Re-Organizing DSB

[falconeer]

If I get 25-30 rec. Despite the neg**** Correction

FIX THE FUCKING EDIT TIME ON THESE FORUMS. 10 minutes to edit. On other forums players can delete their own posts. WTF IS THIS NAZI SHIT.

[Entr0py]

The bot was designed to keep lifetime winR persistent across all leagues, while also having independent winR/powR that resets whenever we start a new league. This is to get accurate win% predictions based on thousands of matches. The only reason you see the same ratings is because it's configured to show the lifetime winR instead of the league's winR.

I'll see if I can play around with it to show the current league's rating instead of the lifetime winR.

[jim the chin]

Just strip away the lifetime ratings and put them on a separate command (like !lifetime), or only put them on the website for people who want to reminisce. Getting two sets of ratings when you do !rank or !stats is damned confusing, and the bot has been going so long now that the lifetime ratings are meaningless; almost none of those people play anymore.

By the way, the WINR ratings aren't even ordered properly. Like I said on the other thread, just do !top10 and see the messed up order.

[Entr0py]

Just strip away the lifetime ratings and put them on a separate command (like !lifetime), or only put them on the website for people who want to reminisce. Getting two sets of ratings when you do !rank or !stats is damned confusing, and the bot has been going so long now that the lifetime ratings are meaningless; almost none of those people play anymore.

By the way, the WINR ratings aren't even ordered properly. Like I said on the other thread, just do !top10 and see the messed up order.

Four things complete:

1) I ran a script to clean up the messed up ratings. It initialized everyone at 1000 winR/powR, and then went through the entire match/prac history one match at a time, and re-computed the winR/powR/carries for lifetime stats as well as all leagues. It then synched the recomputed ratings with the website.

2) In the process of (1), I was able to discover *a/the* root cause of how the ratings got messed up in the first place, and I deployed a patch to fix it.

3) The commands !top and !standings now show the ratings for the league that the bot is currently set to. The only way now to see lifetime winR standings through the bot is to use !lifetimestandings. !top10 shows top 10 lifetime standings - if you think it should instead show the league standings, I can change that.

4) The incorrect sorting of the standings is due to the fact that the sort command needs to be triggered manually. I have instructed Beer and Eridu on how to do this. The standings used to sort automatically at the end of each prac, but as the number of players grew, this became too processor intensive and caused severe lag on the bot, so I moved it to a manual process. I believe Eridu plans to manually sort these every Sunday.

Not sure if that covers everything hoped for - anything else you want/need?

[jim the chin]

So your first two points refer to fixing the -23254636 style ratings, right? Thanks for doing that. There are a few more things though:

3) The commands !top and !standings now show the ratings for the league that the bot is currently set to. The only way now to see lifetime winR standings through the bot is to use !lifetimestandings. !top10 shows top 10 lifetime standings - if you think it should instead show the league standings, I can change that.

1. !top10 should show the current league standings to avoid confusion. If the lifetime standings are available under !lifetimestandings, they don't need to be on the ambiguously named !top10 as well. Please allow the dot preface too (e.g. .lifetimestandings).

2. Lifetime performance should be removed from when you do !rank or !stats. It will confuse people who just want to know their league scores otherwise. Players should be able to do !lifetimestats and !lifetimerank, but if that will take a long time, it might be good just to put the league stats/rank first rather than second in the printout. They need to be properly labeled though. For example, with !stats there is no explanation of which is lifetime and which is league.

4) The incorrect sorting of the standings is due to the fact that the sort command needs to be triggered manually. I have instructed Beer and Eridu on how to do this. The standings used to sort automatically at the end of each prac, but as the number of players grew, this became too processor intensive and caused severe lag on the bot, so I moved it to a manual process. I believe Eridu plans to manually sort these every Sunday.

3. That's good, but they'll also need to be instructed on how to reset the standings too. The idea is to have a ladder-style league that concludes every 6 months with winners declared. If you're not here after 6 months, other people with access will need to do the reset.

4. Obviously, we'd like the league scores to be reset soon so that this can get underway.

5. I saw something weird in the scores just now. Do !rank saiyan and you'll see a winR that is impossibly low for someone with his stats. This may be a sign of a wider issue.

6. AvgC seems to be broken, though I'm not exactly really sure what that is. Maybe just remove if it's too much work. Seems like a similar problem to what the winR's had though.

7. The most important issue may be the hardest. We need to stop smurfs, otherwise good players will smurf to ruin the ratings of their rivals. So there needs to be a !register command of some sort. Perhaps the DB or BABL bot will have most of the needed code.

8. I still think we need a personal !norating command, so anyone who is having a terrible day's praccing will feel comfortable getting in without nuking their rating. Otherwise, we might just scare people away. However, this could be abused to sabotage pracs, and I can't think of an easy way to combat that. Also, Tembest would disagree that !norating is even needed. So, something to think about maybe, but I guess it's not essential for now.

That's all I can think of. Tembest will probably have some useful ideas as well.

[Entr0py]

So your first two points refer to fixing the -23254636 style ratings, right?

Correct

1. !top10 should show the current league standings to avoid confusion. If the lifetime standings are available under !lifetimestandings, they don't need to be on the ambiguously named !top10 as well. Please allow the dot preface too (e.g. .lifetimestandings).

I'll make the change tonight. I can't rememver how to do .lifetimestandings - so I'll defer that in lieu of higher priority items for now.

2. Lifetime performance should be removed from when you do !rank or !stats. It will confuse people who just want to know their league scores otherwise. Players should be able to do !lifetimestats and !lifetimerank, but if that will take a long time, it might be good just to put the league stats/rank first rather than second in the printout. They need to be properly labeled though. For example, with !stats there is no explanation of which is lifetime and which is league.

I'll look into this and see what I can do.

3. That's good, but they'll also need to be instructed on how to reset the standings too. The idea is to have a ladder-style league that concludes every 6 months with winners declared. If you're not here after 6 months, other people with access will need to do the reset.

The standings reset every time a new league is created in the bot. All one needs to do is set up a new league, and !standings/!top will be reset for that new league.

4. Obviously, we'd like the league scores to be reset soon so that this can get underway.

I'm talking with BH/Eridu about this.

5. I saw something weird in the scores just now. Do !rank saiyan and you'll see a winR that is impossibly low for someone with his stats. This may be a sign of a wider issue.

6. AvgC seems to be broken, though I'm not exactly really sure what that is. Maybe just remove if it's too much work. Seems like a similar problem to what the winR's had though.

Yes, I saw that. I planned to look into it first thing when I get home to see what's up.

7. The most important issue may be the hardest. We need to stop smurfs, otherwise good players will smurf to ruin the ratings of their rivals. So there needs to be a !register command of some sort. Perhaps the DB or BABL bot will have most of the needed code.

8. I still think we need a personal !norating command, so anyone who is having a terrible day's praccing will feel comfortable getting in without nuking their rating. Otherwise, we might just scare people away. However, this could be abused to sabotage pracs, and I can't think of an easy way to combat that. Also, Tembest would disagree that !norating is even needed. So, something to think about maybe, but I guess it's not essential for now.

That's all I can think of. Tembest will probably have some useful ideas as well.

7. Leagues can currently be configured through the bot to require registration. So it's not a technical issue, but a question of if we want to require people to register to practice.

8. I don't really care what we do with this. The option is there if we want it, but whether we want it is a discussion for the community.

[Tembest]

Thanks Sever, and Ent for doing all the work.

Sever covered the most important topics but here are a few points:

1. The statistics should be fully reset because as somebody mentioned, none of those players even play anymore, not to mention the
horrible scoring system for pracs with 150 points per kill. Also, with smurfs allowed the scores are not trustworthy in any way. So,
please do a full reset as they currently serve no value. There are many names that cannot be beaten after we disallow smurfs.

2. If we want to call it a league or a ladder, no smurfing can be allowed. I think there should be no debate on this. This is the reason
the current statistics are completely meaningless. It is eg. impossible for vanish to increase his ratings unless he plays with
smurfs. This is also supporting the fact full reset is an absolute necessity.

3. Before ever starting the league or ladder again, fix the goddamn scoring system, and only after this do the full reset.

4. !norating unfortunately cannot be allowed if it's called a league. A league is supposed to be a competitive event, and if you do not
wish to partake in it, you shall play elsewhere. I'm all for having two practice arenas, one with league going on and the other one
without any ratings. I know this will cause people to pretty much play league only, but that just proves there is not enough demand
for an unrated prac arena. The handful of people who wish to play a meaningless game can then go duel or fight in pub.

To sum it up:
1. Full reset. Save nothing.
2. In a league, smurfing cannot be allowed.
3. Fix the scoring system even before the reset.
4. !norating cannot be allowed in a league.

8. I still think we need a personal !norating command, so anyone who is having a terrible day's praccing will feel comfortable getting in without nuking their rating. Otherwise, we might just scare people away. However, this could be abused to sabotage pracs, and I can't think of an easy way to combat that. Also, Tembest would disagree that !norating is even needed. So, something to think about maybe, but I guess it's not essential for now.

Think about what you're saying and you will realise what's wrong here yourself.
You are saying that if you are having a terrible day, you can save your own rating but will nuke others ratings.
If you want to get in half-condition or drunk, you must suffer from the decision just as much as your team mates. I pretty much
never play seriously. I don't care if I win and I also tend to 'emp only' in half of the pracs I play. Now, would it be fair that I'm playing
in a league event like this and have !norating while my team mates will suffer? I mean no harm to anyone, but hell, I just enjoy the
game more when I fool around. So, let me ask you again, do you think I should be allowed to take part in a league event while playing
like that, and having the chance to save my own rating while my team will suffer?

[jim the chin]

To sum it up:
1. Full reset. Save nothing.
2. In a league, smurfing cannot be allowed.
3. Fix the scoring system even before the reset.
4. !norating cannot be allowed in a league.

1. I agree with Temb that lifetime ratings/stats/rank/standings don't need to be updated (added to) anymore, and they don't need to be accessible from the bot anymore. However, I would ask that the lifetime ratings are saved somewhere for people to visit. So I'm talking about a webpage on whatever DSB site we have. People worked hard for those ratings and, this way, people can find them and reminisce.

2. Absolutely agreed.

3. Yes, any more problems with the scoring system need to fixed before doing anything. Are there any more serious issues to be addressed though? I found a couple of small things (saiyan's strangly low rating, and the AvgC stat).

I suppose you want to reduce the base kill value from 150 to something smaller. Bear in mind that the prac bot has a "killbonus" that effectively gives extra flag points, and dilutes the effect of the large base kill value. League doesn't have the killbonus, which is why it's worse there. So I wouldn't reduce the base kill value as much as you've said in recent posts (like to 80). I would suggest that a value of no less than 100 will work for pracs. 120 is probably best. We don't want the game to become completely flagging because that will increase tube whoring and running from pole to pole... which quite frankly, is boring. Also remember that in north, flags are still very important.

4. I suppose I can agree with you on !norating for now. I just think it will hurt the population to not have it. Also, if one person uses !norating, there's nothing stopping others on the team from using it too.

Experience tells me that two arenas is a bad idea (we tried this before in DSB). The league arena won't be the one that's used, so please don't have a "non-league" prac arena too.

[Tembest]

You could also argue the prac bot gives extra kill points rather than extra flag points.
Whichever we use, 120 is far far too high. One kill would account to 4 bell flags in east, and that's a lot, that's not a balance, it's a lot.
It would not change the fact at all that flags would not be worth fighting for at dings. Currently they aren't, in any way.

The value of a kill cannot in any circumstances surpass the value of 3 bell flags if you want it to be a flagging game where tactics
matter the slightest bit. In the 120 you suggested, an east team with 8 flag average could win the game with a 80-70 kill record.
That is no way a balanced situation. See 8:12 compared to 70:80. You lost flags by half. Won kills by what 13%? and won...
My calculations are just quick estimates btw.. nothing else but you see the point.

Isn't it clear those two should be in somewhat relation? Get the average dings, how many are there in a prac? 14? 15?
Calculate how much points would that grant per team if the bell flags were 10-10. 4200 points in a game? right?
Now get the average kills of one team, how much is that? 75 perhaps? Then 4200/75 = There is your kill value (56)

The game is fun only when both elements of the game are equally important. In this care you may see multiple variations in tactics,
etc. as either of focus points could bring you the victory. The best teams master both.

I hold on to my original suggestion one kill should be worth 2-2.5 bell flags, depending which we want to emphasise.
120.... that would change the current gameplay in absolutely no way; flags would remain absolutely meaningless.

Edit: You should remember 150 is the kill value that was designed for 6 vs 6 league, and for the first time everyone praised how
balanced the system for the very first time is. Obviously since we are down to 3 vs 3 games.. the value needs to be roughly cut to half.

[jim the chin]

You could also argue the prac bot gives extra kill points rather than extra flag points.
Whichever we use, 120 is far far too high. One kill would account to 4 bell flags in east, and that's a lot, that's not a balance, it's a lot.
It would not change the fact at all that flags would not be worth fighting for at dings. Currently they aren't, in any way.

If you're protecting a flag-pole, then the killbonus gives you 20 more points for killing an aggressor than the aggressor gets for killing you (killbonus is 5*flagcount, so you get +20 for the 4 flags). In a way, this consideration is doubled to 40 points, because if the other team had the four flags, they'd be getting +20. So the killbonus gives an extra incentive for protecting flags, and it's how I got a figure of 120. Your calculation of 80; plus 40 for protecting a pole gives 120. So I think that's a fair compromise. I would be willing to try 100, but anything less than that will surely be too low. We really don't want a flag-whoring or tube-whoring type of game.

The value of a kill cannot in any circumstances surpass the value of 3 bell flags if you want it to be a flagging game where tactics
matter the slightest bit. In the 120 you suggested, an east team with 8 flag average could win the game with a 80-70 kill record.
That is no way a balanced situation. See 8:12 compared to 70:80. You lost flags by half. Won kills by what 13%? and won...
My calculations are just quick estimates btw.. nothing else but you see the point.

I don't think you're considering the killbonus though. With an average of 12 flags, those 70 kills are worth more. They get a 60 killbonus, while the team with 80 kills only gets 40.

Team 1:
70 kills = 8400 points (at 120 points per kill) + 4200 points (for a 12 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 = 18,000 points

Team 2:
80 kills = 9600 points (at 120 points per kill) + 3200 points (for a 12 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 = 16,400 points

Team 1 wins, and by a suitable amount too I think.

The killbonus almost completely cancels out the benefit of having 10 more kills. Team 2's 80 kills are only worth 200 more points than Team 1's 70 kills, and that's because Team 1 dominated the flagging.

Please check my calculations, but if I haven't made a mistake, 120 for the base kill value seems ideal. The effect of the killbonus on these totals should also make it clear that this stat values flags, not kills.

For league, where there is no killbonus, I agree with you that 80 or 100 would be better. But the pracbot has a killbonus (which I think we should keep), and that makes the difference.