Average of Opponent's Score/Rank on Profile

[BaldAdonis]

If you don't work out all the scores at game end how will you be able to tell the other players' scores from games that you lost in? (Aside from the winner).

The winner takes a certain number from everyone. So say they get 20 from you, and 20 from someone else, then that someone else can be said to have the same score as you. So it all in ratios like that. It won't be perfectly accurate, because scores are rounded to the nearest point, but you'll get a good idea.

That's the only part that Map Rank doesn't do right now, but the information is in the logs.

This suffers from the same problem.
It is missing all of the ranks of players who also lost in the same game as the targetted player's losses.
If target wins, no problem - that number of points gained has taken all the players ranks into consideration.

If target loses then that figure only is a comparison with the winner's score at game end and no other player is considered.

But if you have target/cadet and cadet/cook, then you can determine target/cook.

[saaimen]

So you'd make it work with wins as I mentioned, but what would you do exactly for losses?
Divide the lost points by the average of lost points of all losers, and multiply that with the lost points again? (I know it sounds chaotic )

[BaldAdonis]

So you'd make it work with wins as I mentioned, but what would you do exactly for losses?
Divide the lost points by the average of lost points of all losers, and multiply that with the lost points again? (I know it sounds chaotic )

I think you missed the point. Say I'm player A, and I play a 4 player game with B, C and D. What I want to know are the ratios A/B, A/C and A/D, because those will tell people if I'm playing higher or lower ranks, and how much higher or lower they are. If I win, as you said, it's easy, because the points I get from B (rounded to the nearest point) are 20*A/B. If B wins, all the log gives me is 20*B/A, 20*B/C and 20*B/D. But thats all you need, because (20*B/C)/(20*B/A) = A/C. So, just take the score that the winner got from each person, and divide by the score they got from you, and you'll get your score relative to the third (or fourth and so on) player.

[chipv]

If you don't work out all the scores at game end how will you be able to tell the other players' scores from games that you lost in? (Aside from the winner).

The winner takes a certain number from everyone. So say they get 20 from you, and 20 from someone else, then that someone else can be said to have the same score as you. So it all in ratios like that. It won't be perfectly accurate, because scores are rounded to the nearest point, but you'll get a good idea.

That's the only part that Map Rank doesn't do right now, but the information is in the logs.

This suffers from the same problem.
It is missing all of the ranks of players who also lost in the same game as the targetted player's losses.
If target wins, no problem - that number of points gained has taken all the players ranks into consideration.

If target loses then that figure only is a comparison with the winner's score at game end and no other player is considered.

But if you have target/cadet and cadet/cook, then you can determine target/cook.

Sure but these are ratios. Double everyone's scores and you get the same ratios.

e.g. Target = 100, cook = 400, cadet = 800 would give the same ratios as 200,800,1600 which are different ranks. Ok it's all relative.

[saaimen]

I think you missed the point.

Maybe just a bit, thanks for your explanation. But my question was what you do with those scores afterwards. You can't get them into the win/loss-substraction I proposed, can you? Cause they aren't win/loss points, they're "colleague loser" points

Sure but these are ratios. Double everyone's scores and you get the same ratios.
Ok it's all relative.

Of course... If a captain plays majors, the average rank he plays might be only 200 points above his own score. If a 400p cook always plays corporals 1st class, that's about 850 points higher... Who is the braver one?

It must be relative, not about ranks.

[yeti_c]

This can be simplified can it not...

If you lose a game to someone on the same points as you - you lose 20 points...

Less than 20 points means they're higher than you - more than 20 points means they're less than you...

So using that piece of information - you can work out the ratio of who beats you very easily... of course -> the next step is to take into account who else was playing... -> Again that can be done with the ratios mentioned above.

So - you can create this ratio number that will give you a comparison of who you play against.

C.

PS - obviously this falls down when playing extreme rank differences - due to the cap of 100 points.

[FabledIntegral]

I think you missed the point.

Maybe just a bit, thanks for your explanation. But my question was what you do with those scores afterwards. You can't get them into the win/loss-substraction I proposed, can you? Cause they aren't win/loss points, they're "colleague loser" points

Sure but these are ratios. Double everyone's scores and you get the same ratios.
Ok it's all relative.

Of course... If a captain plays majors, the average rank he plays might be only 200 points above his own score. If a 400p cook always plays corporals 1st class, that's about 850 points higher... Who is the braver one?

It must be relative, not about ranks.

Sad thing is there's virtually no difference between a 400p cook and 799 p cook.

[max is gr8]

Sad thing is there's virtually no difference between a 400p cook and 799 p cook.

Yes there is. I've been below 400, and I'm a high(ish) rank now. The difference between the two is very noticable. I can notice it. People say the difference between the top 10 is indistinguishable by players a lot lower, but among the top 10 they can tell the difference easily. It depends on the position you are in.

From the 2000+ point person it may be hard to tell the difference
From the 300 point person it is a great difference

[FabledIntegral]

Most 400 pt vs 700 pt just auto whatever the hell is next to them - no noticable strategy whatsoever. I refuse to believe that a 700 pt cook is any more competent.

[saaimen]

Hey guys, don't fight over this... I just made my example quite extreme to make it easier to get my point. You could see the difference with a private or a sergeant just as easily.

It must be relative, not about ranks.

THAT was my point.

[max is gr8]

Most 400 pt vs 700 pt just auto whatever the hell is next to them - no noticeable strategy whatsoever. I refuse to believe that a 700 pt cook is any more competent.

Are you saying when I was around 400 points I auto-d everything. What people fail to realise is the only difference between me playing now and then is me playing terminator or assassin and me playing standard. I haven't changed strategy. Ask anyone I've played in the past while and in the far past. No Strategy has changed, either the whole of CC has got worse or you have to admit that I have changed my game settings, they have strategy just strategy that is not suited to the game settings.

[/OFFTOPIC]

Back on topic my average rank would probably be lower, I don't think I should be penalised for my former preferred opponents. I used to play loads of cooks, but now I'm higher it's harder to play people of a similar rank as there are fewer

[saaimen]

That's what I meant. You shouldn't be penalised now, saying you play mostly cooks, as you were a cook when you played cooks. Get it?
The only thing it should show is whether you play mostly opponents with a higher or a lower score than yourself at the time of playing.

[Unbathed]

If this moves up to the "Maybe" list, I think a moving average would be more useful than a raw average, and possibly easier to implement.

"Geometric moving average" is computationally efficient, because you only need to track the current value and the new value.

[Ditocoaf]

I just want there to be, in your profile, the average rank of your opponents. This would be really telling. We can see who just picks off noobs, and who doesn't bother to play corporals like myself.

[saaimen]

But why do you have to see the average rank? As we've said before, your own rank changes overtime so the average rank of your opponents doesn't say anything. The only thing that gives a fair and true image is the relative score of your opponents compared to your own at the time of game end. And that isn't influenced by your own rank.

[Ditocoaf]

But why do you have to see the average rank? As we've said before, your own rank changes overtime so the average rank of your opponents doesn't say anything. The only thing that gives a fair and true image is the relative score of your opponents compared to your own at the time of game end. And that isn't influenced by your own rank.

It's helpful with things like: Klobber only plays New Recruits. Now, we could look at his rank, and see that he has a really weak Average Opponent. Probably about 1050.

[saaimen]

If you do it with the average score he wins, that'll be about 8 points or so? Just guessing. That's obvious enough, all his opponents have a much lower score than he does.
If you give the average rank of his opponent, and Klobber suddenly demotes bigtime... Let's say he becomes a private. Then the average rank he plays (new recruits) is even a bit higher than his! So there's nothing special about that anymore. But if you say "He averagely gained 8 points per win so far", you can see he plays lower ranks, and you can tell he's a puss

[max is gr8]

I play anyone who joins my games. I never join games I only create. Should this be punished? I think it should be the average points on games you join

[BaldAdonis]

It's helpful with things like: Klobber only plays New Recruits. Now, we could look at his rank, and see that he has a really weak Average Opponent. Probably about 1050.

A ratio would tell you something like "Klobber's average opponent has 40% of his score". It still tells you the same thing, but it's easier to do, and more indicative of where people should be playing.

Suppose that you start at 1000 points, and only play people who are exactly at your rank. You're really lucky, so you win 100 games in a row. Now you have 2000 points, and your average opponent has had the same score as you (because you were trying to be fair), so your ratio would be 100%. In terms of rank and points though, your opponents average only 1500, so it looks like you're picking on lower ranked players. You have to use ratios.

[chipv]

It's helpful with things like: Klobber only plays New Recruits. Now, we could look at his rank, and see that he has a really weak Average Opponent. Probably about 1050.

A ratio would tell you something like "Klobber's average opponent has 40% of his score". It still tells you the same thing, but it's easier to do, and more indicative of where people should be playing.

Suppose that you start at 1000 points, and only play people who are exactly at your rank. You're really lucky, so you win 100 games in a row. Now you have 2000 points, and your average opponent has had the same score as you (because you were trying to be fair), so your ratio would be 100%. In terms of rank and points though, your opponents average only 1500, so it looks like you're picking on lower ranked players. You have to use ratios.

If you win 100 1v1 games against exactly players with the same score at game end then you will be on 3000 not 2000 meaning the opponents average is 2000. (You will gain 20pts per game).

The problems using the log to derive the relative figure have already been set out. There is a cap of 100pts meaning if you beat a player more than 5 times your score it will only count as exactly 5 times your score. The missing logs will also skew the data.The whole reasoning behind seeing whether or not the games can be ordered by time finished is to recover the points from the missing logs by effectively redoing the calculation. Even this is not good enough because of terminator games. A time stamp of each points change would be the best way but as I said before this is unlikely but the accuracy with ordering the games is likely to be better. You also have the choice of not enforcing the 100 point cap when redoing the calculation.

Lets proceed along the games log path as it at least is something we can get out hands on (everything else is hypothetical at present).
You will end up with a relative ratio. Since we have added relative ratios together, multiplying that number with the player's current score will give an average rank. If the inaccuracies are acceptable with this method, I may consider adding it to Map Rank.

[chipv]

I think I'm going to go for BaldAdonis' proposal and put it into Map Rank

Proposal:

Only count players who have been beaten. This will give a measure of strength and I guess can be called Strength ratio.

So Strength ratio = ratio of average rank of defeated opponents / target rank at game end time.

I will not be counting terminator kills that do not result in a win.

Any last minute amendments welcome.

(Not counting losses)

[saaimen]

Not a chance we could actually get this into standard CC? (Or something like it, after some more discussion?)

[chipv]

I failed to underline that this is relative ranks, which I believe is exactly what you wanted?

(Rank of opponents / target rank at game end = relative rank at game end)

As for integration into CC, well by all means continue discussing, there is every chance this will be taken up if you're prepared to wait.

[saaimen]

Good, thanks!
As for Map Rank I'm not sure, but since I found out that both Clickies and Greasemonkey only work on FireFox, I've stopped checking all add-ons. I'm kinda hung on my Mac's Safari. That's why I'm eager to get things I like into standard CC

[chipv]

Good, thanks!
As for Map Rank I'm not sure, but since I found out that both Clickies and Greasemonkey only work on FireFox, I've stopped checking all add-ons. I'm kinda hung on my Mac's Safari. That's why I'm eager to get things I like into standard CC

Regarding your small text - Map Ranks works for Explorer also but not Safari.