Defeated opponent per game ratio.

[Badorties]

So the defeated opponent per game ratio is an interesting metric. I am wondering if it is a worthy standin for "overall skill" vs the variable point score.

For the record, I am 1.28. AAFits is 1.14. KungFu is 1.57 Robineete is 1.96.

It is a stat you can't inflate by playing 1v1 games, which is why Fitz's isn't higher.

So what do you think of this stat? Is it possible to game/cheat? does it reflect skill, or does it skew towards certain settings?

 

[Cardinalsrule]

I think its an interesting stat, don't see how you could 'game/cheat' it, because in order to increase your ratio, you need to play and WIN games with higher numbers of players. For example, I avoid 1v1 games like the plague, because I always seem to start 2nd and inevitably have a dicerape on round 2, and then I'm screwed. For that reason, because I play most games with higher #s of players, my ratio is a respectable 1.42. So, its an interesting metric, and when used with others, such as overall score, can be an indicator of skill. By itself, though, it can be misleading, as you can see by the fact that Fitz's score isn't very high, and he's obviously a good player.

Interestingly, Tutor965, who has a thread about his recent success, is currently at 1.03.

 

[corcovado]

I think it is a good stat, the only way i see to "cheat it" would be to play a lot of quads and trips games, if you win your fair share you will average between 1.5 and 2.0 opponents per game instead of 1.0.

 

[ORBOTRON]

I think it's a cool metric. Like Card said, it favors large games, which are inherently more difficult to abuse. I think it would make a great ingredient in formulating an overall skill score.

 

[Rover]

I also think it's an interesting metric.

Check out bean who has a 2.44 ratio (although she has played fewer games than most other people). Anyone know anyone higher?

 

[Restrepo]

can I ask how you can see that stat?

 

[Badorties]

can I ask how you can see that stat?

just divide their opponents defeated by games played. its all on a person page.

 

[Cardinalsrule]

so, to follow up on what Orbo said, in order to formulate an 'overall skill' score, you could use total score, defeated opponents per game, and what else?
should consecutive wins be considered?

 

[ORBOTRON]

While eating dinner tonight, I couldn't stop thinking about this. I think a great way of judging overall score would be to measure a players OD/Game in each of a bunch of settings, then either simply average those numbers, or possibly use a weighted average.
This would go a long way towards curbing farming, encouraging game variation, and rewarding well rounded players.
An example of how it could work:

Make a category for every player setting 1v1, 3, 4, 5 etc. up to 12. Keep track of their average in each category and average them.
This means a player has to get a good average in every category to acheive a good overall score, avoiding a setting results in a 0, and zeros are very bad for an average.

The only possible problem I see with this is a newcomer with beginners luck coming in and winning one of each setting. Even this is short lived though, as they will be forced to play more games or be removed from the scoreboard. I would recomend shortening the inactive rule to like 2 weeks though, to prevent high score sitting like this.

 

[ORBOTRON]

I think this would force cheaters to cheat over a wider range of games also, which would make them easier to spot. Let the multis play 1v1 and attempt to point dump till the cows come home, it will do absolutely nothing for their score.

The more I think about this the more I like it.

 

[mapguy]

So the defeated opponent per game ratio is an interesting metric. I am wondering if it is a worthy standin for "overall skill" vs the variable point score.

For the record, I am 1.28. AAFits is 1.14. KungFu is 1.57 Robineete is 1.96.

It is a stat you can't inflate by playing 1v1 games, which is why Fitz's isn't higher.

So what do you think of this stat? Is it possible to game/cheat? does it reflect skill, or does it skew towards certain settings?

DOG can indeed be gamed very easily, Just play a bunch of 8-man, and 12-man, quads.

 

[Cardinalsrule]

don't think that would help much. An 8-man quad would only get you 4. You'd have to win a very high percentage to make that work.....

 

[mapguy]

don't think that would help much. An 8-man quad would only get you 4. You'd have to win a very high percentage to make that work.....

You do not get it card. If you played say 100 games, you could expect to win at least 50 of them. That is DOG of 2.0. you can further raise your score if you play anything but 1v1s.

 

[Badorties]

You do not get it card. If you played say 100 games, you could expect to win at least 50 of them. That is DOG of 2.0. you can further raise your score if you play anything but 1v1s.

Um... if you play 100 10-man games, you are expected to win only 10% of them, not 50%. Winning 1 of 10 would not raise the DOG. so you need to be better than average to raise it.

 

[mapguy]

Um... if you play 100 10-man games, you are expected to win only 10% of them, not 50%. Winning 1 of 10 would not raise the DOG. so you need to be better than average to raise it.

No, we were talking about 8-man quads.

 

[Badorties]

No, we were talking about 8-man quads.

ah right. team games do kind of debunk that stats usefulness.

 

[ORBOTRON]

I really think the key to this, or any other stat, is to seperate it by setting. If you isolate the DOG of quads games, and simply average it with the DOG of other settings, it totally eliminates the possibility of someone gaming the system with one setting.
For example in Maps example, say they get a DOG of 4 in quads, that's great, but if they play no other setting, that 4 gets averaged out by zeros in 4 or however many other settings we include. Say we only include Singles, Doubles, Triples and Quads. If they only play quads, that 4 turns into a 1.

 

[mapguy]

Also, if you were to calculate the dog for any team games as follows, the math is brought back into line with the single games -
Instead of giving a member of a team the total amount of defeated opponents, only award them their share of the win.
So for example in a 8-man quads. the members of the winning team split the bounty evenly and each receive only 1 dog. lol.

 

[Cardinalsrule]

Orbo's idea would work, but is a pretty complicated formula to set up. For example, is quads settings going to count equally with singles? I don't think it should. Very few people play enough quads games so that they could be compared equally with singles/doubles games. I think you'd have to figure out a weighted average by games played of each type..... and also figure out a fair way to 'penalize' people who only play games of one type. It would probably have to involve a penalty for too high a percentage of only one type compared to the other three.

in short, to make it work, you've got to develop a very complicated paradigm. Who's your resident PhD in statistics?

 

[Badorties]

I've started another thread about this topic.

http://www.majorcommand.com/forums/threads/2319-Developing-the-Command-Score