Recorded best scores.

pier4r
Skynet
Skynet
Posts: 3190

Re: Recorded best scores: After some days of domination, someone went ahead of nullpointer score

Post#61 » 12 Dec 2016, 14:07

Image
http://www.reddit.com/r/Gladiabots/wiki/players/pier4r_nvidia_shield_k1 -> Gladiabots CHAT, stats, insights and more ;

pier4r
Skynet
Skynet
Posts: 3190

Re: Recorded best scores: After some days of domination, someone went ahead of nullpointer score

Post#62 » 13 Dec 2016, 22:53

In this moment, while I lose a lot of points with AIs "3 days old" , the top3 (counting the active players) is:

1. nullpointer, 441/568, 1721 points
2. christian, 805/1262, 1717 points
3. Mgblitz81, 1311/2434 1577 points
Last edited by pier4r on 13 Dec 2016, 23:51, edited 1 time in total.
http://www.reddit.com/r/Gladiabots/wiki/players/pier4r_nvidia_shield_k1 -> Gladiabots CHAT, stats, insights and more ;

User avatar
Ritter Runkel
Neural Network
Neural Network
Posts: 482

Re: Recorded best scores: After some days of domination, someone went ahead of nullpointer score

Post#63 » 13 Dec 2016, 23:15

you earned it

Image

User avatar
NullPointer
Autonomous Entity
Autonomous Entity
Posts: 539

Re: Recorded best scores: After some days of domination, someone went ahead of nullpointer score

Post#64 » 13 Dec 2016, 23:32

Nice, Christian! 5.3 keeps getting better and better

pier4r
Skynet
Skynet
Posts: 3190

Re: Recorded best scores: After some days of domination, someone went ahead of nullpointer score

Post#65 » 13 Dec 2016, 23:37

Ritter Runkel wrote:you earned it

[img]http://www.funusualsuspects.com/uploads/7/4/5/2/7452560/s538191119328260152_p10_i1_w1500.jpeg[.img]


I do not get it :'(

edit
1. Christian 809/1266, 1739
2. Nullpointer 444/571, 1736
3. Mgblitz 1311/2434, 1577
http://www.reddit.com/r/Gladiabots/wiki/players/pier4r_nvidia_shield_k1 -> Gladiabots CHAT, stats, insights and more ;

User avatar
Ritter Runkel
Neural Network
Neural Network
Posts: 482

Re: Recorded best scores: After some days of domination, someone went ahead of nullpointer score

Post#66 » 13 Dec 2016, 23:48

No second meaning. Just congrats.

pier4r
Skynet
Skynet
Posts: 3190

Re: Recorded best scores: After some days of domination, someone went ahead of nullpointer score

Post#67 » 13 Dec 2016, 23:53

I definitely have to redo the resource collection, without it, I'm doomed. Fortunately I collected a lot of examples.
http://www.reddit.com/r/Gladiabots/wiki/players/pier4r_nvidia_shield_k1 -> Gladiabots CHAT, stats, insights and more ;

Christian
Algorithm
Algorithm
Posts: 56

Re: Recorded best scores: After some days of domination, someone went ahead of nullpointer score

Post#68 » 13 Dec 2016, 23:57

Thanks very much. Nothing can beat a nice streak of lucky map assignments. :lol:

Enough for now. I am gonna rest for this night to return rested in the morning and lose about 100 points on meeting point, barred spiral and split team. ;)

MGBlitz81
Automaton
Automaton
Posts: 133

Re: Recorded best scores: After some days of domination, someone went ahead of nullpointer score

Post#69 » 14 Dec 2016, 07:40

I think your split team is pretty good. I think you have beat me every time I played you on that map the last few days. Considering that I don't consider it one of my weakest maps, I'd say you are doing fine.

But then again, you never know what stategies other people are weak against if you don't personally use them.

Christian
Algorithm
Algorithm
Posts: 56

Re: Recorded best scores: After some days of domination, someone went ahead of nullpointer score

Post#70 » 14 Dec 2016, 09:44

MGBlitz81 wrote:I think your split team is pretty good. I think you have beat me every time I played you on that map the last few days. Considering that I don't consider it one of my weakest maps, I'd say you are doing fine.


It has been getting better lately. In 5.2 it was a horrible gamble on my side. Now things improved a bit and I often manage to secure a draw at least.

MGBlitz81 wrote:But then again, you never know what stategies other people are weak against if you don't personally use them.


That is really so. You finetune your ai's to stand strong against players using good, proven ai's and strategies. And then you are caught off guard in a very funny way.

By the way, you just beat me on one of my stronger maps. ;)

User avatar
NullPointer
Autonomous Entity
Autonomous Entity
Posts: 539

Re: Recorded best scores: After some days of domination, someone went ahead of nullpointer score

Post#71 » 14 Dec 2016, 11:22

Christian wrote:And then you are caught off guard in a very funny way.


Story of my life. Don't know if you agree with this, but this game is a lot about collaboration. Some may think it's easier for me to keep a top position because I'm playing less, but it's actually the opposite, this works only for the short term. If I stay away for more than 3 days my AIs get outdated more than one may expect, even though it felt unbeatable for a couple of days.

This, of course, is not a problem, it's actually one of the best things about this game, the constant clash of different ideas leading to AIs no one could've developed on their own when the game started.

pier4r
Skynet
Skynet
Posts: 3190

Re: Recorded best scores: After some days of domination, someone went ahead of nullpointer score

Post#72 » 14 Dec 2016, 11:36

NullPointer wrote:
Christian wrote:And then you are caught off guard in a very funny way.


Story of my life. Don't know if you agree with this, but this game is a lot about collaboration. Some may think it's easier for me to keep a top position because I'm playing less, but it's actually the opposite, this works only for the short term. If I stay away for more than 3 days my AIs get outdated more than one may expect, even though it felt unbeatable for a couple of days.


This is actually is true unless one uses often a testing account, so having always updated AIs. But yes, if I have time I will make a little post about someone with an high score and how quickly s/he can lose it if the winning ratio is not kept high. One could see quickly from here: viewtopic.php?f=7&t=126&p=1253#p1253 that if someone loses as much as s/he wins, with an higher score, the point lost cannot be recovered so easily.
http://www.reddit.com/r/Gladiabots/wiki/players/pier4r_nvidia_shield_k1 -> Gladiabots CHAT, stats, insights and more ;

Christian
Algorithm
Algorithm
Posts: 56

Re: Recorded best scores: After some days of domination, someone went ahead of nullpointer score

Post#73 » 14 Dec 2016, 12:16

NullPointer wrote:Don't know if you agree with this, but this game is a lot about collaboration.


I do agree. Playing many games like I do helps immensely testing and refining you ai's. Considering that there may not always be a perfect ai or strategy you can use many matches to determine "trends" or the meta and set up against it, so that you win more matches than you lose. That is knowledge you only gain by playing many matches. And you will recognize when the meta shifts and can react to it accordingly.

But also the elo mechanism is very forgiving regarding experiments and such. It does not really matter where you start; once your ai's and strategies reach a certain matureness you will rise quickly.

When 5.3 went official I started without any experience regarding the mixture of bots and the new maps (I did not play the release candidates) and started with completely new ai's. I got stomped the first 150 matches or so. But once my ai's got better due to intensive testing in multiplayer, I rose quite quickly back into the top ranks. Given that, I usually test my ai's directly in multiplayer instead of maintaining "enemy" ais to test my ai's. This surely does hurt my w/l-ratio, and I won´t ever be able to reach your ratio but I am fine with the up and downs. :lol:

So, playing many games is certainly no disadvantage.

User avatar
NullPointer
Autonomous Entity
Autonomous Entity
Posts: 539

Re: Recorded best scores: After some days of domination, someone went ahead of nullpointer score

Post#74 » 14 Dec 2016, 12:25

Christian wrote:It does not really matter where you start; once your ai's and strategies reach a certain matureness you will rise quickly.

Yes!

The main reason why my winning ration is high is because in 5.2 I could find only ~5 games per night, so I would play more against myself than against others. And when I could find a match, my AIs were already bullet proofed against more strategies than were actually used against me.

pier4r
Skynet
Skynet
Posts: 3190

Re: Recorded best scores: After some days of domination, someone went ahead of nullpointer score

Post#75 » 14 Dec 2016, 13:03

Christian wrote:But also the elo mechanism is very forgiving regarding experiments and such. It does not really matter where you start; once your ai's and strategies reach a certain matureness you will rise quickly.


Yes and no. If one does not care to go down and then go up, yes, the elo is great (Indeed for me it counts more than the winning rate). Otherwise the elo is forgiving if one loses against someone with 100+ points, one loses maybe 7 or 6. But if you are the one with 100+ points, the elo does not forgive defeats. Yesterday in less than 10 matches (no counter for sneaking sg at the moment) I went from 1600+ to 1460.

At the same time, the day where I had the first good AI in 5.3.1 (and only for that day) I went from 1370 to 1718 in less than 100 matches.

So my point is, unless one uses a testing account to refine AIs without losing points on the main account, keeping the main account high without losing points is not trivial.
http://www.reddit.com/r/Gladiabots/wiki/players/pier4r_nvidia_shield_k1 -> Gladiabots CHAT, stats, insights and more ;

Christian
Algorithm
Algorithm
Posts: 56

Re: Recorded best scores: After some days of domination, someone went ahead of nullpointer score

Post#76 » 14 Dec 2016, 14:04

pier4r wrote:So my point is, unless one uses a testing account to refine AIs without losing points on the main account, keeping the main account high without losing points is not trivial.


I would say it is the other way around. I am way too lazy to maintain a testing account und synchronizing AI's between them by hand. And if I weren´t too lazy, four out of five times I would incorporate some mistakes while transferring the ai's.

So I edit my in-use ai and watch it in multiplayer. When it is bad, you just reverse it and spam matches. Your elo will "repair" itself (unless it was "unjustifiably" high regarding your ai's to begin with; meaning you got a streak of lucky maps and pairings like I did). The whole point in elo is that it really tracks only maybe your last 100 matches or so. Catch some defeats through unoptimized ai's? Optimize your ai and about 100 matches (or more like 50 matches) later there will be no difference in your elo whether you did lose that matches or never played them!

Look at SunwayTaihuLight: Elo 1511 after 115 matches total!
Or SkidRow: Elo 1621 after 145 matches total!

Both propably second accounts of somebody else but that does not matter. They prove that only relatively few matches are needed to get from very low to very high elo. So getting back 50 or 100 points of elo is a matter of few hours, if your ai's and strategies are strong enough to justify that elo. ;)

pier4r
Skynet
Skynet
Posts: 3190

Re: Recorded best scores: After some days of domination, someone went ahead of nullpointer score

Post#77 » 14 Dec 2016, 14:39

Exactly. My point was building the argument (through several posts) that if nullpointer keeps his high elo, although not playing 100 matches per day, it means that his AIs are quite good all the time, otherwise would fall quickly, exactly as you mentioned.

For the point: "The whole point in elo is that it really tracks only maybe your last 100 matches or so" it depends of some other factors, but more or less it tracks the last game performances, yes.
http://www.reddit.com/r/Gladiabots/wiki/players/pier4r_nvidia_shield_k1 -> Gladiabots CHAT, stats, insights and more ;

pier4r
Skynet
Skynet
Posts: 3190

Re: Recorded best scores: After some days of domination, someone went ahead of nullpointer score

Post#78 » 15 Dec 2016, 13:16

Slowly we will have (with a bit of database drilling) a nullpointer number.
https://www.reddit.com/r/chess/comments ... en_number/

How long is the path to reach someone that beaten nullpointer - here the condition - in the last 14 days.
http://www.reddit.com/r/Gladiabots/wiki/players/pier4r_nvidia_shield_k1 -> Gladiabots CHAT, stats, insights and more ;

pier4r
Skynet
Skynet
Posts: 3190

Re: Recorded best scores: After some days of domination, someone went ahead of nullpointer score

Post#79 » 16 Dec 2016, 16:14

Who is bbbbb ? A new player? Pretty effective. I actually would guess that whatever account with 200+ games and not so high win ratio should not be only a testing account.
http://www.reddit.com/r/Gladiabots/wiki/players/pier4r_nvidia_shield_k1 -> Gladiabots CHAT, stats, insights and more ;

ConsciouS-0nE
Autonomous Entity
Autonomous Entity
Posts: 738

Re: Recorded best scores: After some days of domination, someone went ahead of nullpointer score

Post#80 » 16 Dec 2016, 16:38

Shhh. ;)
Lemme "gues"s this kinds of accounts are "inverted test accounts" ;)
Or just pure sub account

Main acount- all ai program failing or winning
Test acounts- testing for the most powerfull programs

"Inverted test" acount- has only the successful ai programs.
Im not the best player, but atleast im something.I'm 8
Science is madness! 3:)

Return to “General Discussions”

Who is online

Users browsing this forum: Erikferus and 2 guests