Wed March 11 RussCon report:
We had another large turnout: 12 people and 3 tracks of gaming! Attendees included Alfred, Bob, Brady, Eric, JP, Ken, Kevin, Marty, Peter, Rick, RussD, and myself (RussW).
Leftover pizza was generously supplied by Kinesoft, a proud sponsor of the 1998 Winter RussCon Games.
5 of us played that Ferro-whatever Pampas game of Alfred's -- it's excellent, even if it does have an unfortunately long name which I can never remember. (We considered calling it "Trains, Plains, and Crayons". Its real name is Ferrocarriles Pampas.) It actually meets my long-time dream of a crayon train game with rational economics instead of the dumb luck of the Empire Builder games. The players are investors in railroads buying shares of stock (each railroad has 5 shares, not 10 like the 18xx games). A very cool game I want to play more.
Ken won Ferro Pampas. Marty introduced Euphrates & Tigris to 3 players and stomped 'em twice. Eric won Settlers twice as is his habit when playing Settlers with Peter. (Peter, you need to start insisting on Loewenherz not Settlers!) Finally the 8 remaining folks played a single big game of Settlers (Peanut! Peanut!) and RussW won.
The 8-player settler game was completed under protest (duly noted) by Eric who fell victim to an honest misunderstanding (or was it...? heh-heh!) on Peter's part. Eric: "I said put the robber on the 6 rock, you said 'The 6?', and I said yes." Peter: "You said put the robber on the 6, I said 'This 6?', and you said yes." A turn later Eric was not allowed to put the robber where he'd really wanted it, which just goes to show that when you have 9 points in Settlers, people can be very unforgiving! Especially when you were mean to them about fixing something earlier in the same game...
It was a first time Peanut map for some players, and the first time since the Millenium game con for Marty and me (where Marty won, snagging a prize copy of Manhattan if memory serves). This was also the first 8 player game since I've been keeping statistics.
I decided to do ratings for this individual RussCon for fun. As always, take 'em with a few grains of salt and several tongues in cheek. These ratings are not intended to be professional advice. Consult your broker or accountant before using them. Offer not valid in Tennessee.
I supply rank ratings and winner ratings, since different people have different preferences for maximizing their rank or maximizing their wins.
Rank ratings for the evening:
The formula is (#you beat - #beating you) / #opponents.
RussW 9/11 = 0.81
Marty 7/13 = 0.54
Ken 5/11 = 0.45
Eric 5/11 = 0.45
JP 1/11 = 0.09
Kevin 0/6 = 0.00
Bob 0/6 = 0.00
Peter -3/11 = -0.27
Brady -5/11 = -0.45
RussD -2/4 = -0.5
Alfred -11/11 = -1
Rick -6/6 = -1
By the way, when 2 people have equal rating, I've been sorting by the player with higher #opponents, #games, or RussCons attended, thus rewarding frequent players. (And don't gripe that this rewards people for playing lots of short games! If someone is choosing which game to play based on favorably breaking ratings ties instead of which game they actually want to play, they need professional help.)
Win ratings for the evening:
This is the first winner ratings calculation I've given in a while (since the formula is more complex). It illustrates some of the properties.
E.g. Marty and Eric both lost the 8 player game, and both of them won their 2 other games, but Marty's 2 other games were 4 player while Eric's were 3 player, so Marty ends up with a higher rating than Eric. That should agree with most people's intuition.
RussW won the 8 player game and lost the 5 player game while Ken won the 5 player game and lost the 8 player game. Ken's rating is higher. Does it feel "right" that Ken's rating is therefore higher? Different people have different intutions. See earlier long RussCon mailing for the derivation and justification of this formula. (Copies available on request for new RussCon list members.)
The formula is:
winner's game points = (n - #winners) / n
loser's game points = -#winners / n
(Normally #winners = 1, but isn't necessarily; some games allow joint victory.)
player rating = Sum(game points) / Sum(|game points|)
Marty = 0.85 = (3/4 + 3/4 - 1/8) / (3/4 + 3/4 + 1/8)
Eric = 0.83 = (2/3 + 2/3 - 1/8) / (2/3 + 2/3 + 1/8)
Ken = 0.73 = (4/5 - 1/8) / (4/5 + 1/8)
RussW = 0.63 = (-1/5 + 7/8) / (1/5 + 7/8)
Everyone else won no games this evening and gets win ratings of -1.
More devil point theorizing:
So the same 4 people placed in the top of both lists, but in different orders (reflecting the different concepts of final rank vs # wins).
How would you pick the overall winner if this were a tournament?
You could just use rank ratings:
You could just use win ratings:
You could (recursively!) look at the combined rankings on both ratings chart:
Marty 3 (2nd + 1st)
Russ 5 (1st + 4th)
Ken 6 (3rd + 3rd)
Eric 6 (4th + 2nd)
One could also consider your ranks in the 2 types of ratings as separate scores and use your best rank, with worst rank as tie-breaker:
Marty (1st, 2nd)
Russ (1st, 4th)
Eric (2nd, 4th)
Ken (3rd, 3rd)
Or do that Euprates and Tigris style (considering your worst rank first, with best rank as tie-breaker):
Marty (2nd, 1st)
Ken (3rd, 3rd)
Russ (4th, 1st)
Eric (4th, 2nd)
One could also observe that Marty beat Ken & Eric in both rating lists, whereas no one else beat anyone in both categories. Thus clearly Marty did better than Ken & Eric, but Russ, Ken and Eric are not clearly differentiated. This makes Marty the overall winner, and Ken, Eric, Russ all tie for 2nd.
Or maybe Russ is 2nd while Ken & Eric tie for 3rd, since Russ wasn't clearly beaten by Marty? I.e. use #clearly beaten - #clearly beating:
This method seems likely to lead to ties too frequently. For those instances (like this) where it doesn't lead to a tie, it certainly seems to be a good arbiter.
You could get meta and apply all 6 of those algorithms, and see who came in first the most by the various methods! Wheels within wheels...
You could also arithmetically combine the 2 ratings, but I think this would be mixing apples and oranges -- I don't think the 2 ratings can be validly directly combined or compared arithmetically. (Even though they're both on a [-1..1] scale, the distributions look different.)
You could sum the 2 statistics:
You could use the max of the 2 ratings:
You could use the min of the 2 ratings (with max as tiebreaker):
Eric 0.45, 0.83
Ken 0.45, 0.73
Cumulative rank ratings over all Wednesday RussCons:
Tom 8/12 = 0.67
EricH 14/27 = 0.52
Jonathan 2/6 = 0.33
Evan 4/13 = 0.31
RussW 23/103 = 0.22
Doug 10/49 = 0.20
James 17/99 = 0.17
Ken 10/73 = 0.14
Tim 1/11 = 0.09
Baron 1/15 = 0.07
Marty 4/66 = 0.06
Bob 0/6 = 0
JP -3/35 = -0.09
Kevin -5/29 = -0.17
Brady -21/96 = -0.22
RussD -10/36 = -0.28
Jay -1/3 = -0.33
Alfred -14/32 = -0.44
Peter -12/27 = -0.44
Rick -13/26 = -0.50
JeffS -3/5 = -0.60
Dawn -4/6 = -0.67
Lance -8/11 = -0.73
Bob gets the Extra Average Award this week. Thanks to Eric for the chance to push his rating down a smidgen. Tom and Jonathan, hope to see y'all soon! Evan reports he may be visiting from the Big Apple in April so perhaps we'll finally get an update to his moldy old rating.
No gaming planned for this Sunday. There was talk of Hispania, but only 3 people interested, so it's a no-go. I'm feeling fatigued and low-energy anyway, so it's probably just as well.
See you 7pm Wednesday March 18!