CTP Ratings (29-APR-2001)

[slamp]

Change in Dutcheese game:
- Edgar, leader of Germans left the game and was replace by Blackice.

Blackice, do you want to be rated ?

[St Jon]

Love Conquers? Turn 109.

1. Mobius
2. Rick Elkins
3. St Jon
4. King Thor
5. (Keygen)
6. Stavros

[Keygen]

Lungmatch 2.2 Turn 29

1. Swissy
2. Keygen
3. Solver

[quinns]

Thanks Birdman for trying to make the formula better. What formula are you using to calculate the ratings? The numbers that I show for ratings are the BEFORE ADJUSTMENT ratings of the players. The Delta is the change. And you're right, I don't show the new rating after adjustment, I know that is confusing, but I thought it would clutter up an already complex looking report.

Edit: I see what you mean now Birdman. Yes we could do that, and maybe that would be fairer. I would have to BACK RATE all of the ratings before commencing with this and write a new program, (mucho trabajo -- much work ). If we have a strong concensus to do this, I could muster up the effort, I think.

The formulas I use have nothing to do with tennis. They are just straight forward formulas which cause a fair "bell shaped" rating adjustment dependant upon the difference in players' ratings. Here they are again. I hope this alleviates your skepticism of the fairness of these formulas.

Ratings Formulae

Formula 3.
Probability of Lower Rated Player winning:

PROBABILITY = 1 /(2 + (HighRated - LowRated)²)

Verbose --> the probability of the lower rated player winning is equal to the value of 1 divided by the sum of the value of 2 plus the square of the difference between the higher player's rating minus the lower player's rating.

Example: If a player rated 20.000 plays a player rated 18.000 then:

PROBABILITY = 1 / (2 + (20 -18)²) = 1/6

That is, the Probability = 1/6 or 0.166667 that means that the player rated 18 has a 16.667 percent chance of winning.


Formula 4.
If the Higher Rated Player wins then:

CHANGE = Probability of Lower Rated Player winning divided by two (2)


Formula 5.
If the Lower Rated Player wins then:

CHANGE = (One (1) minus Probability of Lower Rated Player winning) divided by two (2)

------
Examples of using Formulas 4 and 5:

Example: From the above example, the Probability of the Lower Rated Player winning is 0.166667.

If the player rated 20 wins then:

CHANGE = 0.166667 / 2 = 0.083 (rounded to 3rd decimal place)

So that: New 20-rated player rating = 20 + 0.083 = 20.083

and: New 18-rated player rating = 18 - 0.083 = 17.917

If the player rated 18 wins then:

CHANGE = ( 1 - 0.166667) / 2 = 0.417

So that: New 20-rated player rating = 20 - 0.417 = 19.583

and: New 18-rated player rating = 18 + 0.417 = 18.417


[This message has been edited by quinns (edited May 03, 2001).]

[blackice]

"The formulas I use have nothing to do with tennis."


http://www21.addr.com/~quinns/wtr_info_formulas.html

Oh sorry since they looked exactly like these ones for the Tennis I just assumed sorry.

And personally again I see more than just the way it is scored as the problem. How will JWPP for instance play Klair or some other top ranked player. Will I ever play a newbie again no obviously if I play for points no it would be rather dumb would it not? So I have chosen to play for fun that way it really only matters that I have fun. Which I am
I can then continue to play newbies meet new people and not get miffed or discouraged.

Again I see no proof to contradict that if the win, win, loss as per the example of the last rating will change. If these games stay as they are I will be in last place before long. For winning two and losing one lol
Or the Keygen game example I gave is extremely possible again I see no proof to contradict it as yet.

BTW QUINNS, SOLVER, KEYGEN and the rest don't get me wrong GREAT JOB! I mean that sincerely

This PBEM has provided great fun and I have cyber met some great peeps

Ok Barbaque and kids time too nice a day to be inside

[quinns]

quote: Originally posted by Solver on 05-03-2001 12:24 PM I like the system, and think it's fair. But I repeat, nothing can give an absolutely precise skill representation.

Agreed.

[TheBirdMan]

Then, who says that we have to use (we do) the exact fomular as in tennis?

I mean, in tennis all plays one to one. Here we play one to ?? (at least two and up to severn). We might only have to ajust one of the "permanent" values in the formulas. It is not the same playing one to one as playing one to ??.

And I'm happy to see, that Quinns have "accepted" the substaional of my mail. But to change how we are doing, that is a something different and very difficult to handle.

[quinns]

Alright Birdman. I'm glad you're happy! And I'm glad you realize that it would be alot of work to change, not to mention the potential of messing the whole thing up.

The formula isn't a "tennis" formula, it is my formula that I created to use to handle "any" type of competition rating, in "any" game. The formula is generic for use with all games. I made it up all by me ownself, (admittedly, based roughly upon the International Chess Rating System). I didn't take it from the "tennis ratings manual".

[blackice]

Good job Quinns

Nice work

[blackice]

I appreciate that Quinns thanks

Food for though final word here

GL

1500 wagers 300 20% rank 22
1000 wagers 200 20% rank 54

500 points on the table, 1500 lost to 1000

1000 gets 335-200=135+=1135 new rank 39
1500 gets 165-300=135-=1365 new rank 28

Make it up in one game, no biggy

Reversed, wager same

1500 wins 335-300=35+
1000 lost 165-200=35-

100 points on the line verses 35 nope, yes make it up in one game. Really though the win points do very little change to rank so why bother? choice.

1500 vs 1500 won 300-402=102+
1500 vs 1500 lost 300-198=102-

102 vs 102 ya why not?

Again make it up in one game, possible shift in rank.

You would not play for points, many 1000 point players The rating system is focused on playing people about your same level. Makes for more even win/loss ratio. Plus you would have more opportunity to play people your own or higher ranking online.

Now as we do it more players;


#1 1500 wagers 20% 300 rank 22
#2 1000 wagers 200 54
#3 2000 wagers 400 11
#4 868 wagers 173.6 101
#5 976 wagers 195.2 56

Total wagers 1268.8

#5 wins 520.3-195.2=325.1+ 1301 new rank 27
#2 second 342.6-200=142.6+ 1143 39
#4 third 215.7-173.6=36.7+ 905 89
#3 forth 126.8-400=273.1- 1727 11
#1 last 63.4-300=236.6- 1263 31

Reverse the bottom two, the two higest players for bottom spot you will see that the 1500 point guy stays about the same rank too. The only one that really loses is "The loser" This is exactly why you see top GL players playing three peeps games or four with newbies included.

GL points system 5 player game similar ranks as PBEM, everyone goes up or stays about the same except the loser. Justifiable I would say but it does happen ask Flames. heh or me



[This message has been edited by blackice (edited May 03, 2001).]

[quinns]

Alright Blackice. I will not continue to try to convince you, though I value your opinion on this system. I will apply your penalty for switching from rated to unrated, (via the Elimination Rule above), and remove you from the list, per your request.

I still believe that our CTP PBEM Rating System is much fairer than the GameLeague system regarding drops in ratings. For example, a player rated 20.000 in our system could NEVER lose anymore than 1/2 a point to someone rated even as low as 1.000. That is a maximum drop of only 2.5% in ratings for a losing in a single game to a single player. Not true in GameLeague where a 1500 rated player could potentially lose up to 300 points, 20% of their total by playing, and losing to, a lower rated player.



Edit: And Ice, when I said they have nothing to do with tennis, I meant that the formulas are not based upon tennis. Yes, I used the same formulas, (that I created originally -- i.e. I personally own and developed World Tennis Ratings), for the tennis ratings site as well, but they have nothing to do with tennis, per se, and are strictly based upon mathematical probability, only.

[This message has been edited by quinns (edited May 03, 2001).]

[Solver]

I like the system, and think it's fair. But I repeat, nothing can give an absolutely precise skill representation.

------------------
Solver the "Running Beer" - http://www.aok.20m.com

[Solver]

quote: Originally posted by TheBirdMan on 05-03-2001 04:22 AM It really hurts when beaten by a lower ranked player.

I know. That's how I lost my first positon in SMAC .

------------------
Solver the "Running Beer" - http://www.aok.20m.com

[quinns]

Alright Ice. Sounds like you are happy with GameLeague's system. No problem. To each their own! Wagering points causes top ranked players to shun lower ranked players. You have admitted that. I am far from convinced of the merit of their system -- for whatever that matters!

Quinns

[Solver]

Good said Quinns. I like the inability of player number 1 gaining many points by beating player number 30 and vice versa. But, number 30 beating number 1 will go up, and that's fair. Good system .

------------------
Solver the "Running Beer" - http://www.aok.20m.com