gametechmods
Robot Arena => Discussion => Topic started by: System32 on November 15, 2009, 09:53:40 AM
-
Now, we know the classic bracket incident:
Vet's awesome bot loses.
Ends up in loser's bracket.
Loses to SFTW.
FFUU.
Repeat.
To sort this, and show who the REAL winners are, I have created a round robin/leage/non brackety bracket:
(http://filesmelt.com/downloader/zzz1.bmp) (http://filesmelt.com/)
Pros: All bots get to fight the ones they wanted to fight. No byes!
Cons: May make events longer.
It can have odd numbers like 5 bots, it ensures that bots that can do better WILL do better and makes it even more fun!
Would anyone like to test it? Or shall I use NAR AI for an example?
-
A very interesting idea. Its major downside is that it will make events longer.
-
I like it....
Mind if I copy it for Replica wars 2?
-
A good idea, but it'll take MUCH longer.
-
That's a pretty good idea... i had an idea somehow like this long ago as well, where basically each bot in the tournaments fights all of the other bots.
But yeah, it'll be a HELL of a lot longer than a normal tournament... if i am not mistaken, for a 8-bot tournament, it'll make 28 fights, and a 16-bot tournament, 115 fights :O
About "get the ones they WANTED to fight", the WANTED is pretty unfair.
Vet's awesome bot loses.
Ends up in loser's bracket.
Loses to SFTW.
FFUU.
Repeat.
I think i knew this :P
Or shall I use NAR AI for an example?
Do as you wish.
-
Well, I remember I made a rammer killer and ended up fighting HS based bots RRR1. Not the bot they wanted to fight, but to let that BBRTHSJSTJWTFROFLBBQ bot have it's moment of fame doing what it was made to do.
Got the NAR AI bots needed, using matches now.
-
Wait, so now it's a tournament for the NAR bots?......*gets popcorn and pie*
-
No, that is just a test.
-
Oh, so this like every bot fights every bot and the ones with most losses drop out?
That is a nice idea, but makes big tournaments REALLY long. Prehaps with smaller count of entres, like 8.
-
but there isn`t a final needed is it?
cause the "final bots" already fought each other.
-
So they all fight against each other, and then you get a ranking... we just need a formula for the ranking...
There are plenty of good things in this idea !
-Eliminates the random draw of opponents: the champion (the one with the most wins) is clearly the killer of the tournament, and doesn't wins because of lucky matchups.
-Again, not bot loses because of unlucky matchups.
-No elimination. Even the SFTW gets the chance to fight all bots - even if it doesn't wins a single match (that would deserve a special fail award)
The only real drawback i see is the length of the tournament.
EDIT: @Thyrus: No, there is no final. Just rankings, and the bot that scored best is the winner.
-
-No elimination. Even the SFTW gets the chance to fight all bots - even if it doesn't wins a single match (that would deserve a special fail award)
I deserve this!
-
That's the whole idea nary!
Anyway, it can be done with any number of bots, so there is no bye!
-
In theory... you could also expand the concept to rumbles as well if you wanted. As long as the total bots were in multiples of four... you could have sequential rumbles that would assure that all bots competed with each other one way or another. Just a thought. This could cut back on the total number of matches that need to be recorded.
-
I did a system similar to this for Beetleweights for the Starcore packs. Top entries got included in the pack based on their win:loss ratio.
You mention that it would be more matches and be painful as a result.
Something to consider: Do each match-up as a 1 match winner takes all for that match versus the potential 3 matches of a best 2 out of 3 in other tourneys. Since the tourney winner would be based on overall Win:Loss Ratio with a set of Finals matches for Top Ties you could have quite a bit less effort then initial feelings.
Also I would suggest run a bot against all its foes for that "Row" and then do a highlights video of that bot and and show its final Win:Loss Ratio. Done well people would love to have the video of how their bot did against others. Maybe a Wins Video and a Loss Video for that bot so the owner could 'remember' the wins...
Would not need to do a Losers Bracket this way either.
You would also see how each bot does against a wide variety of foes. It can be quite frustrating to have your bot go against its design nemesis early on when you know it would have defeated most other types of bots if given a chance.
In my experience it can be quite a bit faster if you hold 1 bot in (for example the left slot) and have all the foes it faces switched out in the right slot. Recording each. You slide through all the matches of that bot. Statistically each bot would be in left and right an equal amount of times and face each foe from both positions. Facing each foe twice (once in its 'row' and once as the 'opponent' for the other bot's 'row').
I have done this, think about doing it, it has alot to offer for showing off the over all strength of a bot design...
Starcore
-
It sounds cool and all, but it kind of defeats the purpose of a tounament. The reason why events like BBEANS are so interesting is that upsets can happen. I can guarantee that everybody, except Naryar, was thrilled when ITI beat PITA. What's the fun in a tourney if exactly what you think will happen happens?
-
A tourney is to find the best under a set of criteria.
That I know of this still does that and upsets still happen. Since it is still battles and the battles have winners.
Not sure what you are complaining about except that maybe you like to have the lesser bot win a tourney by upset?
Starcore
-
How do you determine ranking if two bots have the same number of wins?
-
It sounds cool and all, but it kind of defeats the purpose of a tounament. The reason why events like BBEANS are so interesting is that upsets can happen. I can guarantee that everybody, except Naryar, was thrilled when ITI beat PITA. What's the fun in a tourney if exactly what you think will happen happens?
I'm not wanting this to become the standard, I'm just throwing in this as an option.
-
How do you determine ranking if two bots have the same number of wins?
Usually that can be determined by what standard tournaments and even my ELBITE tourneys do to break a tie. but I'll use my system as an example. For a tie, there are a series of methods that are undergone:
Method 1:
After the matches of the bots have completed, Their total point scores are added up (even if your bot uses non-damaging weapons like a snow plow, they are still able to score a point or two in some cases) and organized. Theres a record for each individual match score and the cumulative score (all the points the bot scored from each match it won so far is added up).
Now since there is a high probability that no two bots will have the same score from match to match, the first thing is to compare the two scores of the tiying bots. If one scored higher, quite simply, he is the winner.
Method Two:
Now in even more rare cases, the bots even have the same score in that match (which was not even discovered in my no-weapons table-top bouts) We can simply compare the cumulative points of the two bots and mark the bot with the higher points the winner there.
If even the cumultative scores are tied (which chances are, will never happen in reality) Then its best to just redo the match altogether. Unless it is more of a less serious tournament and everyone is up for a game of head and tails.
-
Wouldn't that mean Flippers are always on the losing end?
-
My thought is if there is a Win:Loss tie just have the 2 or more bots battle it out as a finals Tie-breaker.
Jeff
-
I agree.
-
This is a good idea. But it would take an eternity to do. I suggest that when these tournies come to be that more than one person take control in producing the tourney. Split the work up amongst members who have decent AIing ability and previous tourney producing experience. Just a thought...
-
^^ That's a pretty good idea.
-
myself and DK were talking about doing one later on, one thing that you have to be wary of is like Doomed Bots, DK made these nice looking opening sequence's and I had to try and duplicate them myself for my vids which proved to be a bit of an obstetrical
-
This would be very good for an absolute "best robot" type event. However, please note the fact that real robotic combat often has/had hazards to make it a little more interesting, and the upsets and drama and surprises are what make it exciting in the first place. I personally know for a fact that I can't beat ACAMS or Starcore when I enter BBEANS or whatever tourney, its all in the luck of the draw. That, for me, makes it exciting.
-
There still will be the upsets, just that they won't be so devastating.
-
Idea time:
For a 32 Bot Tournament.
8 Small Brackets of Yours, with 4 Bots Each.
Winner from each bracket goes into 8 Bot Final Tourney, where winner is decided under current standard tournament. Isn't this like how the world cup or world baseball classic is done?
-
yes, that is how it is done.
-
But yeah, it'll be a HELL of a lot longer than a normal tournament... if i am not mistaken, for a 8-bot tournament, it'll make 28 fights, and a 16-bot tournament, 115 fights :O
You are mistaken. (8x8)-8=56 fights, and (16x16)-16=240 fights :O
And a 32-bot round robin tournament: (32x32)-32=992 fights. :O:O
Compare to a 32-bot standard double elimination tournament such as BBEANS. 62 fights. 186 with best 2 out of 3 format, which is still far less than 240.
The only way this format would be practical is if:
a) You don't record every fight, just report the results, or only record brief highlights
b) Break up the bots into smaller groups as Somebody suggested
c) You only do tournaments with 10 or less bots
If you had infinite time, though, this would be a very cool format to use.
-
This should be used to settle small arguement, since the size is too massive.
-
You are mistaken. (8x8)-8=56 fights, and (16x16)-16=240 fights :O
You are counting the number of times all bots have to fight... not the number of fights.
(http://upload.wikimedia.org/wikipedia/commons/thumb/7/73/Complete_graph_K8.svg/600px-Complete_graph_K8.svg.png)
8 corners for 8 bots. Each line here is a fight, and i counted, it's 28.
-
If you had infinite time, though, this would be a very cool format to use.
indeed :P
I think this format should be kept in mind for people hosting tournaments that dont make numbers up,
for example RODE could have done this with only the 8 robots rather then ending after 3 matches ?
-
You are counting the number of times all bots have to fight... not the number of fights.
8 corners for 8 bots. Each line here is a fight, and i counted, it's 28.
Gah... OK, my method was right, but I was counting each fight twice as BotX-BotY and BotY-BotX. It's actually half of what I said... 28 for 8 bots, 120 for 16 bots, and 496 for 32 bots. Still a lot. But if you don't do best 2/3, 16 bots would actually be doable.
-
Honestly, if there are 496 fights for a tournament and my bot is near the end of the bracket, im gona lose interest in that tournament.