General Scoreboard/Scoring Discussion

[Chilly]

On the topic of scoreboards and less on total score. Going off of th-child's idea and chilly's. I like them.

But what about an all time Muy Head Honcho scoreboard. I am of the belief you can't be crowned King of majcom unless you play all game variants and maps. Fixed rate, Assassin, team play, 1v1, multiplayer singles, etc. Can you be a champion without playing fixed force? Or never stepped foot into a team game? Or never took a win on assassins?

So pick the smaller scoreboards, once built out. And choose a set of 5 or so boards that will make up the best player. You would have to have played 20 games with some sort of matrix on playing all levels of players. The last 20 games payed on any scoreboard make up a winning %. Average out the winning % of all boards to determine who is boss.

I'd go with setting up say 7 scoreboards (just picked that out of thin air). Then the top position on each scoreboard gets 100 points, second place gets 99 points, and so on. Then the master scoreboard would be a sum of points you earned on each one. I'd probably double weight 1v1, 2v2 and small FFA (6 or less?). If you were on a team scoreboard more than once (ie. with multiple teammates) you only get credit for the top place.

I might even say go with 10 scoreboards (3 double weighted - so 13 * 100 total pts), but only pick the top 10 scoreboard points. So if you really don't like fixed, or assassin or crusades, you still have a chance to get on the board. Also, this allows people who don't have gold a chance on the master board (since they can't do fixed or crusades). Of course if your lowest was one of the double weighted boards you would only get to drop that and one of the single weighted.

This would put the max score at 1000, but scores above 600-700 would probably be pretty awesome.

Again, just continuing to think out loud.

 

[Bluebonnet]

chilly, liking it more and more.

1 thing i cant wrap my head around though. if you go into a game blind in doubles. and you have to pick your partner first. a lot of people wont be playing. there is a strong contigent of payers who just join teams because someone is needed. i often join team games without a full team and just kick it and wait for someone to show.

also, quads on up will take a big hit. it is easy to set up one side with a team. the other side though is often comprised of people who just fill out the roster as described above.

any thoughts on how to alleviate?

 

[Chilly]

chilly, liking it more and more.

1 thing i cant wrap my head around though. if you go into a game blind in doubles. and you have to pick your partner first. a lot of people wont be playing. there is a strong contigent of payers who just join teams because someone is needed. i often join team games without a full team and just kick it and wait for someone to show.

also, quads on up will take a big hit. it is easy to set up one side with a team. the other side though is often comprised of people who just fill out the roster as described above.

any thoughts on how to alleviate?

It's like the lines at a ski resort. You can register with a teammate, or hop in the singles line.

Basically, you and however many buddies register as a "team" and then can join any game where your team is smaller than the # of slots for that game.

 

[Cardinalsrule]

I'd go with setting up say 7 scoreboards (just picked that out of thin air). Then the top position on each scoreboard gets 100 points, second place gets 99 points, and so on. Then the master scoreboard would be a sum of points you earned on each one. I'd probably double weight 1v1, 2v2 and small FFA (6 or less?). If you were on a team scoreboard more than once (ie. with multiple teammates) you only get credit for the top place.

I might even say go with 10 scoreboards (3 double weighted - so 13 * 100 total pts), but only pick the top 10 scoreboard points. So if you really don't like fixed, or assassin or crusades, you still have a chance to get on the board. Also, this allows people who don't have gold a chance on the master board (since they can't do fixed or crusades). Of course if your lowest was one of the double weighted boards you would only get to drop that and one of the single weighted.

This would put the max score at 1000, but scores above 600-700 would probably be pretty awesome.

Again, just continuing to think out loud.

So, if we had 10 scoreboards or categories that are used to make one 'best overall', what would they be? (we obviously can't use maps, too many)

Singles
doubles
triples
quads
sixers
crusades
assassin
mercenary
tournament games won, or tournaments won
most consecutive wins
points score (where you are on the points scoreboard)
most kills
most decorated (obviously these last 2 favor me a lot , but I think they have to be included)

This is 13 that I think would be worthy, anyone think we should have others? Take some of these off and replace them?
And which would be the 3 double weighted ones? I'd say singles for sure, what else? points?

And if it comes to it, maybe if others think of other categories, boss man can run one of his polls to pick the top 10.

 

[Bluebonnet]

crusades wouldnt work. less than a dozen played i think in majcom history.

mercenary to me is questionable. no different skill sets than chance to win entire game. same play style. just a different point allocation.

i would add fixed force. even though open only to gold members, it does have a different thought process and skill set to play.

some thoughts need to be put out on team games. head to head teams are entirely different than 3 or 4 teams in a game.

i would also add a score board of 1v1. if the grandmaster is a composite of the different scorebaords, i am not sure if i would add it to the grandmaster criteria. but it should have its own scoreboard.

 

[Cardinalsrule]

OK, so we take out crusades and mercs, I'd like to add winning percentage - number of kills/number of games.

For example, my % is 5853/3677 = 1.59

just for edification, a few other notable people:
gtivan - 2374/2359 = 1.01
mereltje89 - 897/807 = 1.11
robinette - 418/239 = 1.75
chilly - 1414/924 = 1.53
dalinar - 661/558 = 1.18
masterjskye - 799/631 = 1.27
ajdedo - 626/399 = 1.57

(OK, I'm going to stop now because I'm tired of it, but you get the idea. If you're unhappy that I did (or didn't) put you in this, sorry) :boxing:

The only thing arguing against using this metric is that people playing a lot of team games (myself, for instance) get a boost from getting credit for killing all of the others. But I think it's still worthwhile to consider.

 

[Bluebonnet]

For edification. I was referring to winning % of games won. Not people beat.

People who play a lot of 1v1 on regular maps face disadvantage of first turn. I still play a lot of 1v1. A good player would have 60% or so game win % (depending on map). But it kills the people won %.

It wont change my game style since I like 1v1 and can live with a 60% overall win ratio on it. Plus way too late to take back the 1,000 plus 1v1 games i played.

But it can all be ironed out later when Sheriff is ready to start playing with this concept.

I would like to see my game win % on 6'ers. I play with a good team.

No need to grab it for me Sheriff, I have a spreadsheet, where if I cared enough it would take me all of 3 minutes to determine.

 

[Cardinalsrule]

For edification. I was referring to winning % of games won. Not people beat.

People who play a lot of 1v1 on regular maps face disadvantage of first turn. I still play a lot of 1v1. A good player would have 60% or so game win % (depending on map). But it kills the people won %.

It wont change my game style since I like 1v1 and can live with a 60% overall win ratio on it. Plus way too late to take back the 1,000 plus 1v1 games i played.

But it can all be ironed out later when Sheriff is ready to start playing with this concept.

I would like to see my game win % on 6'ers. I play with a good team.

No need to grab it for me Sheriff, I have a spreadsheet, where if I cared enough it would take me all of 3 minutes to determine.

That would be a different metric, which I imagine is available to Sheriff/Marcos, and anyone who cares enough to go back through their entire history. But it also skews depending on what style of games a player prefers. Someone like Cagey, who regularly plays a LOT of singles games with 8, 10, 12 players would obviously have a much lower % than a dedicated 1v1 player. For that reason, I'm not sure that I consider it a very good metric to use. But it's worth throwing out there for discussion purposes.

 

[Bluebonnet]

Nope, i still think we are talking apples and oranges on what i proposed.

"So pick the smaller scoreboards, once built out. And choose a set of 5 or so boards that will make up the best player. You would have to have played 20 games with some sort of matrix on playing all levels of players. The last 20 games payed on any scoreboard make up a winning %. Average out the winning % of all boards to determine who is boss."

So some of our current scoreboards have different methodologies to rank top players. High score, most medals, etc. But how do you determine top of the scoreboard on 1v1, doubles, or multiplayer singles? While you could do total games won, it is not a defenitive measure of how good you are. I will agree our point system has some kinks in it so you may not be able to based it on points garnered from that style of play.

But, with the new site, there is hope of good stats becoming available. So you may be able to do a ranking for each scoreboard on game win % of that specific category/setting. There may have to be a ratio of the ranks you battled in the mix for the calculation. I wouldn't want to be at the top of a 1v1 based on me finding noobs who have less than 3 games and i whip em.

Actually, instead rank, the mix could be based on if the opponent had 20+ games on the category. After 20 games anyone should be halfway decent.

Buty that is where i was going with a winning game %.

 

[Cardinalsrule]

MrMarcos is going to need a statistician to work with him to set up all those matrices, lol.

 

[Bluebonnet]

Nah, the math is easy. The trick is to show enough understandable detail without overwhelming the audience.

 

[Dalinar]

If you wanted an overall metric for skill, I'd take what I think of as "the win rate differential"

(actual win%) / (random win%) = multiple of games you win because of skill versus total chance

So in 1v1, if you win 60%, you'd have .6/ (1/2) = 1.2
You win 20% of games more than if things were random

If you won 10% of your 8-player matches, you'd have .1 / (1.8) = .8
You lose 20% more games than if things were random

There are a few issues with this metric, though it's actually the one I use to judge how I'm doing in larger matches. It gives you a base for comparison across game sizes.

First, the ceiling for the number is higher in matches with more players. In matches with 2 people, you can only win twice as many matches as if it were random - with 8 players, with ceiling would be 8.

Second and most important, is that win rates don't account for who you're playing. If Merel and I were to play each other in 1v1s a lot, our win rates would end up close to 50%, giving a win differential of about 1. If we were to hunt kitchen patrols, maybe we'd win 75% and get a multiplier of 1.5 - it depends on who you play.


To make an overall metric, you don't want to average because they don't scale perfectly (1v1 is different than 8-player). To cut down on the number of matches needed, I'd take the numbers from 1v1s, 2v2s, 8-player, and 12-player matches and add them up. That should give you a pretty good view of how skilled someone is.

 

[Dalinar]

Nah, the math is easy. The trick is to show enough understandable detail without overwhelming the audience.

And to get at the data, lol. I looked at setting up a script using the "find games" page to compute some of this stuff and the html isn't usable as is. Whoever sets this up has to have access to the DB or be a lot more skilled than I am.

 

[Dalinar]

Actually, instead rank, the mix could be based on if the opponent had 20+ games on the category. After 20 games anyone should be halfway decent.

But that is where i was going with a winning game %.

I have disagree about 20 games meaning people are halfway decent. There are lots of people who stay kitchen patrols or lieutenants after 100+ matches, and one's win rate against them is going to be different than against colonels. I think the win rate versus players at a given skill level would be a much better metric.
Take win rates for against players below 1000 and then every 500 (maybe 600? I don't have a clear line) point bracket above.
Clearly, a 50% 1v1 win rate versus colonels and 50% against kitchen patrols mean very different things. I think this applies to multiplayer as well, though the only fix I could think of would be to average the ranks of one's opponents at the time the match ended.

P.S. I know I'm triple posting, but it's easier to avoid walls of text.

 

[Bluebonnet]

And to get at the data, lol. I looked at setting up a script using the "find games" page to compute some of this stuff and the html isn't usable as is. Whoever sets this up has to have access to the DB or be a lot more skilled than I am.

Agreed, it has to be something majcom sets up. However, we aren't talking physics for the math. it is just simple addition, division, etc. And therre has been a lot of discussion in other threads on a stats board which then drives the overall scoreboards.

I have an excel where i keep track of my wins. So for team games and maps, i could compute these fairly easilyt.

 

[Bluebonnet]

If you wanted an overall metric for skill, I'd take what I think of as "the win rate differential"

(actual win%) / (random win%) = multiple of games you win because of skill versus total chance

So in 1v1, if you win 60%, you'd have .6/ (1/2) = 1.2
You win 20% of games more than if things were random

If you won 10% of your 8-player matches, you'd have .1 / (1.8) = .8
You lose 20% more games than if things were random

There are a few issues with this metric, though it's actually the one I use to judge how I'm doing in larger matches. It gives you a base for comparison across game sizes.

First, the ceiling for the number is higher in matches with more players. In matches with 2 people, you can only win twice as many matches as if it were random - with 8 players, with ceiling would be 8.

Second and most important, is that win rates don't account for who you're playing. If Merel and I were to play each other in 1v1s a lot, our win rates would end up close to 50%, giving a win differential of about 1. If we were to hunt kitchen patrols, maybe we'd win 75% and get a multiplier of 1.5 - it depends on who you play.


To make an overall metric, you don't want to average because they don't scale perfectly (1v1 is different than 8-player). To cut down on the number of matches needed, I'd take the numbers from 1v1s, 2v2s, 8-player, and 12-player matches and add them up. That should give you a pretty good view of how skilled someone is.

I like the concept a lot. I think it may be boarderline for the average person to understand. But majcom should be above average in understanding of these thoughts.

And yes, as i stated, all settings (1v1 or 8 payer singles) should not be combined. A lot of thought will have to be put into how "who you play" factors into process. But untill majcom gets ready to start on it, I am letting it drop. This will take some more advanced script. Is sheriff willing to pay and how much will it cost??

 

[Bluebonnet]

I have disagree about 20 games meaning people are halfway decent. There are lots of people who stay kitchen patrols or lieutenants after 100+ matches, and one's win rate against them is going to be different than against colonels. I think the win rate versus players at a given skill level would be a much better metric.
Take win rates for against players below 1000 and then every 500 (maybe 600? I don't have a clear line) point bracket above.
Clearly, a 50% 1v1 win rate versus colonels and 50% against kitchen patrols mean very different things. I think this applies to multiplayer as well, though the only fix I could think of would be to average the ranks of one's opponents at the time the match ended.

P.S. I know I'm triple posting, but it's easier to avoid walls of text.

Just brainstorming....

Evenrually, if something is instigated. It should be ran through models to determine if the results match what is expected. No that we have 4 years of data, a lot of testing can be done.

 

[giebenrath]

Open source all of the stats. Seriously. Dump the DB out to a google spreadsheet or something. Let people dig and play. Someone is bound to come up with something interesting. Analyze win/loss rates by rank over time, look at first mover advantages by map to see if something should be tweaked, let people simulate new scoring approaches...plenty of things that could be looked at but that certainly would take too much valuable time from the folks working so hard to just keep things up and running (and re-coded, etc.).

Maybe you get a lot of crap, but you don't have to listen to any of it anyway. At least it would be crap with data behind it...maybe...

 

[AuraCraft]

I sympathise with cardinals' perspective but it's also a fact that you get a lot of compliments just for achieving high ranks. ppl like to rank high and ppl use rank as a measure of skill level.
I was thinking perhaps paying members should lose 20% less points with a maximum of 37 points when losing against grunts.
Part of the reason why we use this points system is the assumption that the more skills you have the more likely you are to win, the less points you get when you beat someone who ranks lower. but the extent to which this is the case is off.
Grunts are restricted to playing 4 games so most of them play a small selection of types of games (settings and maps) and naturally specialise in those few games whereas the variety of games played by ppl who play a dozen(s) of games at a time makes it harder not to make settings mistakes. E.g. a lot of 23oo points realtimer grunts are experts 12 domainers and they would beat tournament majors in best of 3 on 12 domains 1 path. Or as a cadet I understood the evolved map within 1 or 2 games
now this may seem like pay to win which I loathe, but in fact the dice would still roll the same

and since the title is scoreboard; Id prefer all board to be current rather than all time... why should ppl who quit yrs ago still be on the board? With a shrinking population it's hard enough to get udo medals let alone if the board is littered with quitters

 

[Redstorm]

and since the title is scoreboard; Id prefer all board to be current rather than all time... why should ppl who quit yrs ago still be on the board? With a shrinking population it's hard enough to get udo medals let alone if the board is littered with quitters

Yes I agree. Perhaps even a reset once a year would be better. That would allow us to crown an annual champ as well. I do think all time has its value though and have suggested in the past to add to scoreboard with an all time record with say a top 5 or 10 players.