Author |
Topic: Move 28 (Read 5223 times) |
|
woh
Forum Guru
Arimaa player #2128
Gender: 
Posts: 254
|
 |
Re: Move 28
« Reply #30 on: Mar 26th, 2008, 7:32am » |
 Quote  Modify
|
With the results the way they are, I would put option 2 on top.
on Mar 25th, 2008, 5:42pm, Fritzlein wrote:| Also using the dumb tiebreaker of summing all the numbers in the row ... |
|
If you use this between the results of the top three only then the second choise is the winner.
Using Schulze between the top 3 also favors the second option.
|
|
 IP Logged |
|
|
|
aaaa
Forum Guru
Arimaa player #958
Posts: 768
|
|
Re: Move 28
« Reply #31 on: Mar 26th, 2008, 8:12am » |
Quote Modify
|
I would like to point out that Schulze is not supposed to be run again with candidates eliminated as it would violate the monotonicity criterion, unless the eliminated candidates are all outside the Smith set.
So I would advise a revote between all members of the Smith set.
Also, if you're going to break a tie randomly, do it by repeatedly choosing a random ballot before picking a random candidate.
|
|
IP Logged |
|
|
|
RonWeasley
Forum Moderator
Forum Guru
Harry's friend (Arimaa player #441)
Gender:
Posts: 882
|
|
Re: Move 28
« Reply #32 on: Mar 26th, 2008, 11:15am » |
Quote Modify
|
Fortunately the top three candidates constitute the Smith set for this election.
For TheMob members like me who haven't looked this up before, it's the smallest set of candidates that are preferred over every candidate outside the Smith set. In theory it's possible for all candidates to be in this set if you are liberal with my definition. One method for resolving ties would be to have run offs among the candidates of the incrementally decreasing Smith sets until there is a clear winner or no candidates fail to make the Smith set. That would take about an extra day per run off, which seems worth it to me.
Unfortunately we don't have access to the ballots so we have to revert to a different methodology for breaking a tie having an all-inclusive Smith set. A random outcome generator is the fastest means available and I propose choosing from only the "tied" candidates given by the voting site (instead of the entire list).
|
|
IP Logged |
|
|
|
Fritzlein
Forum Guru
Arimaa player #706
Gender:
Posts: 5928
|
|
Re: Move 28
« Reply #33 on: Mar 26th, 2008, 12:08pm » |
Quote Modify
|
on Mar 25th, 2008, 11:06pm, mistre wrote:| Definitely a revote. Of the top 2 moves, I have changed my mind on which I like best. |
|
I actually reversed my preference among these three votes, swapping the top and bottom. In my case a re-vote is essentially equivalent to randomizing a ballot. 
At least the three options are close enough in quality (as far as I know) that I will be happy no matter which one wins.
|
|
IP Logged |
|
|
|
RonWeasley
Forum Moderator
Forum Guru
Harry's friend (Arimaa player #441)
Gender:
Posts: 882
|
|
Re: Move 28
« Reply #34 on: Mar 27th, 2008, 7:16am » |
Quote Modify
|
And our winner is ee3w df7s df6s df5e! A close one. Thanx for voting.
|
|
IP Logged |
|
|
|
Fritzlein
Forum Guru
Arimaa player #706
Gender:
Posts: 5928
|
|
Re: Move 28
« Reply #35 on: Mar 27th, 2008, 7:26pm » |
Quote Modify
|
It is impossible to tell without individual ballot reporting turned on, but there is a fair chance the winning option didn't get 6 of 11 first place votes. It probably won by virtue of being a compromise option between saving the rabbit and attacking with the dog.
. . . . . . . . . 1. 2. 3.
1. ee3w df7s df6s df5e (e->d3, d->g5) . . - 7 6
2. df7s df6s df5s df4s (d->f3) . . . 4 - 6
3. ee3w hd6w hc6w df7s (e->d3, h->b6, d->f6) 5 5 -
|
|
IP Logged |
|
|
|
aaaa
Forum Guru
Arimaa player #958
Posts: 768
|
|
Re: Move 28
« Reply #36 on: Mar 27th, 2008, 9:04pm » |
Quote Modify
|
What does Bomb say? Does the software allow you to determine which of those three moves it prefers?
|
|
IP Logged |
|
|
|
Fritzlein
Forum Guru
Arimaa player #706
Gender:
Posts: 5928
|
|
Re: Move 28
« Reply #37 on: Mar 28th, 2008, 7:58am » |
Quote Modify
|
on Mar 27th, 2008, 9:04pm, aaaa wrote:| What does Bomb say? Does the software allow you to determine which of those three moves it prefers? |
|
At 16 steps depth (4 ply) Bomb would have preferred to play ee3s ee2e Cg2n ef2e, with an evaluation of +0.75 for Silver.
There is no way to tell which of the three moves we voted among Bomb would most prefer except by trying out each of them and letting Bomb think to an equal depth in each case. If we let Bomb think to a depth of 12 steps (3 ply) in each case, its evaluation should be comparable to the evaluation of its preferred move, so we can tell how much of a mistake Bomb thinks we are making.
+0.14 ee3w df7s df6s df5e
+0.11 ee3w hd6w hc6w df7s
+0.01 df7s df6s df5s df4s
So at that particular depth, Bomb likes the move we chose the best of the three moves in the runoff, but thinks we made an error of magnitude 0.61, more than half a rabbit, by rejecting ee3s ee2e Cg2n ef2e in the first ballot. We had Bomb's preferred move in seventh place of our eight options.
I was about to make the obligatory comments about Bomb having no clue in general, and particularly no clue in a quiet position like this one, but to be fair, we have no clue either. I would guess that giving up our rabbit and playing for a cat hostage is the wrong idea in a position like this, because it invites chessandgo to regroup and eventually swarm. I would suspect that with MHR gone from each side, we'd rather be rolling forward than pushing a hostage to our side. But who knows? My own understanding of the position is too weak to poke fun at Bomb too vigorously in this circumstance.
|
|
IP Logged |
|
|
|
Fritzlein
Forum Guru
Arimaa player #706
Gender:
Posts: 5928
|
|
Re: Move 28
« Reply #38 on: Mar 28th, 2008, 8:18am » |
Quote Modify
|
Note that Bomb's decision to go after the cat hostage is strategic, not tactical. After the move sequence
28b ee3s ee2e Cg2n ef2e
29w rb3e rc3x Db4s Ec4e Rf1n
29b Cg3n eg2n Cg4n eg3n
30w Rg1n Ed4e Ee4n Ee5n
30b hd6w hc6w eg4w ef4n
31w Dd2e De2n De3e Df3e
31b df7s rg6e df6e rh6s
Bomb thinks to depth 12 steps and concludes that Silver is still ahead by +0.68. It wasn't just that Bomb was pushing the loss of Silver's rabbit over the horizon, or seeing some tactical compensation. Bomb really likes that cat hostage strategically, which I don't quite agree with in a position like this.
|
| « Last Edit: Mar 28th, 2008, 8:21am by Fritzlein » |
IP Logged |
|
|
|
|
Home
Help
Search
Members
Login
Register
Reply
Notify of replies
Send Topic
Print