用户名: 密码:
新用户注册
监管者: Hrqls , coan.net , rod03801 
 BrainKing.com

Board for everybody who is interested in BrainKing itself, its structure, features and future.

If you experience connection or speed problems with BrainKing, please visit Host Tracker and check "BrainKing.com" accessibility from various sites around the world. It may answer whether an issue is caused by BrainKing itself or your local network (or ISP provider).

World Of Chess And Variants (videos from BrainKing): YouTube
Chess blog: LookIntoChess.com


每页的消息:
讨论板列表
您未权限在该板张贴消息。只有最低脑兵级别的会员才允许张贴在该板。
状态: 所有人能发表
帖子搜索:  

<< <   520 521 522 523 524 525 526 527 528 529   > >>
22. 九月 2004, 21:38:39
coan.net 
题目: Re: Ratings Fallacy
... a good example of why possible ratings should be calculated from the ratings when they first started the game, not when the game is complete.

22. 九月 2004, 21:31:44
ughaibu 
How many named departments did KM have?

22. 九月 2004, 21:28:30
Purple 
题目: Ratings Fallacy
Purple修改(22. 九月 2004, 21:29:32)
Not proud of this but The KM Covert Operations Dept. would occasionally torpedo some arogant hi rated dude by sending in a suicide bomber (often rigged up with a program) who would start a game with say a 2000 rating and..before finishing the game with the target..would tank 7-8 games in a row so the rating would be more like 1400. Then arrogant guy would lose to a 1400 player and take a major ratings hit. This was only done to predatory players but it didn't make it right. It is one example however of how ratings should be taken with a grain of salt.

22. 九月 2004, 21:02:29
ughaibu 
Fancy that, someone was rated above GothicInventor, if that isn't grounds for changing the system I cant imagine what is. Shouldn't all this be on the jokes board?

22. 九月 2004, 20:33:52
Grim Reaper 
<Maybe it would help if you see how an actual USCF elo score is computed. The pre-Glickman technique is very easy to walk through.

Let's look at Nellaf's 7-0 performance against his opposition:

facepan 1957
Scubbabisto 1624
whikki 1677
Thad 1758
juangrande 1805
CMoore 1150
penswift 1975

Under the old Elo system, you get 400 points more
than each player over your first 20 games or so if you win, subtract 400 from them if you lose, and add their rating if the game is a draw, then average the results.

For "obvious results" (like losing to a player > 400 points over your provisional rating, or winning against someone 400 below you) not being factored in to pull you down.

1957 + 400 +
1624 + 400 +
1677 + 400 +
1758 + 400 +
1805 + 400 = 8821 + 2000 = 10821

Over 5 games, this is a 2164.2 rating.

Now the 1150 result is not goverened by
the "+400" rule since is it an obvious result
that a 2164 should beat an 1150.

So, the player is given another game credit at the
provisional rating clip.

So, add 2164 to the running total:


10821 + 2164 + 1975 + 400 = 15360

Divide by the 7 games and you get 2194.

This is no where near 2417, as you can see.

To get a 2417 rating, you need to be undefeated against an average field of 2017. He never even played one person with such a rating.

22. 九月 2004, 20:33:51
coan.net 
*enter BBW again*

... speaking of ratings again, something that has always annoyed me is a rating is calculated from the rating at the time the game ends.

So lets say you are rated 1500 and start a games with someone rated 2000. But by the time the game is complete, his rating drops to 1000. Now you rating will not raise as much as if he was the 2000 rated person you started the game with.

In live games (where most ratings come from), this is not a problem since games are usually started and completed at the same time so rating never has a chance to change.

May I ask why ratings are figured this way and not using the rating the person had when they started the game?

22. 九月 2004, 20:13:31
Fencer 
I was thinking of that. Thanks for your confirmation.

22. 九月 2004, 20:13:29
bwildman 
*enter BBW*
LOL:)

22. 九月 2004, 20:12:40
coan.net 
I think the full 3 points should be given each time for a timeout.

That way if someone is about to lose, they will not let it time out with the possiblility to not lose as many points.

22. 九月 2004, 20:00:15
Fencer 
No, all finished games are figured the same.
Btw, it brings a possible problem with Pro Backgammon (to be implemented soon). Since a game of this type can be won with 3, 2 or 1 point [score points, not BKR], how many points should be given in case of a timeout?

22. 九月 2004, 19:21:18
bwildman 
题目: Fencer
will timed out games be figured differently?

22. 九月 2004, 19:18:56
Fencer 
Speaking of ratings (and finishing all discussions based on ridiculous notes about random numbers), I've just finished my BKR history generator and make the first test run (on my local computer, of course).
The results are very good, I think. Some differences between current values and recalculated ones can occur but nothing too drastic (+-50 points).
It will be launched when I am absolutely sure the generator is correct and when I finish a graph drawing engine.

22. 九月 2004, 19:14:54
DeaD man WalkiN 
题目: every1 keeps talking about ratings
at least BK tries to have 1. It's Fencer site and if he wishes. It could be like IYT site and not have a rating at all.
Yes from time to time I did not think some of my games was right but I just looked at it as, well at least we have something to look at.
BBW, I would like to thank you for what you put. cause with the 1700.2 and 1700.4 it now make a little more sence of why some times you win a game but you don't seem to move in rating.
Again TYVM Fencer for giving as something to look at and a way to kinda conpair who is close to my skill in said game.
:o{P

22. 九月 2004, 16:44:06
Grim Reaper 
When we all started, our Gothic BKR was 1300. So it was very very difficult to "climb up". We had 1900 and 2000 players clubbing each other to break 1500.

Mt 2400 rating was an 1100 point climb the hard way. That other person's was from winning against a "higher rated" player, then mediocre play compounded the rating.

All you have to do is beat a strong person early on, once, and you can pole vault over someone who has won 500 games and lost none.

From then on, just draw every game, and you lose 0 points.

Not a very realistic representation of a real rating system.

22. 九月 2004, 14:58:58
Maxxina 
YOu are right Stevie :)

22. 九月 2004, 14:50:28
Stevie 
It does depend on what bkr the opponents had too though doesnt it

22. 九月 2004, 14:47:04
Grim Reaper 
Conversely, I win over 100 games of Gothic Chess without a loss, and hardly scrape past 2400. Then someone else comes along and goes only 7-0 and was rated higher than me for a few months.

Not very realistic.

22. 九月 2004, 14:46:19
Maxxina 
ED : first 25 games will be your BKR optimal .

If you lost some games now , dont be suprised if your bkr will decrease with hundred points :)

22. 九月 2004, 14:33:28
Grim Reaper 
I played tournament chess for 11 years and never broke 2300. I played here for a week and have a 2600+ rating in chess.

Not very realistic.

22. 九月 2004, 08:53:20
ughaibu 
Established ratings are good enough for assessing a potential opponent's strength relative to one's own, give or take a couple of hundred points, and as one doesn't want to be confined within too small a band of suitable opposition I would say the ratings for the far greater part function usefully.

22. 九月 2004, 07:04:56
Grim Reaper 
No comment from me on the ratings. After all, I am mentioned by Mark Glickman, official USCF Ratings Chairman, as having successfully implemented his new, highly accurate system correctly in his first paper he published on it years ago at Boston Universtity.

I offered to help Fencer way back when, but got the stereotypical "nothing is wrong with the ratings" reply.

I know exactly what Fencer is doing wrong, since he is constantly referring to a very old paper Mark wrote, which DOES NOT consider calculating the so-called "rating period" variable properly the way we need (rate after a 1-game trial, not many provisional games in parallel).

Hope everyone else enjoys the random 4-digit number next to their name :)

22. 九月 2004, 06:56:30
plaintiger 
题目: Re:
ohhh. well if you put it that way...yeah, i definitely would not be happy with people incessantly shaking me awake and demanding explanations of rating systems! :D

22. 九月 2004, 06:51:15
Stevie 
Its 7am I guess? cos its 6am here LOL wohoohohoooo only 2hrs till home time LOL I am still being polite LOL

22. 九月 2004, 06:48:47
Fencer 
Right. But do you know what time is here? I am still sleeping :-)

22. 九月 2004, 06:46:40
plaintiger 
题目: Re:
plaintiger修改(22. 九月 2004, 06:47:21)
c'mon, Fencer, be nice now. everybody else is.

not everyone has encountered your previous explanations, nor will everyone go looking for them on blind faith that they're around here somewhere (you know how many boards there are here, and how many messages are on each one!). as people join BK and see weird BKRs that don't seem to make any sense, they're going to continue to ask this question. and unless and until a system of BKRs is implemented that *does* appear to make sense at first glance, the questions will keep coming. just ignore them if you like; it's evident that you don't *need* to keep explaining because other people who grasp the system, like BBW, will do it for you.

your time is much better spent on updating and upgrading and tweaking and all the other vastly more important stuff that only you can do.

right? :)

22. 九月 2004, 06:43:29
Fencer 
BBW: Yes, it would be useful for me as well :-) I currenly sort bugs by priority, so it shows the open ones at the top of the list.

22. 九月 2004, 06:41:43
coan.net 
题目: Re: Heres another for you
Yea, the bug list is getting long - hard to remember what is there or not.

Fencer - I know it's not a top priority, but would be nice to sort bug list to show still open bugs only.

22. 九月 2004, 06:37:39
plaintiger 
thank you very much, BBW - that's just the kind of explanation i was hoping for!

yea verily dost thou rule. :)

tiger

22. 九月 2004, 06:37:02
Stevie 
题目: Re: Heres another for you
Thanks BBW, I must have missed that one

22. 九月 2004, 06:35:40
coan.net 
Stevie - I think that is a bug if there is a draw (Bug #7)

If that was not a draw, then there is another problem. :-)

22. 九月 2004, 06:34:57
Fencer 
How many more questions about BKR? Whether the formula seems fair or unfair, accept it. I explained it many times and I don't want to repeat it again.
Vacation days should be fixed.

22. 九月 2004, 06:34:09
coan.net 
Doing the math, especially when users have few games you can see a lot of weired things..... and when i say few games, that is under 100.

1) Doing the math, even if you win you could actually lose rating points. (But I believe in the coding of BK, you will never lose points for a win - even if the math says so) - hence, sometimes your rating will just not move.

2) Ratings are rounded. So if says you have 1700, you may actually have 1700.2 --- and then if you win, your rating might raist to 1700.4, but will still show as 1700.

3) The fewer games you have finsihed, the more erradic your rating will be. BK tries to solve this a little to not even show you ratings until 4 games are done, and not show established until 25 games are complete. (Heck, I actually think I read somewhere where 400 games should be the minimum amount of games that need to be complete before getting a good established rating.)

4) Good news is that Fencer is working on a way to recalculate all the ratings from day 1 on all the games soon.... so if people really think things are off a little (And my froglet rating is about 30 points to low because of a bad game timeout), it will be fixed!

22. 九月 2004, 06:27:21
Stevie 
题目: Heres another for you
Stevie: old BKR = 1748, new BKR = 1748 (0)
fullmoonhawke: old BKR = 1799, new BKR = 1799 (0)

I am the one who was and is 1799 and fullmoonhawke is/was 1748

It swapped the bkr, but luckily left them correct on our profiles LOL

22. 九月 2004, 06:21:04
plaintiger 
plaintiger修改(22. 九月 2004, 06:23:52)
i recently had some funky business with BKRs too: here are the results of the last four games of Froglet i completed, in order:

plaintiger: old BKR = 1700, new BKR = 1700 (0)
MsDelete: old BKR = 1301, new BKR = 1240 (-61)

plaintiger: old BKR = 1700, new BKR = 1700 (0)
Paula29: old BKR = 873, new BKR = 871 (-2)

plaintiger: old BKR = 1700, new BKR = 1700 (0)
TJ: old BKR = unrated, new BKR = unrated

plaintiger: old BKR = 1700, new BKR = 1707 (+7)
Lorraine: old BKR = unrated, new BKR = unrated

can anyone explain the logic behind this (without getting too much into mathematics, which will serve no purpose but to bewilder me instantly)?

why did i get no points for three wins in a row - for one of which my opponent lost SIXTY-ONE points, and one of which was against an unrated player - and then get seven points for the fourth game, also against an unrated player?

this makes no sense to me at all.

oh yeah: the ratings of the first two players were established (not provisional).

thanks...

le tigre

22. 九月 2004, 05:20:43
skipinnz 
题目: Re: It doesnt seem fair
Stevie, I think 1500 is the axis so 1600 is way above compared to 1200 which is way below.

22. 九月 2004, 05:12:40
Stevie 
题目: It doesnt seem fair
I was aprox 1650bkr, playing someone aprox 1250.
I win, and get +7, they get 0
we draw other game, I lose 32, they get +42
I could understand changes of that size on a draw when high and low bkr, but those bkr are both mid range id have said :o(

22. 九月 2004, 05:02:57
Eriisa 
rats. I liked have 30 days again.

21. 九月 2004, 17:29:58
Fencer 
Yes, there was a bug. It will be fixed with the next update.

21. 九月 2004, 16:38:35
Fencer 
I'll check it.

21. 九月 2004, 13:52:17
rabbitoid 
题目: vacation question
I thought the vacation counter is reset on January 1st?
I've just looked at mine, and it says 30 left. I know I've taken days in July, I think I've checked 15 days. mind you, I'm not complaing...

21. 九月 2004, 06:35:37
Fencer 
I want to place it to some "smarter" location when I find out how it can be done in Javascript and DHTML :-)

21. 九月 2004, 05:16:35
coan.net 
题目: re: Hot Spot
coan.net修改(21. 九月 2004, 05:18:29)
Yes, any page - right above & to the left of "Main page" in the left hand menu.

I'm not sure how it looks when the page is not in the columns layout - I think I'll go look!

EDIT - Just looked, same place. :-)

21. 九月 2004, 05:13:09
MidnightMcMedic 
题目: Re: Hot Spot
It works from any page, not just the 'main page' What I'm saying is that you can do it from here, from inside a game, or anywhere :)

21. 九月 2004, 04:14:32
musiclover 
题目: Re: Hot Spot
BBW, thanks, I followed your instuctions for eagle eye and it works for me. Live and learn a new thing everyday; that's what they say.

20. 九月 2004, 18:30:40
coan.net 
题目: Hot Spot
I think it is for knights and above only. Plus it sounds like it does not work with WebTV either.

To see it, first you have to change your settings so Javascript will work.

Then in the upper left corner of the screen (above & left of "Main Page"), if you stick your mouse there, a drop down menu will list the first 30 games which it is your turn to move in the order that in on the main page.

20. 九月 2004, 17:58:50
eagle eye 
题目: hot spot?
I'm lost :-) what is a Hot Spot? how can I see it? and what it for?

20. 九月 2004, 05:11:55
Vikings 
first of all you have to be at least a knight

20. 九月 2004, 05:10:29
chupacabraVS2 
题目: I looked in help,FAQ,message boards,server news but...
..cant seem to find out where you upload a photo to your profile.can anyone advise please?
8^)steve

19. 九月 2004, 17:06:55
LongJohn 
LongJohn修改(19. 九月 2004, 17:09:35)
seems Skyking deleted his question again and made everyone looks silly LOL

19. 九月 2004, 17:06:04
Fencer 
Of course I am here. And it's not a bug. It can happen when you are close to the limit and new tournament games are generated for you.

<< <   520 521 522 523 524 525 526 527 528 529   > >>
日期和时间
在线的朋友
最喜欢的讨论板
朋友群
每日提示
Copyright © 2002 - 2024 Filip Rachunek, 版权所有
回顶端