New Score Calculation Scheme: Limiting Farming

[Ramned]

INTRODUCTION

My reason for this system is to promote players to play people nearer their own skill level. That way players cannot simply take a free ride to the #1 spot on the leader-board through playing noobs. This system allows more competition at the top of the board.

This would not destroy farming (for those who enjoy it) but it would hurt the efficiency of farming (for those who do it to achieve high scores). On the simplest level, this system limits the range you may have for farming.

Currently, one may have a score of 6000 and play players with a score of 600 and climb. Under this system, I have attempted a range of about 3,000 points but this may be adjusted by changing a the constant in the formula (as discussed below).









Players are scored using the following formula:

New RATING = Old RATING + K * (Score - Win Expectancy)
----------------------------------------------------------------------------------------
The terms in the above formula are described:


K is a constant (32 for 0-3499, 24 for 3500 - 4799, 16 for 4800 and above)

NOTE: The range of scores (i.e. 0 - 3499) is an educated guess of what would work, this may be fixed. The purpose of the constant is to allow the administrators to decide "what is the range of farming that we can allow? 1,000 points? 3,000? 6,000 (as it is now?))

Score is 1 for a win and 0 for a loss.

Win Expectancy is calculated using the following formula :

Win Expectancy = 1 / (10^((Opponent Score- Your Score)/400)+1)

The Win Expectancy is used in the rating calculation but is interesting on its own. For example, the calculation below is for a rating difference of 200. This shows that if you have a rating 200 points higher than another player, you can expect to beat them, on average, three times for each four games played.

Win Expectancy = 1 / (10^(-200/400)+1) = 0.76

Note: ^ = "to the power of", e.g. 2^3=8.

Two major components
1) Win expectancy takes into account the rating differences, the current formula does on a level so that it is possible to gain points such that a score difference of 5,000+ still receives points for a win. Which is simply ridiculous, as the win expectancy at this point is 100% for the the player with 5,000 points over a player with 100. The win expectancy weights the score. It is possible to have a win expectancy of 100% (a value of 1) thus you gain 0 points for beating complete newbies.

2) The constant k is a second limitation on score inflation. Once you reach some score (determined experimentally) the constant value serves to limit continuous score inflation.
-------------------------------------------------------------------------------------------------------


IMPORTANT NOTE: IN THE FORMULA, THE TERM SCORE IS NOT THE SCORE OF THE PLAYERS. READ THE OUTLINE. RATING IS THE TERM DESCRIBING THE PLAYERS' SCORES.

[GrimReaper.]

sounds good

[PepperJack]

I'm guessing that max (whom I have actually grown to appreciate) will reply something along these lines....


ROFLMAO so what? I only get 4 pts for a win now. Do you think losing out on 3 points will stop me from playing the types of games I enjoy. Farming has its own inherent risks. If I lose a game to a n00b now, its a hundred pointer. If I win its 4 points. You can keep your 3 points with your new system. I'll still play my games.

Good luck turning the entire system on its head...

[Ramned]

In the new system, maxatstuy, who has 5,100 pts, would not gain any points from playing any opponent under the score of 2,000.

Thus, we see from this stat that the scoreboard would condense. The ranking system would have to condense proportionally (i.e. a colonel and major are 200 pts apart, but still hard to get there as you win / lose less).

This proposed formula is a BIG change! Is it worth it? I'm all for stripping down the scoreboard and implementing this new formula, after editing it. (Of course, I am one of the few who'd go this far).

[e_i_pi]

In the new system, maxatstuy, who has 5,100 pts, would not gain any points from playing any opponent under the score of 2,000.

Thus, we see from this stat that the scoreboard would condense. The ranking system would have to condense proportionally (i.e. a colonel and major are 200 pts apart, but still hard to get there as you win / lose less).

This proposed formula is a BIG change! Is it worth it?

If we're talking about changes, then changing sites is most likely the easiest and best solution

[Ramned]

In the new system, maxatstuy, who has 5,100 pts, would not gain any points from playing any opponent under the score of 2,000.

Thus, we see from this stat that the scoreboard would condense. The ranking system would have to condense proportionally (i.e. a colonel and major are 200 pts apart, but still hard to get there as you win / lose less).

This proposed formula is a BIG change! Is it worth it?

If we're talking about changes, then changing sites is most likely the easiest and best solution

The change I speak of is statistical entirely. Scores would range from 0 - 3000 with ideal conditions.

[FarangDemon]

It was confusing that you use the word "score" in the formula but it doesn't refer to what score is commonly referred to on CC. If you wrote "win=1 / loss=0" it would be more readily understandable. This is not a dis, I think yours is a really promising dea and I'd like people to understand it better.

I put the numbers into my excel spreadsheet and played around to see what I would win/loss by playing opponents within +/- 700 points of myself:

opp you exp WIN LOSS
3500 2800 0.02 31 -1
3400 2800 0.03 31 -1
3300 2800 0.05 30 -2
3200 2800 0.09 29 -3
3100 2800 0.15 27 -5
3000 2800 0.24 24 -8
2900 2800 0.36 20 -12
2800 2800 0.50 16 -16
2700 2800 0.64 12 -20
2600 2800 0.76 8 -24
2500 2800 0.85 5 -27
2400 2800 0.91 3 -29
2300 2800 0.95 2 -30
2200 2800 0.97 1 -31
2100 2800 0.98 1 -31

For anybody to have an incentive to play someone 100 points lower than them, they would need to be confident enough to win 60% of the games to break even. I think people could go for that.

But to play someone 200 points less would mean you need to win 75% of games to break even. I think that dice and drops come into play to make it very difficult for a competitive player to be certain they can win 75% of games against another competitive player who is only 200 points below them. So I think no one would want to play anyone 200 points below them.

I can't predict how the scores of all players would distribute over time, but lets assume uniform distribution across each "century" or 100 point range (meaning same number of players between 2700-2799 as between 4300-4399).

Since there are 60 centuries from 1 to 6000, only 4 centuries would consist of players where both parties consider a game to be worthwhile (the people within +/- 200 points of you). That means any player would only realistically be playing with the 1/15 of the total population of players that are near their skill level, which is 6.7% - the 3.3% of players ranked higher than you and the 3.3% below you.

Currently, the top 6.7% of people have scores from 1863-6000 (top guy would only play down to 2122, who would in turn play down to 1863)
the next 6.7% have scores from 1617-1863 (top guy would only play down to 1769, who would play down to 1617)
the next 6.7% have scores from 1460-1617 (top guy would only play down to 1534, who would play down to 1460)

I just picked these arbitrary example ranges to give you a feel for what level guys will and will not be playing each other.

IT ELIMINATES FARMING

I notice that using k=32, playing an opponent with score 800 points below yours gives 0 points. It would eliminate farming altogether because to come out ahead against a guy 800 points below you, you would need to reliably a guy who is only about 13% below you as skill is distributed across all players kill him 97% of the time. Can a conqueror beat a guy with about 1617 points 97% of the time? I'm not the expert on that question but I would think not.

And even if the conqueror can do that, you have proposed k=16 for the highest ranked players, which means that the weakest guy the conqueror can play and still get points (1 for winning, -15 for losing) would be a guy 600 points below him. He would need to win 94% of the time against a guy with a score of 1723. I personally don't find it possible.

But if the above can be done, then I would suggest a sharpening of the win expectation formula which would tighten the playability ranges, leaving harder potential farming targets.

[JimRocky]

If Farangs detailed extended explanation of the OP's idea is accurate; (it seemed like it was to me), it sounds like a good idea.

[Jeff Hardy]

This proposed formula is a BIG change! Is it worth it?

No.

[PepperJack]

After channeling max (a premonition apparently) did some of my own thinking about this. I do not believe it should be implemented. Changing the scoring to this would be overkill, chasing away a fly from one's forehead with a hatchet as the saying goes.

I am against harvesting n00b's but if I beat a 700 point cook, I want all 7 of my points. I play singles 6p-8p public games almost exclusively. I don't want to have to pay attention to a game that is filling to ensure that I can get points from all my opponents. What if I actually limited my play to a particular range of maps and my score shot up? Suddenly the floor on where I can get points has risen.

It seems illogical to me to make any player that actually plays worth nothing. I think you will find that most people will want to be rewarded if they perform well against an opponent regardless of if they SHOULD win. I agree with the sentiment, do not like the execution. Much prefer no points for newbie deadbeats/noob island/ranked game system/mvp of a time period/etc. ideas.

[FabledIntegral]

This screws over people who play in games with 4+ people and public games. Now if you are higher than 3500 and want a public game - and some crap players join in 4/8 slots say, 2 majors join in another two slots, and then lastly a colonel joins, are you telling me if the general wins the game he's going to get like 15 points total? Hardly.

I like the idea however of the expected wins, and your constants could be adjusted, but I feel if some new scoring system like this was implemented it would HAVE to be implemented alongside a system that let you set a point range for people to join. Aka if I hosted a game, only players with 2000+ points could join. Of course ceiling caps would exist, you could use something like "only players between 1200 and 2000 can join," etc. Or even have preset point ranges, so it would be "1-1999, 2000-2999, 3000+" limits, where you could checkmark whichever you want, even all 3 if you really wanted it so it would be the same as now (or don't check any of them).

This belongs and suggs btw.

[FarangDemon]

This screws over people who play in games with 4+ people and public games. Now if you are higher than 3500 and want a public game - and some crap players join in 4/8 slots say, 2 majors join in another two slots, and then lastly a colonel joins, are you telling me if the general wins the game he's going to get like 15 points total? Hardly.

I like the idea however of the expected wins, and your constants could be adjusted, but I feel if some new scoring system like this was implemented it would HAVE to be implemented alongside a system that let you set a point range for people to join. Aka if I hosted a game, only players with 2000+ points could join. Of course ceiling caps would exist, you could use something like "only players between 1200 and 2000 can join," etc. Or even have preset point ranges, so it would be "1-1999, 2000-2999, 3000+" limits, where you could checkmark whichever you want, even all 3 if you really wanted it so it would be the same as now (or don't check any of them).

This belongs and suggs btw.

Good points. A point range specification would be necessary because if anybody plays a player more than 200 points below their score, they get gouged by the expected win probability as you said. So if you can prevent anyone from joining who is more than 100 points below you, the greatest extent to which you can be gouged is winning 6 to losing 10.

To do this, games would automatically be set to be 1800 point games or 1900 point games or 2000 point games, whatever the greatest multiple of 100 below them is. Nobody below the point value would be permitted to join because they would be able to win too much and lose too little while being just slightly lower skilled. Also give players the option to let themselves be point gouged by creating lower point games than they actually are, just so the same freedoms are maintained.

[FabledIntegral]

This screws over people who play in games with 4+ people and public games. Now if you are higher than 3500 and want a public game - and some crap players join in 4/8 slots say, 2 majors join in another two slots, and then lastly a colonel joins, are you telling me if the general wins the game he's going to get like 15 points total? Hardly.

I like the idea however of the expected wins, and your constants could be adjusted, but I feel if some new scoring system like this was implemented it would HAVE to be implemented alongside a system that let you set a point range for people to join. Aka if I hosted a game, only players with 2000+ points could join. Of course ceiling caps would exist, you could use something like "only players between 1200 and 2000 can join," etc. Or even have preset point ranges, so it would be "1-1999, 2000-2999, 3000+" limits, where you could checkmark whichever you want, even all 3 if you really wanted it so it would be the same as now (or don't check any of them).

This belongs and suggs btw.

Good points. A point range specification would be necessary because if anybody plays a player more than 200 points below their score, they get gouged by the expected win probability as you said. So if you can prevent anyone from joining who is more than 100 points below you, the greatest extent to which you can be gouged is winning 6 to losing 10.

To do this, games would automatically be set to be 1800 point games or 1900 point games or 2000 point games, whatever the greatest multiple of 100 below them is. Nobody below the point value would be permitted to join because they would be able to win too much and lose too little while being just slightly lower skilled. Also give players the option to let themselves be point gouged by creating lower point games than they actually are, just so the same freedoms are maintained.

If you had to play players within 200 points - 8 player speed games would be nonexistant. Especially when the main group in such games compose of majors, colonels, brigs, and generals. At the very least it would be within 1000 points, not 200. You going to tell me with what you propose, certain generals shouldn't even play each other?

[FarangDemon]

If you had to play players within 200 points - 8 player speed games would be nonexistant. Especially when the main group in such games compose of majors, colonels, brigs, and generals. At the very least it would be within 1000 points, not 200. You going to tell me with what you propose, certain generals shouldn't even play each other?

That would be bad, but it wouldn't happen.

Assuming scores are distributed uniformly from 1-6000, each century contains 1/60 of all players. 1/60 of all CCers corresponds to the guys with scores between 2385 and 5902. In the proposed system these guys would be between 5900 and 6000. So the conqueror would be able to play a guy with a current score 2385 and still get 6 points for winning or -10 for losing, meaning he'd have to win 62% of the time to break even against him.

I have no idea how uniform the distribution would actually be though.

[FabledIntegral]

If you had to play players within 200 points - 8 player speed games would be nonexistant. Especially when the main group in such games compose of majors, colonels, brigs, and generals. At the very least it would be within 1000 points, not 200. You going to tell me with what you propose, certain generals shouldn't even play each other?

That would be bad, but it wouldn't happen.

Assuming scores are distributed uniformly from 1-6000, each century contains 1/60 of all players. 1/60 of all CCers corresponds to the guys with scores between 2385 and 5902. In the proposed system these guys would be between 5900 and 6000. So the conqueror would be able to play a guy with a current score 2385 and still get 6 points for winning or -10 for losing, meaning he'd have to win 62% of the time to break even against him.

I have no idea how uniform the distribution would actually be though.

First of all - scores are never uniformally distributed, they would be in a bell curve. If you'd taken statistics you'd know that. And CC atm is what I would consider most likely a bell curve. So your initial statistics are off.

Second - this would promote farming more than it would hurt. Currently - if you have a rank 6000 play a rank 2385, he'd get something like 8 points for winning and 45 points for losing. That was an estimate. But yeah - you can see the difference. SO how would this help farming...

[Ramned]

Well thanks for all of the suggestions guys. I have realized that primarily due to 4+ player games it would be hard for this to work. Besides, they eliminated farming well enough by....banning it.

[FarangDemon]

First of all - scores are never uniformally distributed, they would be in a bell curve. If you'd taken statistics you'd know that. And CC atm is what I would consider most likely a bell curve. So your initial statistics are off.

My statistics class did not teach me that the distribution of CC scores is normal.

There are many different types of distributions in real life. The normal distribution is just one of them. While normal distribution is used often in statistics class, you cannot just say that something is normally distributed without any reason to back it up.

Distribution of scores is a very complex matter.

I want it to be a pure function of peer-to-peer skill. Currently it is a function of skill and the amount of points gained by farming. So even if we know how skill is distributed (which we don't), the distribution of scores gets distorted by the effect farming has on inflating some players scores.

I suspect it won't be exactly uniformly distributed and that it will taper off somewhat toward the higher ranked.

Second - this would promote farming more than it would hurt. Currently - if you have a rank 6000 play a rank 2385, he'd get something like 8 points for winning and 45 points for losing. That was an estimate. But yeah - you can see the difference. SO how would this help farming...

I don't get your point.

If you understand the formula for gaining points suggested here, you can see that farming would be impossible - the math has already been explained in detail in my first few posts and I don't want to clutter the thread by repeating it all again. The crux is that you cannot successfully farm someone that is only within a few percentile below you on the skill continuum.

[FarangDemon]

I have realized that primarily due to 4+ player games it would be hard for this to work.

Why wouldn't it work for those games?

Besides, they eliminated farming well enough by....banning it.

It won't do anything to make CC more competitive because you can still get tons of points from low ranked.

It is still possible to gain points by farming noobs in team games, too.