Re-Organizing DSB

General discussion

Moderators: Beer Hunter, Tembest, Entr0py

jim the chin
Posts: 315
Joined: Sat Dec 24, 2011 1:17 pm

Re: Re-Organizing DSB

Post by jim the chin »

Sorry the edit time on this forum is crap. Team 2 has an 8 flag killbonus (not 12), but I got the numbers correct. The team with 10 less kills would win by 1600 points if they dominated flagging 12:8. So I think a base kill value of 120 points is ideal.
Please delete this account. I want nothing to do with this place any more.

User avatar
Entr0py
Posts: 135
Joined: Sat Dec 17, 2011 7:54 pm

Re: Re-Organizing DSB

Post by Entr0py »

Another update:

What has been done

1. I fixed the issue with saiyan's rating (and a few others) in the recalculation of the ratings. The problem stemmed from two situations - one where people changed the capitalization of the letters in their names (which caused the program to treat the different versions of their spellings as different players), and a second where people removed themselves from leagues, causing them to not show up in the player lists.

2. I looked into the carry problem in !stats. As of right now, you can trust the Total stats line. I am aware that the ratings/carries for the specific sectors are still messed up, and will fix that tonight.

3. New/redefined commands:
3a. !top/!top10/!standings - Shows the standings for the current league
3b. !ostandings/!lifetimestandings - Shows the standings for the lifetime overall ratings
3c. !stats (Optional: Player name) - Shows the stats of yourself or requested player for the current league
3d. !rank (Optional: Player name) - Shows the rank of yourself or requested player for the current league
3e. !ostats (Optional: Player name) - Shows lifetime stats of yourself or requested player
3f. !orank (Optional: Player name) - Shows lifetime rank of yourself or requested player

What still needs to be done

1. Fix the issue with ratings/carries in !stats for the specific sectors

2. Some additional behind-the-scenes data and bot cleanup

Open questions/issues

I'm not sure what exactly you are wanting when you request that the stats be reset. Are you asking that I reset the lifetime ratings, and if so, do I stop there, or also reset lifetime kills/deaths/pracs played/etc?

I will say up front that I am against the idea of resetting lifetime stats/ratings. They're called "lifetime" for a reason. They have built up over 4 years to the point where I believe they actually do predict match outcomes fairly well (if you'd like, I could write a script that will tell us how well they predict match outcomes), and they also paint a vast picture for over 2600 players. Many of those players don't play anymore, but I'm not convinced that's a valid reason to erase their stats history.

As for "archiving" the current lifetime stats and then starting over, I don't really see any benefit from that, either.

But I will read your thoughts on it and think about it further.

Anything else right now?

Tembest
Posts: 692
Joined: Sun Dec 18, 2011 10:12 am

Re: Re-Organizing DSB

Post by Tembest »

I did not include the kill bonus in my calculations as it's under discussion whether it should exist or not.
At one point we tried kills only based on flags held. That made game horrible. Then we moved to Fixed kill + kill bonus based on flag
count, which has received more neutral feedback. Some people say it's too hard to follow scoring and the purpose of it is very minimal.
Somehow people like the fact only dings count. But when the idea was introduced, it received much positive feedback.
I don't know which way to go with it so I did not include it in my calculations..

Anyhow, I don't think it's relevant whether we have the kill bonus in these calculations. One could argue it emphasises the importance
of flagging while others could claim kills become more valuable. The truth is, it encourages you to hold on to your flag, but it does not
give a reward UNLESS you kill someone, hence I would state the bonus makes kills worth more. If you don't kill anyone you get 0 bonus
points, meaning holding on to the flags alone gives you nothing. But if you have 0 flags, and kill someone, you still get 150 points.
Thus, I would say it increases the value of kills in the game if anything. Importance of flagging, yes, but there is no reward unless you
KILL. At the same time, there is no bonus if you have no flags. In other words, no matter how big the kill bonus is, it is irrelevant in our
calculations as it automatically makes both flags and kills equally important, hence it is a neutral factor.

The fixed kill points should account to roughly as many points as bell flags in a prac.

This all said, I cannot agree with your calculations.

I'll edit your scoring to what it should look like.

Team 1:
70 kills = 4200 points (at 60 points per kill) + (possible flag killbonus).
12 flags at each ding = 360 points for the team at each ding = 5400 points (assuming 15 dings per 20 minute game).
Total = 9,600 points + (possible bonus)

Team 2:
80 kills = 4800 points (at 60 points per kill) + (possible flag killbonus).
8 flags at each ding = 240 points for the team at each ding = 3600 points (assuming 15 dings per 20 minute game).
Total = 8,400 points + (possible bonus)

If the kill bonus would be 5*flag count, it would roughly mean that:
1/3 points come from fixed kills
1/3 points come from bell flags
1/3 points come from flag/kill bonus

Tembest
Posts: 692
Joined: Sun Dec 18, 2011 10:12 am

Re: Re-Organizing DSB

Post by Tembest »

The problem with lifetime stats is that if we do not do a full reset, it is impossible for players to surpass the best players ratings in the
future. It also means players such as Beastie cannot improve their lifetime rating. Literally impossible, as he cannot play with smurfs
anymore. So, the best ratings can never be beaten, absolutely never, and the best can never improve once we disallow smurfing.

The lifetime statistics give a picture of who's good and who's not, but they have been put together wrongly, hence they matter very
little at the moment. When we change the rules of the game, it is crucial the statistics are reset due to reasons stated above. See
when I played with Beastie and Makron on smurfs (1200) rating on my team, and the opposing team had Sulla, Pater, Infinite, opposing
team had a higher rating but I guess it was obvious from the beginning who was going to win. Playing with smurfs has been the only
option for Beastie for a long time to increase his rating.

This is why I suggest the lifetime statistics are reseted. The whole league will be different, for the first time it will actually be a league...

I don't care if you put those on a website, but as Sever said, perhaps some want to have the chance to see those somewhere.
The !lifetime rankings on the bot, should however be reset.

Edit: So, what needs to be done is:
1. Fix the scoring system before anything else. (Kill value)
2. Reset all statistics and start an actual league/ladder that means something.
3. Make people !register in order to play. IP/ID check.

Edit 2: You don't have to reset pracs played, kills, deaths etc. Those I don't care about. But what you have to do is fully reset
winR and PowR as they are not reliable due to numerous reasons and many can never improve those.
Perhaps it's a good idea to save the pracs played, kills, whatever else there may be. That way statistics won't disappear. Only winR
and PowR are problems here.

Tembest
Posts: 692
Joined: Sun Dec 18, 2011 10:12 am

Re: Re-Organizing DSB

Post by Tembest »

jim the chin wrote:Sorry the edit time on this forum is crap.
I tripled the value to 30 minutes due to complaints from you and Falc.

jim the chin
Posts: 315
Joined: Sat Dec 24, 2011 1:17 pm

Re: Re-Organizing DSB

Post by jim the chin »

Thanks for fixing those things.
Entr0py wrote:Open questions/issues

I'm not sure what exactly you are wanting when you request that the stats be reset. Are you asking that I reset the lifetime ratings, and if so, do I stop there, or also reset lifetime kills/deaths/pracs played/etc?
Yes, I think Temb and me agree that everything should be reset. This means that for the first "league" we do, !lifetimestandings and !standings will give exactly the same printout. They will only begin to differ when the second league starts. I only request that you save the current lifetimestandings/stats/ranks/standings just before deleting them from the bot; dump them in a file; and send them to whoever can put them up on the website for posterity. They mean something to some people, but as Temb said, it's useless adding to them after these changes.
I will say up front that I am against the idea of resetting lifetime stats/ratings. They're called "lifetime" for a reason. They have built up over 4 years to the point where I believe they actually do predict match outcomes fairly well
Once people start playing in the new league, won't it sort itself out again pretty quickly though? Also, if everyone starts at 1000 again, do we really want a piece of code making it harder for some people to increase their rating because the bot has decided they're a better player?
Anything else right now?
Not beyond what I've already said. The smurf-stopper is needed, and Temb and me need to agree on a base value for kill points.
Please delete this account. I want nothing to do with this place any more.

Tembest
Posts: 692
Joined: Sun Dec 18, 2011 10:12 am

Re: Re-Organizing DSB

Post by Tembest »

Just to demonstrate why everything needs to be reset and why current ratings are not truthful:

In the following are 3 of my smurfs I have used in pracs. Starting at (1000) winR and powR. Look how I have performed and then imagine how it has affected
my teammates and the other team as well. This is what smurfing causes.

Code: Select all

DSB-PracBo> Overall lifetime statistics for player: cine
DSB-PracBo> +-----------------------------------------------------------------------------------------------------------------------------+
DSB-PracBo> |Sector|  Ki/De   Ratio   As  EAs TKs |  FTs/BFs  AvgBFs  FE |  TRat Points    Time|  W-L    GP  PL | PowR WinR   C%  Cs  AvgC|
DSB-PracBo> |-----------------------------------------------------------------------------------------------------------------------------|
DSB-PracBo> |  East|21.1/16.4  1.29  6.1  8.2 0.0 | 97.2/78.1   5.46 164%|  2875   3030 1003:51| 44-9    53   0 | 1000 1000   0%   0   0.0|
DSB-PracBo> | North|22.5/15.4  1.46  6.0  5.2 0.0 |138.2/92.0   6.42 161%|  4723   4986  149:23|  8-0     8   0 | 1000 1000   0%   0   0.0|
DSB-PracBo> | South|23.8/16.0  1.48  7.0  3.5 0.0 |142.0/80.8   5.57 139%|  3207   4124   80:00|  4-0     4   0 | 1000 1000   0%   0   0.0|
DSB-PracBo> |  West|15.0/ 8.0  1.88  2.0  2.0 0.0 |109.7/173.7 12.12 202%|  2725   3348   60:00|  3-0     3   0 | 1000 1000   0%   0   0.0|
DSB-PracBo> |=============================================================================================================================|
DSB-PracBo> | Total|21.2/15.9  1.33  6.0  7.3 0.0 |105.3/84.3   5.89 141%|  3104   3338 1293:14| 59-9    68   0 | 1180 1333 100%  31  25.6|
DSB-PracBo> +-----------------------------------------------------------------------------------------------------------------------------+

Code: Select all

DSB-PracBo> Overall lifetime statistics for player: clits
DSB-PracBo> +-----------------------------------------------------------------------------------------------------------------------------+
DSB-PracBo> |Sector|  Ki/De   Ratio   As  EAs TKs |  FTs/BFs  AvgBFs  FE |  TRat Points    Time|  W-L    GP  PL | PowR WinR   C%  Cs  AvgC|
DSB-PracBo> |-----------------------------------------------------------------------------------------------------------------------------|
DSB-PracBo> |  East|19.8/11.9  1.67  5.0 10.6 0.0 | 82.7/80.3   5.60 168%|  3304   3053  247:00| 13-2    15   0 | 1000 1000   0%   0   0.0|
DSB-PracBo> | North|10.0/ 4.0  2.50  2.0  3.0 0.0 | 48.0/48.0   6.00 150%|  2709   2770   10:00|  1-0     1   0 | 1000 1000   0%   0   0.0|
DSB-PracBo> | South|20.5/11.5  1.78  6.5  0.0 0.0 |153.5/116.5  7.77 194%|  4406   4018   40:00|  2-0     2   0 | 1000 1000   0%   0   0.0|
DSB-PracBo> |  West| 0.0/ 0.0  0.00  0.0  0.0 0.0 |  0.0/ 0.0   0.00   0%|     0      0    0:00|  0-0     0   0 | 1000 1000   0%   0   0.0|
DSB-PracBo> |=============================================================================================================================|
DSB-PracBo> | Total|19.9/11.7  1.70  5.2  9.0 0.0 | 92.7/85.7   5.92 142%|  3526   3267  297:00| 16-2    18   0 | 1127 1128   0%   0   0.0|
DSB-PracBo> +-----------------------------------------------------------------------------------------------------------------------------+

Code: Select all

DSB-PracBo> Overall lifetime statistics for player: insecto
DSB-PracBo> +-----------------------------------------------------------------------------------------------------------------------------+
DSB-PracBo> |Sector|  Ki/De   Ratio   As  EAs TKs |  FTs/BFs  AvgBFs  FE |  TRat Points    Time|  W-L    GP  PL | PowR WinR   C%  Cs  AvgC|
DSB-PracBo> |-----------------------------------------------------------------------------------------------------------------------------|
DSB-PracBo> |  East|23.8/15.4  1.54 10.7  0.0 0.0 |124.6/96.2   6.82 205%|  3352   3131  251:10| 12-2    14   0 | 1000 1000   0%   0   0.0|
DSB-PracBo> | North|27.0/25.0  1.08  7.0  0.0 0.0 |147.0/108.0  7.71 193%|  3859   4575   20:00|  1-0     1   0 | 1000 1000   0%   0   0.0|
DSB-PracBo> | South|15.0/ 8.0  1.88  2.0  0.0 0.0 |155.0/95.0   6.33 158%|  3432   3350   20:00|  1-0     1   0 | 1000 1000   0%   0   0.0|
DSB-PracBo> |  West| 0.0/ 0.0  0.00  0.0  0.0 0.0 |  0.0/ 0.0   0.00   0%|     0      0    0:00|  1-0     1   0 | 1000 1000   0%   0   0.0|
DSB-PracBo> |=============================================================================================================================|
DSB-PracBo> | Total|23.4/15.6  1.50  9.9  0.0 0.0 |128.2/96.9   6.85 164%|  3392   3245  291:10| 15-2    17   0 | 1096 1133 100%   2  46.5|
DSB-PracBo> +-----------------------------------------------------------------------------------------------------------------------------+
Perhaps now you understand why everything must be reset... and smurfing stopped. This cannot be called a league otherwise.

That's like a 90-10 win-lose record.. With a rating close to 1000.

jim the chin
Posts: 315
Joined: Sat Dec 24, 2011 1:17 pm

Re: Re-Organizing DSB

Post by jim the chin »

Tembest wrote:One could argue it emphasises the importance
of flagging while others could claim kills become more valuable. The truth is, it encourages you to hold on to your flag, but it does not
give a reward UNLESS you kill someone, hence I would state the bonus makes kills worth more. If you don't kill anyone you get 0 bonus
points, meaning holding on to the flags alone gives you nothing. But if you have 0 flags, and kill someone, you still get 150 points.
Thus, I would say it increases the value of kills in the game if anything. Importance of flagging, yes, but there is no reward unless you
KILL. At the same time, there is no bonus if you have no flags. In other words, no matter how big the kill bonus is, it is irrelevant in our
calculations as it automatically makes both flags and kills equally important, hence it is a neutral factor.
I thought my example made it perfectly clear:

A team with 80 kills but a flag average of 8 receives approximately the same points for those kills as a team with 70 kills and a flag average of 12. The difference was only 200 points. Without the kill bonus, the difference in kill points would have been much, much larger! Therefore the kill bonus rewards flagging, not killing. I don't think I can make it any clearer than that. If you get 120 points for a kill, and 60 extra points for holding 12 flags when you get that kill, then 1/3 of those points come from flags. It's really that simple.
Team 1:
70 kills = 4200 points (at 60 points per kill) + (possible flag killbonus).
12 flags at each ding = 360 points for the team at each ding = 5400 points (assuming 15 dings per 20 minute game).
Total = 9,600 points + (possible bonus)

Team 2:
80 kills = 4800 points (at 60 points per kill) + (possible flag killbonus).
8 flags at each ding = 240 points for the team at each ding = 3600 points (assuming 15 dings per 20 minute game).
Total = 8,400 points + (possible bonus)
Teams usually end up with between 50 and 70 kills per team (after 20 minutes in east). So lets start off by saying that the average is around 60 kills, rather than the 75 you're basing things on. That would require the bass kill value to rise to maintain the equilibrium between flag and kill points that you want. So, 80 would be better. Also consider that there are more kills in east than in any other sector, so perhaps it needs to be even higher than 80.

For the aforementioned reasons, this needs to increase to 120 when there is a killbonus of 5*flagcount for every kill.

When you're saying that 1/3 of points should come from kills, 1/3 from killbonus, and 1/3 from flags, what you're really saying is 2/3 should come from flags because the killbonus only considers flags (5*flagcount); that's all that goes into the calculation. It doesn't matter that it's only awarded when you kill. Kills are inevitable. There will not be a situation when there are no kills in a prac. So if the killbonus is included (which it should be imo) then the base kill value should go up to compensate (from 80 to 120).

The example I gave above made it abundantly clear how the killbonus affects the scores. In your example, Team 2 would lose by another 1000 points for only holding 8 flags while they're killing. In fact, Team 1 would get more points for their 70 kills than Team 2 would get for their 80 kills. That should make it obvious.
Please delete this account. I want nothing to do with this place any more.

User avatar
Entr0py
Posts: 135
Joined: Sat Dec 17, 2011 7:54 pm

Re: Re-Organizing DSB

Post by Entr0py »

I should point out here that lifetime ratings are not a "league". They are only supposed to be a reflection of how good someone is and help with match outcome predictions. There is no "winner" for lifetime ratings. Rather, the focus is supposed to be on the ratings for the current league, for which there is a winner.

Also, there are a few ways to increase your ratings when they are very high. One is to play with players on your team who have low ratings. This would lower the avg. team rating to a point where the teams are more even, and then winning will increase the ratings. Another way is to play AGAINST stacked teams, so that you're not expected to do as well. Then, you're not penalized as much if you lose, and if you do win, you're rewarded with higher ratings.

Those are just ways to increase already high ratings. Smurfing will not help increase the high ratings, because you start over at 1000. I do agree, however, that smurfing should not be allowed in a LEAGUE, because it throws off the ratings and predictions.

The easiest way to do that, is to create a new league, e.g. Indi League 2, and require registration. Problem solved.

Addressing your concern about comparing apples and oranges when it comes to ratings between games of different point values/types, the rating system actually takes that into account. It normalizes and scales the ratings based on the point system into a dimensionless quantity, so that they really are comparable. I can't say much more than that right now because the details are complex and I don't have them in front of me to show to you.

Finally, it makes sense that few people can get a lifetime rating as high as beastie/kemi/etc, because few people are as good as them. Even if we reset everything, nobody would approach those rating levels. People eventually would level out to their current lifetime ratings, only with thousands of other players removed from the list.

Tembest
Posts: 692
Joined: Sun Dec 18, 2011 10:12 am

Re: Re-Organizing DSB

Post by Tembest »

The average kills is probably relatively easy to check. So, if it's closer to what you said, so be it.

I agree on everything else but how you view the kill bonus. You consider it as part of flags, while I consider it neutral.
As the BONUS comes from kills, it directly increases the value of a kill. It does not increase the value of a flag.
It is a bonus that will not work without both of the factors affecting (flags held + kills), it is automatically considered a neutral factor.

That's where we have the problem.

I don't care if 1/3 come from the bonus, it could be less... as long as roughly Total Kills = Total Bell Flags (excluding bonus)

Naturally the Fixed kill value should be different according to number of players and flags in the game. Fly talked about something like this
before. Anyhow, as a quick fix dropping the value to anywhere between 60 and 80 will work.

And as mentioned, I already switched it to 80 more than a year ago, but Eridu changed it back because we had a prac where one team had more kills AND
more bell flags and still lost the game, because of the kill bonus. This is the reason I'm not sure if we need the bonus. You cannot see it in statistics... so
a team with more kills and bf's might lose a prac. It's rather confusing...

Post Reply