Double Elimination

Simon Windisch · Post by **Simon Windisch** » Fri Oct 27, 2006 6:12 am

You could throw a fight to get an easier ride, but I think that if you looked at which robots went through the loser's side, and which went through the winners side you'd find that there were equally scary robots on each side - just a theory.

Also, double elimination gives you two lives. As long as you're winning you know you can take a loss and still keep in. As soon as you lose a match you know you've got no lives spare. Would anyone throw away a safety net like that? And anyway, spinners don't always win, or by now every robot would be a spinner?

Simon

peterwaller · Post by **peterwaller** » Fri Oct 27, 2006 9:31 am

Well spotted I have now corrected them.

Andrew_Hibberd · Post by **Andrew_Hibberd** » Fri Oct 27, 2006 11:02 am

from personal experiance if someone looses/throws away a fight then they are not helping themselves. If you have more fights the chances of loosing go up. You could quite easily face a spinner in the lossers bracket.

The double elimination is a good system with the correct numbers, and i agree that leagues may be better. Fencing compertitions run a league system followed by a direct elimination. People are put into the leagues via rankings, however if it was done randomly then it should work.

For instance if you had 18 robots, then you could have 2 pools of 4 and 2 pools of 5, leading to 12+20 fights. Then take the top so many (or even all of them then put them into a direct knock out table. this may lead to byes, however these would be given to the top robots. If the ranking is worked out by having victories/no. of fights . Then equal scores worked out by melees.

peterwaller · Post by **peterwaller** » Thu Dec 07, 2006 2:38 pm

I have now further refined my double elimination charts so they can handle any number from 8 to 64 robots. By having groups with 4, 5, 6, 7, or 8 robots you can handle odd numbers. For 8 to 15 robots you have two groups 16 to 32 four groups and 33 to 64 eight groups.
The chart below shows the number and size of groups for a given number of robots plus it shows the variations in the number of fights to get through the group stage.
http://i109.photobucket.com/albums/n66/ ... psize2.jpg
The structure for each of the group sizes is shown in the charts below.
http://i109.photobucket.com/albums/n66/ ... WGroup.jpg
http://i109.photobucket.com/albums/n66/ ... Group2.jpg
http://i109.photobucket.com/albums/n66/ ... WGroup.jpg
http://i109.photobucket.com/albums/n66/ ... WGroup.jpg
http://i109.photobucket.com/albums/n66/ ... WGroup.jpg
Once the group stages are complete you are left with 4, 8, or 16 robots giving equal numbers of battles. For teams with multiple robots they will be placed in different groups to avoid clashes as long as possible.
I plan to trial this at AWS22 unless anyone can see a major flaw.
Links for 7way group and group size updated as per comments.

Rhys · Post by **Rhys** » Thu Dec 07, 2006 9:03 pm

That is superb!

One thing I would say, is I think it would be better if the 2 group fights were grouped closer. ie. where we have 13 entrants, do a 6 & 7 way instead of a 5 & 8 way. I think that would be a bit fairer, as both 6 & 7 have the same number of fight to qualify, where as 5 & 8 don't. And also the same kind of thing for 4 group. For example, it would be fairer to have 4 x 5 way fights instead of 3 x 4 way fights and an 8 way. It would be unfair on people in the 8 way. They could have to endure 5 fights, where as someone in a 4 way would only have to win 2. Only a suggestion in the idea of fairness. But I may totally be missing something that makes my idea unfesable.

But once again, amazing system.

Post by **BeligerAnt** » Thu Dec 07, 2006 9:32 pm

Peter, there's something wrong with the 7-way group. It's a bit hard to explain but here goes...

If you number the entrants from top to bottom#1 to #7, then the loser of the 1st round fight between #5 & #6 goes into a fight against a robot that has appeared from nowhere!

I agree with Rhys that it seems fairer to keep the groups as close in size as possible. I can't think of any downside to that.

The groups still need some minor tweaking I think.
Currently 2 robots can meet each other in the first round, then again in the third round (losers bracket). The losers from the second round (winners bracket) should swap to the opposite sub-group of the losers bracket. They could still meet in the play-off for second place in the group, but I think that is unavoidable.
There is still the problem of losing in the first round being better than losing in the second round! This might only apply to the 6- and 7-way groups. For the 6-way group, losing in the first round requires 4 fights to get to 2nd place, whereas losing in the second round then requires a total of 5 fights to get to 2nd place. At least the difference is only +/-1 fight now. I don't know if it can get any better.

Thanks for putting the effort in Peter 8)

peterwaller · Post by **peterwaller** » Fri Dec 08, 2006 8:17 pm

Slight slip on the seven group now corrected.
Rhys's idea on keeping the group size as close as possible is a great idea I have now updated the links to the new version.
I am not convinced that having to fight the same robot twice is that much of a problem compared to how difficult it is to draw it on a diagram.

At the moment each chart prints out at A4 with the boxes just the right size for an Avery label. If we can reach agreement on this I will probably print out the charts and laminate them. Then with sets of labels for each robot with the name printed on we can place them on the chart and build up a record of the event. I will have another look to see if any of the differences in the number of battles can be reduced so if any one has any idea's please let me know.

peterwaller · Post by **peterwaller** » Wed Nov 07, 2007 8:50 pm

I know I should get out more but I have again been refining my double elimination charts. I decided that 8 groups was to many and it would be nice to get it all on one chart. So what I have done is produce diagrams for groups of 5,6,7,8,9,10,11 and 12 to cover 20 - 48 enties. I have also produced a large A2 chart of the four group finals with boxes the right size to take the group sheet. The idea is once you know the final number of entries you stick the four appropriate groups sheets onto the a2 chart and end up with one complete chart.
http://i109.photobucket.com/albums/n66/ ... i4gpA2.jpg
http://i109.photobucket.com/albums/n66/ ... /12way.jpg
http://i109.photobucket.com/albums/n66/ ... 5way-1.jpg