Season 3 Top Player and All-Stars announced...

[Dako]

Current scores. Or clan seeds in the league (their ending positions).

Anyway, I'd implement some kind of weight system for the games. And that will resolve win percentage, because with the weight we would be counting difficulty system.

As for the results - are you trying me to pick one at random when we are talking about math solution? I can pick one, but that would not be the correct result because I have not enough data.

[jpcloet]

I'm simply giving examples to show how not easy this is. Current scores would be much easier than league final seeding; plus that would not work for wars. I'd have to check how maprank works in terms of Relative rating to see if I could use that.

[Dako]

I know that it is not easy. And we are talking about league, not wars. Let's worry about wars a bit later.

[jpcloet]

I know that it is not easy. And we are talking about league, not wars. Let's worry about wars a bit later.

I wish that maprank had the extra game box like the tournament script's last update. I would love to be able to maprank a player on say 10-20 games.

[Leehar]

7-1?

Why would you say 7-1? Was 3 games not enough even though they all got medals?

Perhaps, but I'm looking at it somewhat from Rodions reasoning. That the possibility of the 3-0 individual of getting to 7-1 would be significantly harder then the one who's on 7-1 by virtue of him being there already.
tbh, I'm just content with leaving it as a judgement call rather than complicating it overly much with data dives. Sometimes there are just things you can't tell purely based on numbers (I'm wondering if I'm somehow regurgitating the power rankings argument... ). Who knows, a voting system by the cla reps based on as much factual evidence you could provide wouldn't hurt either

[angola]

Who knows, a voting system by the cla reps based on as much factual evidence you could provide wouldn't hurt either

In some ways I agree, but most likely it would just turn into a popularity contest and that wouldn't reward the random folks that show up on the all-star teams.

I'll leave the math to the rest of you folks, however.

[danryan]

Another question is how do you balance home/away performance. I'd argue that winning away games should carry more weight than home games because inherently you'd expect performance to be worse on away games.

You're never going to make everyone happy, but hey, you made me happy.

[Dako]

Yes, I think away games are like 20% harder.

[jpcloet]

Another question is how do you balance home/away performance. I'd argue that winning away games should carry more weight than home games because inherently you'd expect performance to be worse on away games.

You're never going to make everyone happy, but hey, you made me happy.

Away is slightly harder, however, most cases team join based on map anyways. Feudal home vs Feudal away should not be any different. I've been watching patterns of behaviors. One of the ones I'm weary of is a player going 12-0 then stopping and not playing the last few weeks etc.


Can someone give or find me a weighting example that could work for MVP's?

[nagerous]

Another question is how do you balance home/away performance. I'd argue that winning away games should carry more weight than home games because inherently you'd expect performance to be worse on away games.

You're never going to make everyone happy, but hey, you made me happy.

Away is slightly harder, however, most cases team join based on map anyways. Feudal home vs Feudal away should not be any different. I've been watching patterns of behaviors. One of the ones I'm weary of is a player going 12-0 then stopping and not playing the last few weeks etc.

Well they are only harming their own clan if they do that, a selfish act that I can't see many following through with.

[FarangDemon]

Kill ratio is totally lame. Most valuable players is the frontier? I have not heard about a clan that loads only one guy because he is better at attacking. It always differs from game to game.

So... kill ratio as an MVP criteria is lame, I think.

I agree with Rodion, Denise and Dako - ditch kill ratio.

Although it makes sense intuitively that Away games are harder than Home, it cannot affect score in a way that makes sense. Consider two players are tied at 7-1, but one performed better on Away games. You could equally argue:

• The one that won more Away games (or has higher win % on Away) should be ranked higher because those wins were harder.
• The one that lost more Home games (or has lower win % on Home) should be ranked lower because those wins should be easier.

However much harder Away is than Home, Home is that much easier than Away, so you run into this contradiction every time.

[lord voldemort]

away in theory should carry more weight...dako had a good point that its just what player/team best fits the map or match up.
kill ratio isnt a great way imo for the obvious stated reasons

i think an interesting way would be for an example for league play.
after each matchup a clan leader submits a 3,2,1 vote for the opposing team just based on their performance for that week...tally the votes and you have a winner.
will it be the guy with the best record...maybe. but it should be a true indication of who had the best impact as a player overall. As a leader you should ask for opinions from your fellow clan mates.
just an idea...taht could be better developed

another would be to round up the top 10 or so statistically best results have the cla vote on them.

[Leehar]

Who knows, a voting system by the cla reps based on as much factual evidence you could provide wouldn't hurt either

In some ways I agree, but most likely it would just turn into a popularity contest and that wouldn't reward the random folks that show up on the all-star teams.

Thats why I suggested that there's initially some factual basis for choosing the worthy individuals and then let intuition or perception become the decider.
LoVo's suggestion has merit as well

Anyway, looks like it's the 3rd page and no one's congratulated Steve yet, so let me go ahead and be the first to do so, Gratz steve

Looks like kort was well represented with 2 more in the top 5 as well.

[Dako]

How can you rate opponents if you cannot read their game chat?

[josko.ri]

congrats to all MPV's, especially steve066. I am very proud that lot of KORT slots are in MVP's awards (7 different players awarded). also, after CLA quarters, steve066 continued with his clean sheet performance and now he is on 15-0 before semis

I would like to give my point about ranking system. kills ratio definitely shouldnt be included IMO. tactics doesnt depend on whose player makes kill, and tactics can be manipulated on the way that team who has possible MVP player make their strategy that always that player makes kill, and other players skip killing even if they have chance to do it (on the time when game is already decided).

here are 2 examples which I think are controversy:
1. Theldin is better than Rodion in Division D triples.
Theldin has 5-1 score. he played 1 game vs #1 (lost it), 2 games vs #3 and #4 and 1 game vs #5 clan. his clan was #6. so, he practically doesnt have single win against #1 and #2 clan in Division. he played only 1 game vs 1-2# clans in division. (lost it)
Rodion has 6-1 score. he played 2 games vs #2(won both), 1 game vs #3, 3 games vs #4 and 2 games vs #6 clan. so, he not only has 1 win more than Theldin, but also he played 3 games vs 2-3# clans in Division (won all 3). it shows that he had much stronger schedule than Theldin, and won one game more than him.

2. josko.ri is better than cachejob on MVP total list.
josko.ri 14-4 score (77,7%)
cachejob 21-2 score (91,3%)

while example (1) is little difference in only one more game, example 2 is huge difference and I think it shows how killing ratio is pointless. I suppose my 11-1 score in quads/triples make my killing ratio huge while 3-3 in doubles didnt reduce it too strong. he has his 2 looses in quads so probably that reduces his killing ratio a lot and finally he came below me, what I think shouldnt happen in any way considering winning%, what should be the strongest factor in MVP ranking. plus, all my wins are on different maps so probably that factor were favorized me. his winning % and much more games played are so much higher than mine that I think all others factors in overall shouldnt overcome this one.

you asked fro suggestions how to improve MVP scoring system. it is very hard question, and I may offer one solution how I think it can work better. this solution would require a ton of work but I think finally would give more reliable results. so, factors that needed to be included should be (with this order of importance):

1. winning %
2. strength of schedule (considering how many average winning % has all opponents of particular player)
3. average player opponent's opponents winning %. factor 2 also depends on this factor, so this one can maybe just be tiebreaker solution.

I think swiss system in chess works on similar system, there are just #2 and #3 points used for tiebreakers while #1 point is the main thing. but, we may give weight factors to all of these 3 points. I know it is hard and a ton of job to do, and I may help with volunteering for it for next clan league if you need me.

also, I am not sure how many games played were minimum to be considered. but, I think it shouldt be games played the factor, it should be games won. for example, if minimal games played need to be 12, then player with 11-1 are considered for MVP award while player with 11-0 is not considered for MVP award. for that reason "minimal games won" should be factor for a player to become eligible for MVP.

and finally, thanks to jp a lot for doing so lot of his effort for clan stuffs. I imagine how many hours of volunteering had to be done to make this MVP list

[jpcloet]

josko-cache comparison is off as it was only quads in the analysis. A perfect quads score ended up needing a 90% win rate or higher.

j-8-1 (88.8%)
c-15-2 (88.2%)

I think swiss system in chess works on similar system, there are just #2 and #3 points used for tiebreakers while #1 point is the main thing. but, we may give weight factors to all of these 3 points. I know it is hard and a ton of job to do, and I may help with volunteering for it for next clan league if you need me.

% alone is not good enough for me. See above 7-1 vs 3-0 argument. The swiss system is fine for head to head but does not quite apply to a league let alone an individual. In the league's case it would apply to a team and we don't need that actually. I could see the argument that Kill Ratio should be a tie-breaker and not part of the main scoring system. I'm still thinking about how to incorporate home-away since it means more manual work which I'm trying to avoid.

[Leehar]

here are 2 examples which I think are controversy:
1. Theldin is better than Rodion in Division D triples.
Theldin has 5-1 score. he played 1 game vs #1 (lost it), 2 games vs #3 and #4 and 1 game vs #5 clan. his clan was #6. so, he practically doesnt have single win against #1 and #2 clan in Division. he played only 1 game vs 1-2# clans in division. (lost it)
Rodion has 6-1 score. he played 2 games vs #2(won both), 1 game vs #3, 3 games vs #4 and 2 games vs #6 clan. so, he not only has 1 win more than Theldin, but also he played 3 games vs 2-3# clans in Division (won all 3). it shows that he had much stronger schedule than Theldin, and won one game more than him.

I would think that with strength of Schedule, Theldin had the harder time of it. Every game he played was against clans better than his, while everyone Rod played was in the weaker?

[Rodion]

Part 1

Does A still trump B due to % and "surplus"

Part 2

Does A still trump B due to % and "surplus".

I see what you're trying to do here.

While my answer would be "yes" to both questions, I don't mind the idea of trying to reward players that consistently leave their "comfort zones". I would like, however, to suggest that whatever formula you use, you make it public before CL4's first games start.

Pro: there won't be people complaining that a new criterium was used to alter the intuitive results (for instance, it is "intuitive" that 8-1 should beat 7-1; like Josko pointed out, it is also "intuitive" that a 21-2 record should beat a 14-4 record in the overall list).

Con: people might "game" the system (if the formula is pretty darn good, maybe the system can't be gamed?)

Weighing pro and con, I think the suggestion should pass.

Here is a clan war example of the top 9 players, which one would you pick for MVP?

Record Win
3-0 100%
7-1 88%
6-1 86%
5-1 83%
6-2 75%
6-2 75%
6-2 75%
5-2 71%
7-3 70%

I would consider 3-0 too small of a sample size and deprive it from eligibility. 7-1 would then win.

7-1?

Why would you say 7-1? Was 3 games not enough even though they all got medals?

Perhaps, but I'm looking at it somewhat from Rodions reasoning. That the possibility of the 3-0 individual of getting to 7-1 would be significantly harder then the one who's on 7-1 by virtue of him being there already.
tbh, I'm just content with leaving it as a judgement call rather than complicating it overly much with data dives. Sometimes there are just things you can't tell purely based on numbers (I'm wondering if I'm somehow regurgitating the power rankings argument... ). Who knows, a voting system by the cla reps based on as much factual evidence you could provide wouldn't hurt either

You're misusing my reasoning.

While in the 8-1 vs. 7-1 case it is crystal clear that 8-1 is a better position, you can't say the same when comparing 3-0 and 7-1.

Why?

You could probably try to apply the reasoning by saying it is expected that 3-0 becomes 6-2 (or 5-3), thus below 7-1. However, someone that started with a 3-0 record is presumably an above average player that will presumably win more than lose. We have no reason not to believe he wouldn't reach 8-0 with 5 more games (and thus beat 7-1), so my "reasoning" can't apply here.

Voting system - HELL NO! -> make it a purely mathematical formula that won't screw someone just because he's not a community favorite.

Another question is how do you balance home/away performance. I'd argue that winning away games should carry more weight than home games because inherently you'd expect performance to be worse on away games.

You're never going to make everyone happy, but hey, you made me happy.

Away is slightly harder, however, most cases team join based on map anyways. Feudal home vs Feudal away should not be any different. I've been watching patterns of behaviors. One of the ones I'm weary of is a player going 12-0 then stopping and not playing the last few weeks etc.

Well they are only harming their own clan if they do that, a selfish act that I can't see many following through with.

Maybe they aren't harming anyone at all? Consider a clan that did so well in the first 4 weeks that they have already locked their #1 seed. They can sit week 5 out with no harm to the clan. Unlikely, but possible.

Kill ratio is totally lame. Most valuable players is the frontier? I have not heard about a clan that loads only one guy because he is better at attacking. It always differs from game to game.

So... kill ratio as an MVP criteria is lame, I think.

I agree with Rodion, Denise and Dako - ditch kill ratio.

Although it makes sense intuitively that Away games are harder than Home, it cannot affect score in a way that makes sense. Consider two players are tied at 7-1, but one performed better on Away games. You could equally argue:

• The one that won more Away games (or has higher win % on Away) should be ranked higher because those wins were harder.
• The one that lost more Home games (or has lower win % on Home) should be ranked lower because those wins should be easier.

However much harder Away is than Home, Home is that much easier than Away, so you run into this contradiction every time.

I don't think they were trying to oppose someone that lost 4 home and won 4 away to someone that won 4 home and lost 4 away. The idea is probably to pick both "7-1"s and see who played more away games and who stuck with his favorite map/settings (so that a 7-1 with 8 away games would beat a 7-1 with 4 aways and 4 home games).

Rodion has 6-1 score

7-1.

josko-cache comparison is off as it was only quads in the analysis. A perfect quads score ended up needing a 90% win rate or higher.

So, you gave perfect quads ratings to people above 90%? That means a 100% quads guy wouldn't have gathered any advantage against a 94% quads guy? Sounds flawed.

I would think that with strength of Schedule, Theldin had the harder time of it. Every game he played was against clans better than his, while everyone Rod played was in the weaker?

I think standard SOS is absolute, not relative. That means they consider "how good were the teams you faced" and not "how good the teams you faced were compared to yours".

An interesting thing to add besides SOS would be WOP (weakness of partners), where someone would be rewarded by winning alongside less than stellar teammates. Good idea on paper, but hard to apply, in my opinion.

[jpcloet]

So, you gave perfect quads ratings to people above 90%? That means a 100% quads guy wouldn't have gathered any advantage against a 94% quads guy? Sounds flawed.

The concept stems from QB rating and unlike the QB rating (which goes above 100 to a "perfect" score of 158.2), my aim is to go to 100 score but not in 100% perfect kind of way. Meaning you could still go 10-1 and get a 100 score.

This is a good article that points out how the QB rating was developed. It is essentially a comparison to the average player in the league. You can find several flaws in this system as well.

http://www.bluedonut.com/qbrating.htm

[Rodion]

So, you gave perfect quads ratings to people above 90%? That means a 100% quads guy wouldn't have gathered any advantage against a 94% quads guy? Sounds flawed.

The concept stems from QB rating and unlike the QB rating (which goes above 100 to a "perfect" score of 158.2), my aim is to go to 100 score but not in 100% perfect kind of way. Meaning you could still go 10-1 and get a 100 score.

This is a good article that points out how the QB rating was developed. It is essentially a comparison to the average player in the league. You can find several flaws in this system as well.

http://www.bluedonut.com/qbrating.htm

I see. I'm familiar with it. I think 77.5% completion is good enough to get max. 1 TD every 8 pass attempts, 0 INTs and more than 12.5 yards per pass attempt?

Also, what do you think of my suggestion for CL4 of a predefined formula?

[jpcloet]

Also, what do you think of my suggestion for CL4 of a predefined formula?

At least people know what it is then. I like the idea of having min requirements being say min 4-4-4 in game types etc. to qualify. There were a number of top player candidates that did not play any doubles, although due to the lower number in S3, it was understandable.

[Rodion]

Also, what do you think of my suggestion for CL4 of a predefined formula?

At least people know what it is then. I like the idea of having min requirements being say min 4-4-4 in game types etc. to qualify. There were a number of top player candidates that did not play any doubles, although due to the lower number in S3, it was understandable.

Also, since doubles are usually considered to be more luck-based settings, top players are expected to help with their talents more on quads games and less on doubles.

I made a suggestion here ( viewtopic.php?f=441&t=125953&start=300&hilit=reserves - page 21) regarding eligibility minimums.

[quote=Rodion]Clans played:

90 total games (290 spots) - for MVP and reserve all-star team consideration, I think a minimum of 20 games would be appropriate (6.9% of that clan's spots).
20 doubles (40 spots) - for doubles all-star team consideration, I think a minimum of 3 doubles games would be appropriate (7.5% of that clan's spots).
30 triples (90 spots) - for triples all-star team consideration, I think a minimum of 6 triples games would be appropriate (6.67% of that clan's spots).
40 quadruples (160 spots) - for quadruples all-star team consideration, I think a minimum of 10 quadruples games would be appropriate (6.25% of that clan's spots).[/quote]

The suggestions was meant to get the ball starting, but it never did. Anyway, my proposition was of 3/6/10 minimums instead of 4/4/4.

[jpcloet]

On the doubles front there were a number of 3-0's. Again, brings us back to is 3-0 better than the one player at 8-1 or the 3 players at 6-1?

[Rodion]

On the doubles front there were a number of 3-0's. Again, brings us back to is 3-0 better than the one player at 8-1 or the 3 players at 6-1?

The suggestions was just to show how category minimums should be calibrated considering spot distribution between game types was uneven. The 3/6/10 numbers could be tweaked.