a12 / a13 matchmaking

User avatar
GFX47
Dev
Dev
Posts: 2914

Re: a12 / a13 matchmaking

Post#21 » 24 May 2018, 18:50

Another possible solution : exclude ghosts that lost more than x (50?) points.

pier4r
Skynet
Skynet
Posts: 3390

Re: a12 / a13 matchmaking

Post#22 » 24 May 2018, 20:30

I wouldn't hurry to set small limits . There are ghosts that in some moments lose a lot of points to then gain many when the playerbase forget to optimum against them. Maybe one can go with streaks against lower rated players . If the ghost loses 5 times in a row to a player with a score 150 points or more lower than the ghost.

Rationale : if the opponent is raising , he will be quickly closer or above the ghost, while farming him. So it wouldn't count . If the ghost is broken, will be farmed by many before settling to a way lower score . Therefore it will likely have a lot of losses by opponents with much lower score .

But it has to be 5 times (or more) in a row as the occasional upset happens .
http://www.reddit.com/r/Gladiabots/wiki/players/pier4r_nvidia_shield_k1 -> Gladiabots CHAT, stats, insights and more ;

User avatar
GFX47
Dev
Dev
Posts: 2914

Re: a12 / a13 matchmaking

Post#23 » 24 May 2018, 22:44

You lost me there ^^

pier4r
Skynet
Skynet
Posts: 3390

Re: a12 / a13 matchmaking

Post#24 » 25 May 2018, 07:42

You piglet. If I get a keyboard...
http://www.reddit.com/r/Gladiabots/wiki/players/pier4r_nvidia_shield_k1 -> Gladiabots CHAT, stats, insights and more ;

pier4r
Skynet
Skynet
Posts: 3390

Re: a12 / a13 matchmaking

Post#25 » 25 May 2018, 08:05

pier4r wrote:...
But it has to be 5 times (or more) in a row as the occasional upset happens .


Hmm ok let's simplify, we don't need to be too precise.
I don't know if it is possible to keep track of the initial score of the ghost in the database.
If it is possible, then disable a ghost if it has lost more than 200 points (surely not a small amount like 50) from the starting score.

A ghost losing 50-100 points is not that impossible if a raising player is playing. That is, a player that has, say, 1800 score points, but his real strength is 2100. The raising player is going to farm various ghosts so some of them may lose a lot of points to this particular player.
Or could be a nemesis. For example my ghost often loses against mumpsimus but not others.

Therefore making a larger range, say, 200 points (or 150 to be really strict), may determine really buggy AIs rather than AIs that are being farmed by their nemesis.

In the case of the good bot ghosts mentioned above, since it is 2100 now and it started from 2470, it would have been disabled when reaching 2269 points or lower.

Edit. On the other side if say that the available ghosts of good bot are in sequence: 2430, 2450, 2470. The ghost with 2470 was chosen as the last deployed, and got disabled once it went under 2270. Then the ghost with 2450 become active. Likely it will be farmed too and it will be disabled once it hits 2250 and so on.

At the end those three gost above may release 600 points together, while keeping only one ghost, as it is the system now, will release only 300 points.
The problem with the actual ghost system - that I find still well done - is that ghosts are starting with a lot of points so if they are farmed, due to buggy AI or outdated AI they inject a lot of points in the system. That should be expected. Moreover even "old" ghosts with possible outdated ai are not that bad too. Vectorpleximus played the last time, I think, in March (60+ days ago) and it is still a nice AI. I was at first proposing that after 10 days a ghost would be disabled, but due to Vectorpleximus and kingsnowman (or Miojo) I took this request back. They are nice AIs that is nice to have around.

All in all I would say that the current ghost system is pretty neat and one should just accept that scores will be pimped (as the points are always available. There is not anymore the phenomena that player quits playing, taking their points away from the system). What is important, and it was always so is the score distance. viewtopic.php?f=7&t=472
http://www.reddit.com/r/Gladiabots/wiki/players/pier4r_nvidia_shield_k1 -> Gladiabots CHAT, stats, insights and more ;

User avatar
GFX47
Dev
Dev
Posts: 2914

Re: a12 / a13 matchmaking

Post#26 » 25 May 2018, 08:59

Let's try 150 starting now.

User avatar
GFX47
Dev
Dev
Posts: 2914

Re: a12 / a13 matchmaking

Post#27 » 26 May 2018, 15:13

Did you notice the difference so far?

pier4r
Skynet
Skynet
Posts: 3390

Re: a12 / a13 matchmaking

Post#28 » 26 May 2018, 15:29

Not yet. I need a bit of data normally (that is, some 2/3k games in grandmaster. I don't follow much the other leagues) so give it time. I guess at the end of the next week we can see it more.

Surely it seems that some bugged ai disappeared but at the same time, as wrote above "On the other side if say that the available ghosts of good bot are in sequence: 2430, 2450, 2470. The ghost with 2470 was chosen as the last deployed, and got disabled once it went under 2270. Then the ghost with 2450 become active. Likely it will be farmed too and it will be disabled once it hits 2250 and so on."
the other next bugged ai is deployed and farmed.

So at the end this may release even more point in the system than keeping one buggy ghost.

Anyway for personal learning/curiosity I am slowly building a little score simulator. Though you need a hp 50g (or an emulator) to run it. I may port it to a desktop language or I may not (I may port directly RPL to the desktop. Since I find it really neat)

https://app.assembla.com/spaces/various ... reSim.urpl
http://www.reddit.com/r/Gladiabots/wiki/players/pier4r_nvidia_shield_k1 -> Gladiabots CHAT, stats, insights and more ;

Return to “Bug Reports”

Who is online

Users browsing this forum: No registered users and 1 guest