Proposed leaderboard for alpha 14

pier4r
Skynet
Skynet
Posts: 3188

Proposed leaderboard for alpha 14

Post#1 » 15 Jul 2018, 13:05

GFX posted this on discord

---
Proposition of a new leaderboard format
https://stats.gladiabots.com/?v=1400
inactive players displayed by default
impact of inactivity on displayed score:
* -50 pts per inactive week
* minimum score = current league demotion score(edited)
oh and players inactive for more than a year are not displayed at all

then he added that the ghosts would not be affected by the -50 points.
----
http://www.reddit.com/r/Gladiabots/wiki/players/pier4r_nvidia_shield_k1 -> Gladiabots CHAT, stats, insights and more ;

pier4r
Skynet
Skynet
Posts: 3188

Re: Proposed leaderboard for alpha 14

Post#2 » 15 Jul 2018, 14:01

My view.

First I would like to know what is the objective behind the new leaderboard, otherwise we may discuss about pointless stuff.
For example is the new system meant to take away points from the system?

- inactive players displayed by default
I disagree with this, I would like to see how I am doing against people that played recently, not being in a league filled by people that do not play.

At the same time, the "all of fame" (or inactive players) leaderboard that exists now is somehow informative about how people fared in the relatively recent past.

- 50 pts per inactive week plus 'minimum score = current league demotion score'
I disagree against this too. As in my opinion it doesn't incentive more activity as players that play every now and then but they are strong they will have to climb every time back to their right scores, skewing things a lot or even frustrating people.

If the "-50 points a week" is a system to lower the inflation, inflation meant as "points produced out of thin air", then the main culprit are the ghosts. (Really no one saw this when ghost were introduced? It was so obvious I mentioned it quite early viewtopic.php?f=7&t=60&start=340#p15448 1 March 2018)

At times I am baffled. Sometimes some pretty nice ideas are introduced, like the ghosts. Sometimes ideas are proposed that I think "do people even test their ideas checking the evolution of a system for a couple of times or they just say what they think without any analysis?". I mean one can do it of course, but it is producing a negative effect on his reputation in the topic. Anyway I digress, back to the topic.

Now presuming that the "-50 points a week" is a system to lower the inflation I would first test a possible solution slowly instead of trying to fix everything at once, or cramming a solution, normally produces poor results. Especially for a dynamic system like the scores.

If we want to take points introduced by ghosts, the ghosts themselves should be affected as well. Otherwise there will be often plenty of ghosts available with a lot of points (for a while at least, until those become all old and farmed), like the mcompany pimped ghosts around 2300.

Second point is that I am not thinking that a score can be skewed with too much decay. As when a player goes down his real score (say someone has an ai that is 2100 but the score is decayed down to 1800), one has to grind it back and it is frustrating for the players that are really 1800.

According to the motto "let's try slowly" let's reason a bit.
Thanks to ghost we see that players that didn't play for relatively long have still good ghosts (at least with the current system), see bhead, joni ja pojat, masterplayer, vectorpleximus, zgeneral and others.

Therefore can I say "ah this player didn't play for 4 weeks, he is surely 200 points weaker"? Not really, reality doesn't agree.

So my suggestion is as follows. If we go on the path of removing points for inactivity, we could do:
- "-25" points a week, for a maximum decay of -75 points. (so maximum 3 weeks)
- all the ghosts of that player get the same decay, that won't count against the "deactivate" threshold (so the decay won't count for the 150 points that one ghost needs to lose be deactivated)
- the point above may happen multiple times for the player and his ghosts (imagine a player playing on and off, say a week each month), so the limit lower cap is 150 points less than the original score of the ghost. (The player score doesn't matter, as the cap is -75, then each time he plays he has a new score with decay points set to 0)

This also means that every ghost would be identified by the tuple: ["original score", "actual score", "removed points for inactivity"]. Where "removed points for inactivity" cannot get past 150 points and it is increased at most with "25" points steps and the deactivation threshold should be counted like this.

Code: Select all

if ( ghost.actualScore() + ghost.removedPointsForInactivity() + 150 < ghost.originalScore() ) then
  ghost.deactivate()
end

The tuple is not needed for active players as they get a maximum decay of -75 each time and they can always recover getting better. Instead the ghosts are fixed.

In this way tons of points are removed from the system (especially from inactive players with hundreds/thousands of ghosts), but gradually and without bringing the system in a strong deflation circle. It doesn't sound that smart from "oh, we have strong inflation" to "oh, now we have strong deflation".

Quick short example.

The player PLAYER1 stops playing at a score of 1950. After one week, his score will be 1925, all his ghosts will have -25 points less.His most highest ghost, say GHOST1A that is at 2070 is now dropping at 2045.

Player1 plays again. Then he stops playing again and after one week his last score is 1980. Suddenly drops to 1955. His ghosts loses also -25 points again. So the GHOST1A mentioned before that is now at 2040, drops to 2015, adding up the decay points to -50.

Player1 is not really playing, after 3 weeks it gets the maximum decay of -75. So its score is now 1905 and will stay so until he plays again.
In the meanwhile GHOST1A got also -75 decay points plus the -25 that he got in the cycle before, so it has now -100 decay points. Nonetheless its ai is quite good so he is farming back people, floating around 2000.


The example shows that ghosts of players that have good ai slowly go down and bring down with them the active players that do not have so much better ai. This will happen somehow gracefully. If instead a vectorpleximus is at 1750, having not played for 4 weeks (he was 1900+ before stopping playing, losing 50 points per week), playing a bit with an ai that is really much stronger than 1750 will just produce a strong force (read ghosts) that tends to bring all the active players down.

Without even mentioning the case that if players go down too much, active players that retain their points because they play (see mumpsimus, bockwurst, elvenmonky, milkhunter) will be quickly without opponents, as all the other players that didn't keep going just go too down while the ghosts stay up.
http://www.reddit.com/r/Gladiabots/wiki/players/pier4r_nvidia_shield_k1 -> Gladiabots CHAT, stats, insights and more ;

pier4r
Skynet
Skynet
Posts: 3188

Re: Proposed leaderboard for alpha 14

Post#3 » 15 Jul 2018, 14:23

On a side note. I am not sure whether GFX already noted that milkhunter is already almost without opponents (for his merits *). A not graceful decay done "a la brexit" will leave them alone for good.

* In the elo system if one is better than the others and always wins (or really often wins), one just goes indefinitely up.
http://www.reddit.com/r/Gladiabots/wiki/players/pier4r_nvidia_shield_k1 -> Gladiabots CHAT, stats, insights and more ;

sollniss
Automaton
Automaton
Posts: 164

Re: Proposed leaderboard for alpha 14

Post#4 » 16 Jul 2018, 11:07

Why not count the most recent ghost as player activity? Maybe matching ghosts of inactive players against eachother could also help.

Or you just let them decay and/or disable them.

mumpsimus
Algorithm
Algorithm
Posts: 52

Re: Proposed leaderboard for alpha 14

Post#5 » 16 Jul 2018, 18:49

pier4r wrote:On a side note. I am not sure whether GFX already noted that milkhunter is already almost without opponents (for his merits *). A not graceful decay done "a la brexit" will leave them alone for good.

* In the elo system if one is better than the others and always wins (or really often wins), one just goes indefinitely up.


Not really, because no opponents will mean Milkhunter can't play games, so will be inactive, so will lose the 50 points until someone is in reach again... ;)

eduran
Hello World
Hello World
Posts: 6

Re: Proposed leaderboard for alpha 14

Post#6 » 16 Jul 2018, 21:05

pier4r wrote:- inactive players displayed by default
I disagree with this, I would like to see how I am doing against people that played recently, not being in a league filled by people that do not play.

It's just the default view, i.e. you can change it back to hide inactive players.
Attachments
Untitled.jpg
Untitled.jpg (70.71 KiB) Viewed 296 times

User avatar
GFX47
Dev Team
Posts: 2569

Re: Proposed leaderboard for alpha 14

Post#7 » 16 Jul 2018, 21:15

The primary objective is to give a better idea to new players about the overall activity of other players.
It has nothing to do with a potential fix of the inflation problem. That's why it doesn't impact ghosts.

The one week activity filter is too exclusive right now.
Hiding inactive players is actually a shame because a lot of them have accomplished big things with the game when they were active and should be remembered for that.
Then there's the problem with players achieving a high score and then stopping any activity just to avoid losing points and their top ranking.
That's why there should be a way to "deprecate" a past achievement and only reward the achievers that stay in the competition.

I get why you are upset (I just wish it would be expressed with a bit more tact ^^) and you have some interesting points. I find your solution too complicated though.

After all I feel it could obfuscate too much the real value of players and make the whole system even harder to grasp that it's already.
It comforts me in my opinion that a season system (with score reset to the current league entry score) coupled with a complete leaderboard (all players active during current season) could better solve this plus the inflation problem.

Thanks for your input!

pier4r
Skynet
Skynet
Posts: 3188

Re: Proposed leaderboard for alpha 14

Post#8 » 16 Jul 2018, 23:26

Sorry for the lack of tact. Only when I suppose the direction is X (but then I was wrong) and I see certain proposals I'm really saying to myself "Why?" Multiple times.

Anyway if the objective is to let shine achievements of people not playing , then I still don't understand the -50 per week until the base league score (that from bronze is bottomless). In 4 weeks one is going to drop 200 points that are a lot. Even recovering them will take some 30+ games.

I am not against the idea per se. I find unreasonable the fact to lose even a thousand of points due to inactivity (someone at 2700 that goes down to 1750).

I can understand losing a fixed amount like 75,100,150 but more than 200? Not really.

Plus once again t800 or tct did great things but with the proposed system they are no better than the new promoted player from master league.

To be honest rather than a merciless decay system I would find more reasonable a season reset (plus hall of fame) starting from the minimum score of the league. Say every 4+ weeks.

In that way even if the ghosts keep their points , every reset will take away points like a tax, eventually balancing out.

The reset won't change much in terms of score differences, but better than the decay.
http://www.reddit.com/r/Gladiabots/wiki/players/pier4r_nvidia_shield_k1 -> Gladiabots CHAT, stats, insights and more ;

User avatar
Revenge
Neural Network
Neural Network
Posts: 357
Contact:

Re: Proposed leaderboard for alpha 14

Post#9 » 17 Jul 2018, 18:28

Mfw miojo falls 1.5k points down to the bottom of GM for inactivity. :lol:

In all seriousness though, this will punish GM players way harder than master rank because of elo score discrepancies.

masterplayer
Script
Script
Posts: 45

Re: Proposed leaderboard for alpha 14

Post#10 » 22 Jul 2018, 12:33

pier4r wrote: - inactive players displayed by default
I disagree with this, I would like to see how I am doing against people that played recently, not being in a league filled by people that do not play.

- 50 pts per inactive week plus 'minimum score = current league demotion score'
[...] I am not thinking that a score can be skewed with too much decay. As when a player goes down his real score (say someone has an ai that is 2100 but the score is decayed down to 1800), one has to grind it back and it is frustrating for the players that are really 1800. [...]

+1

GFX47 wrote:The primary objective is to give a better idea to new players about the overall activity of other players.

Why not add the 'games played last week' from the stats page to the leaderboard?

GFX47 wrote:[...] Hiding inactive players is actually a shame because a lot of them have accomplished big things with the game when they were active and should be remembered for that.

We have the stats page.

GFX47 wrote:Then there's the problem with players achieving a high score and then stopping any activity just to avoid losing points and their top ranking. [...]

They won't do this, because they want to climb the ladder and unlock more skins.

Please don't try to force players activity.
:? I apologize for my childish name and my bad English.

ElvenMonk
Automaton
Automaton
Posts: 129

Re: Proposed leaderboard for alpha 14

Post#11 » 22 Jul 2018, 21:40

masterplayer wrote:
GFX47 wrote:Then there's the problem with players achieving a high score and then stopping any activity just to avoid losing points and their top ranking. [...]

They won't do this, because they want to climb the ladder and unlock more skins.

Please don't try to force players activity.


Pudhing player activity is ok, but works better, when targeting casual players. Just give more ephemerial bonuses for any in-game action and people will play.
Problem described above might exist for top players and can be safely ignored.

Return to “Release candidates”

Who is online

Users browsing this forum: No registered users and 0 guests