more futile than that? fat chance.

why should a draw be worth more than a survival?

because somebody else wins = you lose

Since there are stalemate lines in dip, it is very destructive, if people start helping the big guy just to get 2 more centres instead of forming a coalition to secure the stalemate.
The problem is not people going for a draw, the problem is people playing the dog for the big guy only not to be attacked.

Oxim: both are considered problems by some people; both are considered legitimate tactics by the majority of players; both are considered good, even priority strategies by the rare few.

if you care about either, get a group of people who share your views and play only with them.

Already in there Kestas!

Well, Zarathustra, appreciate the effort, but the current point system is far superior than the one you suggested. Not just in tweaking in discussion, but in pure theory.

Theoretically, I think TheGhostmaker's system based on the Chess Elo system is probably the most sound. I understand that Kestas is skeptical, and I was at first as well, but it's really fairly elegant. It's a very simple concept that is quite applicable.

>No-one can agree on what parts of the current system need >to be changed, so I really doubt we're going to see any >agreement on an entirely new system.

>The fact that every new points system thread has only ever >ended up with unresolved debates and disputes makes >them really tiresome and repetitive. All these threads are the >same, the same issues, the same arguments, the same >end result

Dare I say it, the most likely reason for them being futile is that the man hosting the site is not interested in a change. If there was more than the faintest of glimmers of hope, then we would probably have a better, longer debate and reach a majoritory view. However we have six people in favour of Elo-rating in some form, of those that posted in this thread: myself, aoe3, Darwyn, Churchill, belsherj and Chrispminis. Of those four who haven't supported it, only yourself and Zarathustra have proposed something different or rejected it, the other two haven't commented.

>The only compelling reason I can think of to change the >points system is that if it gets intricate and convoluted >enough no-one will be able to understand it, so no-one will >complain (if we implement ELO I doubt anyone will be >complaining about the unfairness of logistic curve derivatives >or whatever)

The reason that people are complaining is not because it is simple, but because we see discrepancies between our evaluation of player skill and the point system's evaluation, and because there is a very serious flaw- playing a weaker player pays out as well as playing a stronger player, even though it takes less to win.

actually, i think Kestas would implement it right away if someone programmed it.

> If there was more than the faintest of glimmers of hope, then we
> would probably have a better, longer debate and reach a
> majoritory view.
We constantly have long debates, we're in one right now :-| but they've never produced any majority view, just more debate (and more alternative systems)

As aoe3rules says though I wouldn't be coding it myself anyway, but I wouldn't merge it in either without seeing a very clear majority in favor of it, and seeing that it solves all the issues the current system does
I won't go through a whole load of hassle to replace one set of tradeoffs with another, one set of complaints with another set.

It may seem obvious to you that your system is correct, as much as it's obvious to Zarathustra that his is, but look at it from my POV with all these alternate systems and no unity on which system is best (or even on what's wrong with the current one). 58 people currently logged on and the best agreement on an alternative has 6 people behind it? It shouldn't be hard to see why I'm not interested in change

see new thread.

i didnt think mine was the best, just better because it attempted to account skill into ranking and deal with the problem of playing lots of games making you look like you just had a big defeat or were less experienced.

i really like the ELO system now that i know what it is. it is the type of equation I was looking to get. Im in support of it.

Yeah ELO looks good to me, with one observation:

In a PPSC game. a player with 1400 points plays 6 players with 100 points. What is his expectation? your formula gives 1400/2000 = 70%.

70% of 34 = 23.8, but the game ends when you get to 18 SC.
So I guess experienced players won't often pay a field of newbies at PPSC because they could lose points even if they win.

Zara-

Unless I'm reading this wrong, RankPoints becomes inflated over time for each particular player; 4 survivals outrank 1 win. Perhaps try instead:

Point Pool * Rank Factor = Rank Value
where Rank Factor = (Rank Points / Games Finished)

By my thinking, you'd track Rank Factor over a player's whole career, where it kinda-sorta approximates a win percentage. I suppose to avoid any misleading results you'd also want to not formally rank anyone with under 10 or 20 finished games.

(If I am misunderstanding and Rank Points does not persist across more than one game, I don't see how one's overall Point Pool belongs in the formula as it is of a different "scope".)

In this case I think also someone's suggestion that the weightings are off has some merit, but that's just a tweak. I'd suggest:
Eliminated = 0
Survived = 1
Draw = 3 (or 4)
Win = 7

NevilleChamberlain wrote: "
Yeah ELO looks good to me, with one observation:

In a PPSC game. a player with 1400 points plays 6 players with 100 points. What is his expectation? your formula gives 1400/2000 = 70%.

70% of 34 = 23.8, but the game ends when you get to 18 SC.
So I guess experienced players won't often pay a field of newbies at PPSC because they could lose points even if they win."

100 is the average rating. Rait is at about 200-odd The case of a player ranked at 1400 is really, very unlikely.

Is see your point (1400 is unlikely), but:
1) you have underestimated Rait's rank.
In actual fact, raits rank would approach 860, because his win ratio of completed game is 59% (assuming WTA, if playing 6 average players, to get expecation of 59%, 860/1260 = 59%).
2) Doing the same calculation for Rait with PPSC yields a different result:
(assumng all wins at 18.5 SCs, all draws and survivals at 11 SCs)
Expectation: 59%*18.5SCs + 27%*11SCs =13.8SCs = 41% of SCs, leading to a rank of 415. (415/1015)
There's a big difference between 860 and 415, so you'll need some wizardry to line them up, unless you you have a separate ELO ranking for each type of game?

Also, what value of V are you proposing?

Finally, I'd like to say that I support ELO being implemented alongside the current system, so people can see it in action and get used to it. Swapping out the current system straight out for a new system would annoy too many existing players.

Once people see it in action, you may be able to get more support for a change.

I am quite happy to write the PHP/MySQL code for this.

They are treated differently, and faithfully to the way the point system treats them.

With V=6 Rait had a rank of over 190, but less than 200. Remember that he is such an outlier, and 6 is rather a small value for V, that he was actually still building his rank to a reliable amount.

If you can put the work in for writing the code, I shall do everything possible to make sure that the system is as good I can make it. The only reason I stopped where I did was because there was no point in continuing if nobody would write the code.

Thanks

No worries, I can write the code. Am going sailing for a few days, will look into it when I get back. Am thinking about implementting a javascript drag-drop interface for entering orders anyway, so I ought to get phpDip up and running on my Mac.

Ok, I am going to run some tests with new V values then. (=computer down whilst my woefully inefficent software runs :D).

Do you have real data to run it on then (ie the SQL game results data table from the main phpDip site)? Or are you doing like me and putting rough approximations in Excel to get a quick answer?

I have the data from July 30th. I am trying different values of V, to see where things are going as I change it. V=6 seems to be a bit too low. I am now trying V is 14, having done V as 10. From that we see that some of the younger, better players have climbed, which suggests that V=6 was too low. We shall see what happens this time.

What's wrong with V=50?
A player losing their first game would only drop to 93 - a winner would go to 143. (WTA)

Good point. I'm going out for a walk now (flashman, since you seem to have been stalking me, now's your chance :P). I'll consider that. V=18 is about to run.

@modesty_johnson:

Yea, I was thinking that the rank factor would be divided by total games but i forgot to include it. Also, i agree that the weights are off, but i knew that they were very rough when i wrote it.

I like your insight on the newbie dilemma. i had thought about that problem when i wrote the formula up, but figured that the fact that Newbies have low point pools would take care of the problem. I figure that most players will get better over time and thus increase their point pool. this lets them play higher stake games and gain rank.

i think the greatest benefit of the system i proposed was the simplicity. Also, according to my limited knowledge of programming, it would be easy to implement. If ELO can be done and someone is willing to do it, great. if not, i think with the modifications we have suggested here to my formula, we could get a pretty good ranking system without a lot of hassle

The hassle is more with the fact that we need to get Kestas to agree with the system rather than getting someone to program it. No matter how complicated or simple the formula, it is still a case of type in and go. We need somebody there who knows how to type it in, and an agreement on what to type in.

As for V=50, the problem is the innaccuracy- if somebody wins a game, their rating goes up by a huge amount, which is not necessarily correct I reckon to V=25 at the moment.

I think that in actual fact it is more inaccurate to have low value for V.

Regardless of the value chosen for V, there is an ELO rank score that accurately reflects how good a player is.

No value of V will make a player's score instantly be that value, rather everyone's score will gradually approach that value, but never quite get there. Having larger values of V means that the score approaches the true value quicker.

When a player doesn't have many games under his belt, a large value of V will make his score jump a lot when he wins a game. This is not a problem with V, it is a result of the fact that he has jumped from 0/3 to 1/3. His ranking score should jump up when this happens.

Once an average player has won 7/52, going to 8/53 will change his ranking score less.