Darwinbots Forum
Bots and Simulations => Bot Tavern => Topic started by: bacillus on September 04, 2008, 08:39:42 PM
-
F1 BOT STANDINGS
Last updated 11/10/08
1 - Saber
2 - Fruit_Fliesv0.21
3 - Etch Mk II(F1)(abyaly)-15.03.2008
4 - Occura Amplificis(F1)(bacillus)-25.03.08
5 - TieFighters
6 - Pacifist (5G) (Moonfisher) 19-02-08
7 - Spinnerv1.52
8 - Locust(F1)(Moonfisher)
9 - Ebola (F1Vir) (Moonfisher)
10- Multiply4(F1)(F2)(Peter)19-12-07
11- Excalibur 1.12 (F2)(Light)-06.03.07
12- Callidus(F1)(Shen)-05.04.05
13- Blue on blue (F2)(Jez)-05.08.06
14- Bubbles (F2)(Jez)-21.08.06
15- Reaper (F1)(Googlyeyesultra)-17.07.07
16- Detonator v1.2(F1)(bacillus)-24.04.08
17- Martian Tank 3 (F2)(Martian)-06.07.2007
18- Singula Haloculus v2.1(F1)(bacillus)21.04.08
19- The One (1G)(Shen)-23.04.05
20- Republican Wasp (F1)(Commander Keen)
-
Ok, well. Etch is the best that is for sure. Doesn´t matter what method you use.
That Excalibur would be beaten should be pretty certain too. I beat it with multiply, and after I enthousiatic anounced it beaten the league(I did not really expect this) everyone started testing their bot on mine.
Overall I gues that these league standings are pretty accurate. The whole top uses as normal typical dirty tricks only, maybe I should use that kind of tricks.
Are there any objections for using the bacillus kind of F1-league and replacing the old version with it. Or let just introduce the FB-1 league (Fbacillus-1 league). I reread the original topic just and it tends to have some kind of objections with some.
-
I have no objection with him posting results It's too unwieldy for new users, though, to be the "official" run. Part of the idea of the league is that you automate it to see how a version of your bot does compared to another version. You need a ladder for that to make sense.
But since F1 is presently on hiatus, bac's method fills something of a vacuum, so I guess I shouldn't complain
-
Well, the all-v-all method was initially used to create a fairly represented ladder on which challenges were to be based on, but since I lost progress I decided to continue with the 25-odd I had already done.
-
Note when I say ladder I don't mean just a linear ranking. I mean the idea that in order to advance you have to beat everyone below you at the time you place. Once you get high in the ladder you just have to be able to beat the guy one below you to keep your rank. Or beat the guy one above you to advance. It makes for a more intimate league. And managable with lots of bots in it, in a way that a n^2 FFA just can't be.
-
Perhaps develop ranking like in chess, where eveybody has a point average. I don't know the details, but if a low point beats high point he gains a lot (and the high looses a lot) if the low point looses he looses a little (he was expected to do so). In such a system you don't have to play everyone, you just organise appropriate contests. It allows for incremental updating if a new player appears.
-
I have no objection with him posting results It's too unwieldy for new users, though, to be the "official" run. Part of the idea of the league is that you automate it to see how a version of your bot does compared to another version. You need a ladder for that to make sense.
But since F1 is presently on hiatus, bac's method fills something of a vacuum, so I guess I shouldn't complain
To set the record straight, I don't believe bac or anyone else wants to run this league the same way in the future.
You can agree that is the top of the league there where some weaker bots, that where defended by Excalibur. This league table puts a more serious kind of leaguetable then one where guardian was even in the league.
I would (as far that wasn't clear already) like to propose this as the new league table, and let all new league-challangers be run at the old way.
Perhaps develop ranking like in chess, where eveybody has a point average. I don't know the details, but if a low point beats high point he gains a lot (and the high looses a lot) if the low point looses he looses a little (he was expected to do so). In such a system you don't have to play everyone, you just organise appropriate contests. It allows for incremental updating if a new player appears.
No, that is complicated enough. If you look into it a rating system as with chess or draughts could be complicated to setup in DB, I wouldn't know where to start, to me this sounds as an 'líttle' absurb proposal.
To give an example for the complexity, there where multiple chess/draughts worldtitelskeepers that kept first in the ranglist by not playing (almost)any game for serveral years. One of the best proposals would be the only ever chess-world title holder that came from the USA. He got the worldtitle never payed again, and it took years after that he won it to lose it. Another one would be the current tittle holder of draughts. The dutchman Ton Sybrands didn't really played that many draughts matches the last several years. If believe he has quit now, but it will take time before official ranglists get it.
Just some random useless knowledge I know from thinking sports.
-
I don't see how the Swiss avg. point system will work with this, but you're right, It was very elaborate making the starter table, and no method's perfect eg. a bot may beat Etch but not anything else.
-
Excalibur was under ranked in later versions (I think a bug was introduced that lowered its effectiveness, then it was fixed later). Ideally when a new version is released the current league table is rerun, in league order. So Excalibur would have been the #1 bot. It's still possible to get stuck on a bot that you can't beat, but beat all the ones above it. But I think that's fine, it should be an easy or even semi-hard thing to get #1 in the league. It should be really, really hard. That's the whole point.
-
Excalibur was under ranked in later versions (I think a bug was introduced that lowered its effectiveness, then it was fixed later). Ideally when a new version is released the current league table is rerun, in league order. So Excalibur would have been the #1 bot.
I'd go with the rerunning the league order. I didn't thought about that. That elliminates 'stupid old' bots like guardian and aura too.
About that excalibur could be first. I tend to miss a bot Well a bot that I would possible see as first too.
Where is the first etch? Did you forget him, or did he just blew completely.
It's still possible to get stuck on a bot that you can't beat, but beat all the ones above it. But I think that's fine, it should be an easy or even semi-hard thing to get #1 in the league.
Look up, look below. Do you have different oppinions.
It should be really, really hard. That's the whole point.
Anyway, your point is made. I'm over to your side. He, wait you persuaded me fast.
Bac., how did you run all the bots from the starting gate into your league. You posted the results with all the other starting bots left a few hours after the first ranking without them. That is fast.
-
I mean Excalibur would be #1 of the bots that are currently in the league. Not counting any new bots made in the last 1-2 years or however long it's been since the league was run.
-
I mean Excalibur would be #1 of the bots that are currently in the league. Not counting any new bots made in the last 1-2 years or however long it's been since the league was run.
I wasn't eather, lookup the old league table(here (http://www.darwinbots.com/Forum/index.php?showtopic=8)) and etch is already placed before excalibur.
You where not really specific speaking of F1 so if you meant F2, you are completely right. I give you that.
-
Hmm... I don't remember. I reran one of the leagues a couple weeks back just to make sure it worked and didn't crash. After 24 hours running on my laptop (Guardian is a pain to run, and it dropped WAY down in the rankings, so there were a lot of battles) Excalibur was #1 and my computer had crashed.
-
Reading the original thread might be time-consuming, but will stop a whole lot of discussions. I said that I was only going to run the latest versions of bots, or at least only the best version (for bots which I've already run.)
-
Reading the original thread might be time-consuming, but will stop a whole lot of discussions. I said that I was only going to run the latest versions of bots, or at least only the best version (for bots which I've already run.)
Oh, alright.
and
Bac., how did you run all the bots from the starting gate into your league. You posted the results with all the other starting bots left a few hours after the first ranking without them. That is fast.
-
Actually, when you look through, there's quite a few that can be ignored because they're veggies (correct me if I'm wrong). Besides, I had some time spare
-
Okay, I just updated the league standings inclusive with the newer versions; I won't actually make any new posts, just update the current one.
I think there should be a standard when entering updated bots. For example, Etch MK II could be copied twenty times, each time making a minimal change it, then post them as different versions and take over the league. At the moment I'm going on trust which is good enough, but I thought I should raise the point anyway.
-
It's a valid issue. Especially since a one line tweak might make a bot behave totally different. I'd give these "bill of rights and responsibilities" for leagues:
1. When a new bot is submitted, the author pledges that it represents a "significant" departure from any other bot currently in the league. Honor system here, until the first time we have bad abuse and then we'll try to invent something better. "Significant" is judged by the community and bot author.
2. A bot author can pull a bot of theirs from any league they want for any reason they want
3. A bot in a league already can always challenge the bot above it in the league in case program changes make it stronger than that bot now. Ideally leagues a re-run when new versions of the program are released, so this would happen automatically.
4. If an author updates a bot, it most invoke points #2 by removing the bot from the league and resubmitting it. This way you can't change you bot to get it past a hard bot then change it back.
5. Unless the author objects, the bot will be run in all eligible leagues. Meaning that all F2 bots should get run in F1 as well.
-
Okay, I just updated the league standings inclusive with the newer versions; I won't actually make any new posts, just update the current one.
Alright, and I see excelibur is going to average, did something break in that bot or something.
I think there should be a standard when entering updated bots. For example, Etch MK II could be copied twenty times, each time making a minimal change it, then post them as different versions and take over the league. At the moment I'm going on trust which is good enough, but I thought I should raise the point anyway
Not sure whatever you mean.
It sounds eather like the mutation-bot-league botsareus made up.
Or something like you want bots to be posted multiple times and run them all.
Sorry but I don't like the idea of a one-bot league, beacouse that way you would get.
winner 1
winner 0,9b
winner 0,9a
winner 0,8
winner 0,7
.....
Ok my opinion about those updated bots. Where as a example you can take the fruit flies. There really was a minimal change.
First: You throw the earlier change out. Then the next bot challenge the league from the start.
Second: If you know the bot is going to be updated soon, ignore it. Pretty useless to let a league run then.
-
It's a valid issue. Especially since a one line tweak might make a bot behave totally different. I'd give these "bill of rights and responsibilities" for leagues:
1. When a new bot is submitted, the author pledges that it represents a "significant" departure from any other bot currently in the league. Honor system here, until the first time we have bad abuse and then we'll try to invent something better. "Significant" is judged by the community and bot author.
2. A bot author can pull a bot of theirs from any league they want for any reason they want
3. A bot in a league already can always challenge the bot above it in the league in case program changes make it stronger than that bot now. Ideally leagues a re-run when new versions of the program are released, so this would happen automatically.
4. If an author updates a bot, it most invoke points #2 by removing the bot from the league and resubmitting it. This way you can't change you bot to get it past a hard bot then change it back.
5. Unless the author objects, the bot will be run in all eligible leagues. Meaning that all F2 bots should get run in F1 as well.
This seem pretty straighforward and logic. Isn't his something to put into league-rules.
-
As you can see, I have pulled out Fruit Flies 0.2, a.) because it's virtually identical to 0.1, and lost to it, and b.) because Moonfisher admitted that v0.1 was probably better, and I should pull it out.
-
Actualy I was just trying to point out that v0.2 was better in some ways even though it lost to v0.1....
But I feel a new version should atleast beat the old one, so I'll try and get around to making a slightly more complex v0.3 to knock out v0.1
Either way loosing to v0.1 shows some weakness in the new design which I should look into weather I like it or not...
-
It looks like the population control system from version 0.21 is more stable that in previous versions. Any other changes?
-
I put in the F2 league. As you can see, apart from This'nThat sneaking into last place, nothing much happened. Also, I'm thinking I should rerun the League, as I noticed that Occura Amplificis got broken and can't even beat This'nThat (not sure about other ones though)
-
bac, would you post a copy of your F1 league table and the bots in it?
-
I'm not sure why the league table has 21 bots in it. Getting the table up to the default 30 slots would be another reason to rerun...
#F1
1 - Fruit_Fliesv0.21
2 - Etch Mk II(F1)(abyaly)-15.03.2008
3 - Occura Amplificis(F1)(bacillus)-25.03.08
4 - TieFighters
5 - Pacifist (5G) (Moonfisher) 19-02-08
6 - Spinnerv1.52
7 - Locust(F1)(Moonfisher)
8 - Ebola (F1Vir) (Moonfisher)
9 - Multiply4(F1)(F2)(Peter)19-12-07
10- Excalibur 1.12 (F2)(Light)-06.03.07
11- Callidus(F1)(Shen)-05.04.05
12- Blue on blue (F2)(Jez)-05.08.06
13- Bubbles (F2)(Jez)-21.08.06
14- Reaper (F1)(Googlyeyesultra)-17.07.07
15- Detonator v1.2(F1)(bacillus)-24.04.08
16- Martian Tank 3 (F2)(Martian)-06.07.2007
17- Singula Haloculus v2.1(F1)(bacillus)21.04.08
18- The One (1G)(Shen)-23.04.05
19- Republican Wasp (F1)(Commander Keen)
20- ThisnThat1.01(F2)(Peksa)12.05.08
-
New standings as of 3/11/08:
#F1
1 - Quickdraw(F1)(Moonfisher)
2 - Saber
3 - Fruit_Fliesv0.21
4 - Etch Mk II(F1)(abyaly)-15.03.2008
5 - Occura Amplificis(F1)(bacillus)-25.03.08
6 - TieFighters
7 - Pacifist (5G) (Moonfisher) 19-02-08
8 - Spinnerv1.52
9 - Locust(F1)(Moonfisher)
10 - Ebola (F1Vir) (Moonfisher)
11- Multiply4(F1)(F2)(Peter)19-12-07
12- Excalibur 1.12 (F2)(Light)-06.03.07
13- Callidus(F1)(Shen)-05.04.05
14- Blue on blue (F2)(Jez)-05.08.06
15- Bubbles (F2)(Jez)-21.08.06
16- Reaper (F1)(Googlyeyesultra)-17.07.07
17- Detonator v1.2(F1)(bacillus)-24.04.08
18- Martian Tank 3 (F2)(Martian)-06.07.2007
19- Singula Haloculus v2.1(F1)(bacillus)21.04.08
20- The One (1G)(Shen)-23.04.05
-
bac, would you post a copy of your F1 league table and the bots in it?
Could you do it for the new league table too.
F2 hasn't been updated for a while too I see. In the current F2 I've got a F3-bot getting in top10.
-
Yeah, I'll have to update it. Another thing to put on my to-do list