pledge of allegiance ruling will alienate God: "at a time when America is at peril, like right now, we do not need to alienate God"... the controversy over this decision makes me feel WAY out of step with mainstream America...

i am not a number

Russ