[Attempt at] New PBEM rating system - continued

[Solver]

See, Pete, I believe that it should NOT anyway continue till a fully developed program is available - one that stores ratings in a database, etc. I think that at one time I even wanted to write it, but for some reason did not, oh whatever.

And, I disagree that score should be taken in account, I pretty much believe that the PG provides a fair indiciation of how is a player doing in a game - although, well, score does have its upsides. I'm not so cathegoric here.

Nonetheless, I still believe that the Tennis system, when used, was better. Another indication of it is, that it worked for longer than the 10% system did. Tennis system worked ever since the ratings were originally invented, and all the way till a bit after Quinns took his long break from Apolyton.

Again, I repeat - should there be a program that does it, I would be very glad to again reinstate the Tennis systems.

And, in those times, the ratings got published in CtP news all the time, etc, the tennis system got more recognition from players.

[blackice]

Solver it was not the formulas most disliked, it was the fact one player playing one game and winning could sew up top spot endlessly. The "Clair factor" which you and I both know most all hated that aspect.

Color it how you will that was the bigest flaw in the system and it could not be corrected. You want multi game participation that system was not the way to get it. Keeping people in games nope it did not do that either... What was good about either of those things?


That while the rest of us played 20 games and bounced up and down the rankings with no end in sight. No consideration for the unique aspects of CTP, land, diplomacy, etc. People quiting games because of land position it was endless...

A top player could lose 20 spots simply because of land. So we made up more rules upon rules just to get the system working. That in itself marks a flawed system, we hit what a full page of rules and more were to added just to get the system to work.

The new system needed minor rules not endless streams of them.

BTW not to get into this again but the tennis ranking system is not very well liked in most forums, let alone the tennis circle.

Actually not so much player count as it could use 2 to infinity as players. It was more dependant on score and pg as I understand it from the originator at GL. The percentages paid out were averages of the percentage if you conbine the score and pg totals vs the other players. IE skill... I will point out it was designed for CTP not chess, tennis, or any other non-related game.

They also took into concideration landscape as the game progressed.

Designed for CTP, simple, more fair, less rules and more game participation were the keys reason's for the change. Up until partisan politics, personalities got way out of hand most liked the new system.

I threw up my hands in dismay at the childishness of the whole affair. It became not worth the effort, rather than seeing it as a change for the better it became a call to arms for the few partisans to bring the whole thing down. enough said...

[Mathemagician]

alright,

first of all, i think Excel can do the calculations.
all i would need to enter is the position of each player in the latest power graph of the corresponding game.

assume, ratings are re-calculated every 10 turns, all i would need is someone telling me who is ranked where on the power graph (counting only the human players).

i enter that into excel and voila: the new results.

i admit, i havent done it yet, but i thought about it, and i dont see any relevant lacks in the excel logic.

only a sorting algorithm for the final table i havent found yet, but im sure that could easily be done manually if the ranking is updated, say twice per month.

[Solver]

I agree that one player being well ahead in a game was the biggest flaw of tennis rating, but that is, IMO, rather easily fixed. I don't want to go into detailed maths here now, but by adding another variable to the formula, which indicates the turn number, this could be balanced. That is, the player atop of power graph gets less and less rating boost, while the loser also is taking less of a loss in the rating. When I have the time, I can do the maths and see what the coefficients would have to be for it to be reasonably balanced.

See, there will always be people disliking a certain ratings system, but there's a good reason the tennis ones are used so widely, and it's that they work fairly well - with flaws, but so does everything else.

Therefore, I would be rather willing to contribute my thoughts to adjust the tennis rating formulas, but they were, IMO, preferrable.

It's getting the players back here again that would be problematic.

[Solver]

Originally posted by Mathemagician
alright,

first of all, i think Excel can do the calculations.
all i would need to enter is the position of each player in the latest power graph of the corresponding game.

assume, ratings are re-calculated every 10 turns, all i would need is someone telling me who is ranked where on the power graph (counting only the human players).

i enter that into excel and voila: the new results.

i admit, i havent done it yet, but i thought about it, and i dont see any relevant lacks in the excel logic.

only a sorting algorithm for the final table i havent found yet, but im sure that could easily be done manually if the ranking is updated, say twice per month.

In this regard, you better trust me. Excel does not work good enough, and I say that as someone who has been doing that stuff with Excel. Firstly, entering and doing all that stuff still takes certain time, and it's easier to make a mistake. And, it takes enough time for you not to look forward to doing it all. Excel won't keep you ratings of players unless you work on it a lot, which is a pain in the butt.

A program that worked, though, would easily do it all so quick that the ratings could even be updated every week, provided that there are enough rated games - and the last time they were.

I speak of experience here.

[Mathemagician]

now to the system itself.
i was hoping to not have to explain it in detail.

here is an overview:

some of you might know the ELO system i referred to above (that as i know is used at least for the Magic world ranking, there including 10,000s of players).
no idea why it is called ELO. anyway...

the formula there, for each finished two player game, is:

new score = old score + K-factor * {result - 1 / (1 + 10^((own old score - opponents old score)/400)) }

i hope all the brackets are right and understandable.

K-factor
simply weighs the importance of the tournament the match was played for. it therefore is the maximum number of points you theoretically could gain or lose in the game.
for equally scored players, the winner will take K/2 points and the loser will lose K/2 points.

result
is the result of the match being played.
1 if you win
1/2 for a draw
0 if you lose

the complicated thing in "(result - complicated thing)"
thus gives a value from -1 to +1 depending on the score (think:skill) of both players.
this is NOT the in-game score, its the ranking score of those players in the global ranking system.
the whole bracket then gives a value from -1 to +1, meaning you can lose 100% of K (for a -1) or win 100% of K (for a +1) or something in between.

the main problem was to replaced the difference of two scores in a two player game by something equivalent in a multi-player game without losing the important characteristics of the system.
one of these characteristics was that the complicated part of the bracket, calculated for player A would always be the negative of that for player B, so each player would gain what the other loses.
thus the whole ranking system would not lose or gain any points.

i now simplified this last bracket part in a somewhat more linear way,
mainly by using r(i) and s(i), so the bracket part for each player would be ( r(i) - s(i) ).
actually, the r(i) should be r (rank(i)) where i is the index of a specific player.

the r(i) would sum up to 1, like in the original.

im still working on the s(i), but what is for sure is that they depend on the skills (score) of ALL participating players in the game and of course in a seperate way, on the score of the player with the index i.

..

so much... want more ?

[Mathemagician]

Originally posted by Solver
In this regard, you better trust me. Excel does not work good enough, and I say that as someone who has been doing that stuff with Excel.

i see that as a challenge to prove you wrong.

[Solver]

I see. I'm good at maths, so you can hardly scare me with those big long formulas. And I can say that to work effectively in games with many players, the formula would need to undergo some changes.

one of these characteristics was that the complicated part of the bracket, calculated for player A would always be the negative of that for player B, so each player would gain what the other loses.

I'll take your word for it... don't want to check it now, and that's another thing that I dislike. IMHO, that's not good for a game such as CtP.

And finally, the whole K-factor is wrong here. I know how it's right in Chess and such games, but we can't do it here... unless we actually make K some constant value here.

So, for now, you don't have me convinced, I believe that the system would only work reasonably for 1v1.

[Solver]

i see that as a challenge to prove you wrong.

Try to. Give me a fully automated program and let's see if you can do it quicker with Excel than I can with a program written solely for the purpose of calculating these ratings .

Just hardcode the whatever formulas, tennis, 10% or ELO into the program, and there it is .

[blackice]

rather easily fixed.

It was with the new system, you see the new system had that built into it.....

No need to fix it the new system did all that and more...

No time to view it now Math, have to go to work thanks I will have time later though.

Drawing people back Solver will take a group effort if partisan politics and iron fisting were to contniue, well I agree it will be a problem.

You see some think for the community as a community not power or politics...

In the community poll most prefered the new system, remember?

[Mathemagician]

thats why im still working on the formula.

i didnt say i want only the bracket changed, i said it was the hardest part.

K is somewhat irrelevant. i dont know what you think is wrong about it, but it simply scales the (-1,+1) interval to the desired length in which the score scale is.

so far, i didnt completely check this system, but i can say that at least the score the weakest player will win if he wins the game is equal to the score the strongest player will lose if he loses the game.
me thinks that also all the scores some up to 0, but i havent mathematically proven it yet.

and what exactly was the thing you disliked about that last quote, solver ?
that the whole score wouldnt change ? why not, it gives the whole ranking the (in my eyes) ideal characteristic that you CAN win more points by playing more games, but also increase your risk of losing more points. plus, the higher you are on the ranking list, the higher the average place you need to make in your games, or simply put, the higher the percentage of games you need to 'win' in order to keep your ranking position.

[Solver]

Drawing people back is possible, though, with an effort from the sane and old-time members of the community.

As said, I have nothing against the 10% system, I just believe it not being better than the old one. But, of course, I am rather democratic. Many people do, though, I believe, not clearly enough see the good and bad points of each.

I do, for instance, trust that the rating system with tennis formulae provided better adjustements to ratings overall, the tables being more dynamic and interesting. But this is just my opinion.

And one thing Pete, I'm sorry to remind you, but you were also involved in some of that iron fisting, and what not .

Well, I admit that I *may be* biased towards the tennis system somewhat, too, for reasons that people who've been around as long as you can probably understand.

[Mathemagician]

about the politics/poll, i think first we should sort out which characteristics the miscellaneous systems would have and then everyone, also the mathematically, less inclined, can vote on the system they want, judging by desired characteristics.

some can check the mathematical statements, others might just trust them.

[Mathemagician]

solver, could you please tell me more about this tennis system ?
was it the one gavrushka used that needed a hell of a lot of tables ?

did i mention my system, apart from more or less complicated equations , needs only the results, previous scores and number of players as input and directly puts it into the formula ?

[Solver]

Mathe, the thing I was disliked was exactly that, the losing players loses the amount of points the winner gets with a minus sign.

That is, |lost| = |won| , isn't correct, IMHO.

A player rated 1700 beats a 1610 player and both get the same adjustment to their ratings. It works fine, but in 1v1 games. In a 7 player CtP game, one player has 6 wins essentially, the next one 5 wins and 1 loss, and so on. Won't this formula give the winning player too big of an advantage over the others?