Author |
Topic: player of the month (Read 11663 times) |
|
PMertens
Forum Guru
Arimaa player #692
Gender: 
Posts: 437
|
 |
Re: player of the month
« Reply #15 on: Mar 2nd, 2005, 2:01am » |
 Quote  Modify
|
Quote:| On the other hand, it does encourage people to play the bots. Is that good? |
|
it does encourage to play them in a known way - and that is not good.
I think that nobody will disagree if I say that there is a certain way to beat bomb every game 
(and since there are so many bots to play only few players would take the risk to do st. new)
Playing speedy is a totally different story, since that thing is evolving and hopefully gets harder and harder.
(Same for other evolving bots)
|
|
 IP Logged |
|
|
|
Fritzlein
Forum Guru
Arimaa player #706
Gender:
Posts: 5928
|
|
Re: player of the month
« Reply #16 on: Mar 2nd, 2005, 12:17pm » |
Quote Modify
|
Since there are about 3 copies of every bot, it makes perfect sense that wins against humans should count for 3 points and wins against bots only one point.
Mouse's proposal that it should count for more points to beat a player rated higher than onesself is interesting, although it makes it harder for high-rated players to accumulate points, since there won't necessarily be any even-higher-rated folks available for them to play against. Perhaps it is good to level the playing field in this way, because player of the month is currently vastly easier for stronger players to win.
An alternative way to level the playing field would be for the wins in the winning streak to each count for as many points as the rating system thinks you had a percentage chance of losing. If I beat Bomb2004, the system says I have a 13% chance of losing, so that win would only get me 13 points, but if I play 99of9 and win I would get 51 points since I have a 51% chance of losing. (Actually much higher, but that's what the ratings say.) It would still be easier for stronger players to win player of the month, but not by much, since lower-rated players could collect points against weak bots which would give almost no points to the stronger-rated players.
Probably if this idea is used, it should be coupled with the idea of weighting human games three times more, so that ones rating relative to the rating of the bots isn't the primary determining factor.
|
| « Last Edit: Mar 2nd, 2005, 10:00pm by Fritzlein » |
IP Logged |
|
|
|
PMertens
Forum Guru
Arimaa player #692
Gender:
Posts: 437
|
|
Re: player of the month
« Reply #17 on: Mar 2nd, 2005, 8:37pm » |
Quote Modify
|
it would be interesting to hear Omars opinion 
|
|
IP Logged |
|
|
|
mouse
Forum Senior Member
Arimaa player #784
Gender:
Posts: 45
|
|
Re: player of the month
« Reply #18 on: Mar 3rd, 2005, 10:11am » |
Quote Modify
|
on Mar 2nd, 2005, 12:17pm, Fritzlein wrote:Mouse's proposal that it should count for more points to beat a player rated higher than onesself is interesting, although it makes it harder for high-rated players to accumulate points, since there won't necessarily be any even-higher-rated folks available for them to play against. Perhaps it is good to level the playing field in this way, because player of the month is currently vastly easier for stronger players to win.
|
|
Thats why I would also give extra points if you play someone with a little lower rating.
If you want more than a few players to try for the player of the month you will have to level the playing field. For the momment only players who can beat bomb on a regular basis will have a chance with all the bomb versions.
on Mar 2nd, 2005, 12:17pm, Fritzlein wrote:
An alternative way to level the playing field would be for the wins in the winning streak to each count for as many points as the rating system thinks you had a percentage chance of losing. If I beat Bomb2004, the system says I have a 13% chance of losing, so that win would only get me 13 points, but if I play 99of9 and win I would get 51 points since I have a 51% chance of losing. (Actually much higher, but that's what the ratings say.) It would still be easier for stronger players to win player of the month, but not by much, since lower-rated players could collect points against weak bots which would give almost no points to the stronger-rated players.
|
|
There is a extra advantage with this method it will vastly decrease the likelihood of a two players ending with the same number of points at the end of the month.
And I am not sure this method will actually give a big advantage to players with a high rating. A player with a rating arround 1500 who can get a win against Arimaanator will have a very good chance of winning the player of the month with this scoring scheme.
|
|
IP Logged |
|
|
|
omar
Forum Guru
Arimaa player #2
Gender:
Posts: 1003
|
|
Re: player of the month
« Reply #19 on: Mar 3rd, 2005, 6:16pm » |
Quote Modify
|
It's been about 2 years since I started the player of the month contest. At the time we didn't have many humans or bots so even getting 5 or 6 people on your list was a challenge Thing have definitely changed since then and I think it might be time to redefine the player of the month contest altogether. I'd like the new definition to meet the following goals:
* Any player should be able to win the contest regardless of how strong they are
* It should encourage more human vs human games
* It should encourage more completed games
* The rules should be fairly simple and easy to understand
The simplest thing I can think of which meets these goals is: the player who has played the most humans (unique) regardless of whether they won or lost the game, wins the contest.
To discourage someone from playing a game and resigning quickly just to add another person to their list, lets not add the game to a players list if the person resigns or loses on time (still counts if the opponent resigns or loses on time).
But because some people are having problems with the flash client crashing or network problems maybe we should not count the games where either player lost on time.
Of course only rated games count and the game must also begin and end in the same month. In case of a tie the player who played more recently wins.
|
|
IP Logged |
|
|
|
Fritzlein
Forum Guru
Arimaa player #706
Gender:
Posts: 5928
|
|
Re: player of the month
« Reply #20 on: Mar 4th, 2005, 8:01am » |
Quote Modify
|
Now that you lay out your goals, Omar, it makes sense to completely alter the structure of player of the month. However, I am leery of simply counting the number of distinct humans played. There is no incentive to play one's best. On the contrary, there is a strong incentive not to prolong a struggle, because the quicker you lose, the quicker you can challenge another human to get another point. Resigning and losing on time are not the only ways to lose by giving up; one can simply play randomly with no thoughts of strategy, on intentionally play poor moves.
At a minimum I'd add an additional incentive for winning. For example, give one point for every human you have played in the month but not beaten, and two points for every human you have beaten. To a certain extent, quantity can overcome quality. Everyone still has a shot at winning, and the most active players have the best shot. On the other hand, winning games still matters, and the player of the month won't necessarily go to the person who can throw the most games most quickly.
|
|
IP Logged |
|
|
|
PMertens
Forum Guru
Arimaa player #692
Gender:
Posts: 437
|
|
Re: player of the month
« Reply #21 on: Mar 4th, 2005, 9:14am » |
Quote Modify
|
I agree somehow with Fritz
I guess we count only the first unique hh game ?
Or is it more like 1 point for every uhhg and 1 extra point for winning in at least one of the games against that particular human ? (or would that backfire and prevent from playing a second game that month since I could try not to give him a chance to get his second point ...)
Should silver/gold be taken into account since statistics tell us a slightly higher chance of gold winning ?
2 Points for winning would of course be much better for the top players, since they need half the games of a beginner - but then how do you define player of the month (in Omars version the player could have <1k points and lose just every game - well, that could still happen with the 2:1 points but less likely)
|
|
IP Logged |
|
|
|
PMertens
Forum Guru
Arimaa player #692
Gender:
Posts: 437
|
|
Re: player of the month
« Reply #22 on: Mar 4th, 2005, 9:16am » |
Quote Modify
|
oh and by the way:
the player who plays most recently makes again no sense omar - that lead to Fritzl playing bomb last minute and finishing just 5 minutes to late - and there was no point for me in playing another game before him.
should be the other way round
|
|
IP Logged |
|
|
|
Fritzlein
Forum Guru
Arimaa player #706
Gender:
Posts: 5928
|
|
Re: player of the month
« Reply #23 on: Mar 11th, 2005, 4:12pm » |
Quote Modify
|
on Mar 3rd, 2005, 6:16pm, omar wrote:
* Any player should be able to win the contest regardless of how strong they are
* It should encourage more human vs human games
* It should encourage more completed games
* The rules should be fairly simple and easy to understand
|
|
Omar,
I've been thinking more about your goals. I don't see a way to meet all four of them at the same time: you can only have three of the four. Your proposal to count the number of distinct humans played is simple, encourages HvH games, and can be won by anyone, but doesn't encourage completed games, at least not in the sense of trying hard the whole way. There is no incentive for committed play.
I have a proposal that meets a different three of your goals: it isn't totally simple, but anyone can win, it encourages HvH games, and it encourages completed games. Here's the idea:
For each human you beat during the month, you score the root of your odds of losing to them according to the ratings. A table of examples:
| rating diff | odds on favorite | favorite gets | underdog gets | | 0 | 1:1 | 1 | 1 | | 241 | 4:1 | 1/2 | 2 | | 382 | 9:1 | 1/3 | 3 | | 482 | 16:1 | 1/4 | 4 |
So if I went around playing people rated 482 points below me, I would need to beat 16 of them to get 4 points in the contest, but if any one of them beat me, that one person would get four points in the contest from that one victory. The risks are balanced, assuming the ratings are correct (which they aren't; I'm overrated from playing bots, but that's my fault )
I propose not counting bots, not counting losses, and not counting repeat victories. There would never be a risk for playing new people, as there is under the current system when you have to put your winning streak on the line. You just try to beat as many distinct humans in the month as possible.
I think this would really give folks an incentive to mix it up, and to try hard every game. Because of the weighting of rewards, truly anyone can win, and your standing in the contest gets better and better (on average) the more distinct humans you play. Whoever plays the widest variety of human opposition has the best chance to win, regardless of their rating.
|
| « Last Edit: Mar 11th, 2005, 4:16pm by Fritzlein » |
IP Logged |
|
|
|
omar
Forum Guru
Arimaa player #2
Gender:
Posts: 1003
|
|
Re: player of the month
« Reply #24 on: Mar 11th, 2005, 6:36pm » |
Quote Modify
|
I guess, someone could play to goal, but not really try hard. I was thinking that since these would be rated games, people wouldn't do that to protect their ratings. But maybe winning the contest could become more important than maintaining the rating. Or someone might get the idea of creating a second account to use for the contest games; not good.
I like your proposal, Karl. Would it still be OK if we did it in a discrete way like:
Rating difference, favored gets, underdog gets
450 or more, 1, 16
350 to 450, 1, 9
250 to 350, 1, 4
0 to 250, 1, 1
Maybe we should also consider the side and thus allow two unique games against the same opponent where the sides are different.
I see PMertens point about the tie breaker. If you only acheive the same score then the person who a acheived it first should still be considered the winner.
OK let me throw out another proposal for how to score the player of the month contest. How about winner gets 3 points and loser gets points based on how long they could defend: 3 points if the game is 60 moves or more; 2 points if the game is 40 moves or more and 1 point if the game is 20 moves or more. I think this would encourage a person to not lose and at least try and defend as long as possible. The one who plays the most unique human games this way would win the contest. Though a good player could win with fewer games a weaker player might be encouraged to win the contest by playing more games because after each game they see their score going up (most anyone can defend for 20 moves if they try).
|
|
IP Logged |
|
|
|
Fritzlein
Forum Guru
Arimaa player #706
Gender:
Posts: 5928
|
|
Re: player of the month
« Reply #25 on: Mar 11th, 2005, 7:49pm » |
Quote Modify
|
on Mar 11th, 2005, 6:36pm, omar wrote:
OK let me throw out another proposal for how to score the player of the month contest. How about winner gets 3 points and loser gets points based on how long they could defend: 3 points if the game is 60 moves or more; 2 points if the game is 40 moves or more and 1 point if the game is 20 moves or more. |
|
This new proposal gives an incentive to try harder, but now favors stronger players again, so you are still getting only three of your four goals. My more complicated proposal should be exactly fair to stronger and weaker players (assuming those mythical accurate ratings exist, of course).
Quote:
I like your proposal, Karl. Would it still be OK if we did it in a discrete way like:
Rating difference, favored gets, underdog gets
450 or more, 1, 16
350 to 450, 1, 9
250 to 350, 1, 4
0 to 250, 1, 1
|
|
Sure, we could discretize it, but the way you scaled it means that the most evenly matched games will, on average, count the least, when they should, on average, count the most. I would suggest instead:
0-50 5/5
50-120 6/4
120-240 8/3
240-440 12/2
440+ 24/1
But I'm sceptical that discretizing actually makes it simpler. True, everyone's score in the contest will be a whole number, but by the same token you have to list the ranges and hard code them, etc. rather than using a single formula. It's a matter of taste, I guess.
Quote:
Maybe we should also consider the side and thus allow two unique games against the same opponent where the sides are different.
|
|
I like the idea of allowing two games per pair of players so that there would be an incentive for a rematch for the winner of the first game. On the other hand, I really like keeping the focus on playing as many different humans as possible.
Quote:
I see PMertens point about the tie breaker. If you only acheive the same score then the person who a acheived it first should still be considered the winner.
|
|
I quite agree, but one advantage of using the continuous version of my proposal is that ties won't happen.
|
|
IP Logged |
|
|
|
Fritzlein
Forum Guru
Arimaa player #706
Gender:
Posts: 5928
|
|
Re: player of the month
« Reply #26 on: Mar 11th, 2005, 7:56pm » |
Quote Modify
|
I just thought of a horrible disadvantage to my proposal: it creates an incentive to intentionally lower your own rating, because the lower your rating is, the more points you get for each victory. So in between my games against humans, I could lose a bunch of times to Arimaazilla to keep my rating low, in order to maximize the points I get in the POTM contest. Yuck.
This flaw is so glaring that I think it would, in fact, be better to just count the number of distinct humans played without regard to wins and losses. True, that would make the contest itself dumb, but at least people wouldn't have such an incentive to mess up their own ratings. 
An alternative would be to use Omar's latest idea (3 to the winner and 3, 2, or 1 to the loser) but make it simpler: The winner gets 60 and the loser gets the number of moves in the game, capped at 59.
|
| « Last Edit: Mar 11th, 2005, 8:01pm by Fritzlein » |
IP Logged |
|
|
|
Adanac
Forum Guru
Arimaa player #892
Gender:
Posts: 635
|
|
Re: player of the month
« Reply #27 on: Mar 12th, 2005, 11:53am » |
Quote Modify
|
I much prefer human vs. human games for many reasons but I’m surprised that the player of the month award may not include bot games anymore. Given that arimaa was invented to promote AI research, shouldn’t there be some small incentive to play the bots too? Obviously it’s a waste of time for 2000+ players to play the 2004CCbots but it would be a shame if every brand new player logged onto the site and found themselves swamped with match requests from every player trying to boost their point totals while all the bots sat around idly. Here’s my alternate proposal, which encourages established humans to play one another while not ignoring the bots altogether (it’s very similar to Fritzlein’s original proposal).
3 points for each game played against a distinct human with an RU below 100
1 point for each game played against a computer
1 point for each victory against a distinct opponent.
What do you think of my version of the ‘3 out of 4’ proposal?
On a completely separate issue (which I probably belongs in a separate thread), another problem with the current system is that you can double-count by playing bots at regular and fast time controls, while it only counts once for humans regardless of the speed of play. Is there any plan to eventually have 1 account for each bot with different ratings for different speeds? Of course, I mean after the Postal Championship when more time is available
|
|
IP Logged |
|
|
|
omar
Forum Guru
Arimaa player #2
Gender:
Posts: 1003
|
|
Re: player of the month
« Reply #28 on: Mar 17th, 2005, 12:21am » |
Quote Modify
|
on Mar 12th, 2005, 11:53am, Adanac wrote:
On a completely separate issue (which I probably belongs in a separate thread), another problem with the current system is that you can double-count by playing bots at regular and fast time controls, while it only counts once for humans regardless of the speed of play. Is there any plan to eventually have 1 account for each bot with different ratings for different speeds? Of course, I mean after the Postal Championship when more time is available
|
|
There is a discussion of how to include different speeds into the rating; but it applies not just to bots, but humans also since humans play very differently at different time controls also. I think the thread is:
Omar = OmarFast , bot_bomb = bot
Hummm, should we includes bots with reduced scores as suggested by Adanac or not include them at all. I could go either way on that. I'd be interested to see what others think before we finalize it.
|
|
IP Logged |
|
|
|
PMertens
Forum Guru
Arimaa player #692
Gender:
Posts: 437
|
|
Re: player of the month
« Reply #29 on: Apr 5th, 2005, 10:48am » |
Quote Modify
|
any changes ?
|
|
IP Logged |
|
|
|
|
Home
Help
Search
Members
Login
Register
Reply
Notify of replies
Send Topic
Print