wxyz wrote on 01/29/14 at 11:21:48:
Hi Mr Danger

2 of the bottom 16 exactly in the Neo example (Roach and Clbrun)... But Flo Chollet say the same things two week ago and I understand your point of view (also because I'm a mid player...). So I thought about that and improved the system by giving what i call a "Gala match" : that give for the last match (number 15 if the table beyond) to the 8 first people who have not yet encounter a top 8 the possibility to fight against them (the exact rule is to have a gap of 5 or more ranks between the higher opponent and the top player)! So with that new rule (Gala matches) you would have meet Neo instead of Wild to die on the Battelfield!!!

Here are a 3rd example with that rule (http://www.ffsmk.org/forum/index.php?topic=5727.0)


There you see that the rule is benefic to Ben, Anton, Raph, Alex, Lighfoot, Djo, Narnet and Rudmarker

It's also interesting to compare your result in the two example. Considering, you have the level to finish around 16th place. In the first example you finish 20th and on the other hand, in this last example, you finish 12th!

=> In the Neo example you do a poor tournament, losing important matches against your direct opponent (lose against Conor, Flo, and Raph and you also do a tie against Leyla, Freddy and Djo). So the program give you necessarely easier match and in the end you just have meet Clbrun and Roach as top players (+ Neo with the Gala rule).

=> In the last example , you do a very good tournament by winning all your important matches (win against Conor, Leyla, Salim and Ogasawara for example), you do also a tie against Roach. And with that you have a very good reward since you could match-up six of the height stronger players (Neo, Geo, Harold, Sami, Karel and Drew)!

So it's important to say that nothing is written in advance and only your result of the D-day will give you the reward or not to match-up more and more top players! If you win important matches, the program will think that you are a good player and you will play against stronger players!

I know that the system is hard to assimilate but with more than 32 people, if we want to keep a reasonable number of matches (the number of 15 come often) AND a very accurate ranking for the top players  there is not thousands of possibilities

djo wrote on 01/29/14 at 13:06:08:


Roach and Cilbrun were not in the bottom 16 (though you could argue Tif was as he was joint 15/16/17/18th last along with Graf who was 8th last)

djo wrote on 01/29/14 at 13:06:08:


Don't discount the level of practice I'm putting in for this year, I plan on vast improvement over previous years (except for BM, which I'll just drive around aimlessly shooting nowhere near the opponent as usual).

djo wrote on 01/29/14 at 13:06:08:


Not hard to assimilate (but then I do have a maths degree and work as an analyst), in fact to try to understand it better I recreated how I thought it work earlier and ran a few simulations of my own but I guess with slightly different ways of choosing the matches, determining results (I only looked at win, draw, loose rather than the score in the game) and assigning points. I looked at a few different set-ups (though each revolved around 40 players playing 15 games each) just with differing numbers of random matches added in as well as ranked matches and different ways of assigning points (usually aimed at rewarding success against "better" players).

I still think it should be made that the top players do play at least a few games against lower opposition as even in your latest example the top 6 players only play anyone in the bottom 20+ places in the very first 2 games (when the draw is essentially random). To almost completely isolate the top players from nearly the entirety of the "bottom" half of players seems to go so much against what the spirit of cdm has seemed to be about (at least the years I've been going).

If 15 matches is what people are looking at then maybe adding slightly more of a random-esq element would at least give people a little more exposure to players of a differing ability. Say for example at game 9 (by which point the table will start to approach what it will look like at the end) you had a round where players couldn't be drawn against anyone within say 16 ranks of their current placement and repeat that at game 11 and 13. that would give players about 10 matches around their skill level and 5 matches that are fairly random (games 1 and 2 are going to always be luck of the draw, and to a lesser extent game 3 as well). By your example Geo only plays 1 player outside the top 15, by adding in those 3 randomish ties he would play at 4 (well at least 3 anyway as the draw would be altered from round 9). I don't think having the top players play 4 or 5ish games with lower players would be so bad as to bore them to tears (in fact it might offer a nice chance to get a bit of a break from hours of constantly close games) and it would avoid the situation of the tournaments turning into a case of the haves (skill) and the have nots hardly mixing at all.

Being quite shy (yes really) I think that there would be quite a few players that I wouldn't have spoken to at all if not for a forced introduction of having to play a game against them.

Since a lot has been said about this being used to determine the true ranks of players does that mean that there will be no knockout-stages to follow?

Sorry for my semi-illiterate ramblings


@Aron, yeh I only noticed today that I never played against you until 2013 (and then we played in all 3 modes) http://www.ffsmk.org/index.php?task=h2h&idp1=99&idp2=80

djo wrote on 01/29/14 at 15:20:11:


Not really, 2 of the top4 were in each half (Harold/Neo top half, Scoob/Geo bottom half) I guess you were referring to Drew being in the same half as Scoob/Geo but since Harold won (most wins ever) and Neo has finished top4 for the last 7 years in BM I'd argue that pairing to be at least as good as Geo/Scoob if not better. There had to be one side of the draw with 3 of the big 5 and another with only 2, no matter what kind of group matches are played this will always be the case in the knockout rounds.
So I can't figure the reasoning for changing anything based on that?? Or were you thinking or something else?

wxyz wrote on 01/29/14 at 15:49:23:
Of course not ... the main purpose of the SB is clearly to have the more accurate ranking with a lot of players (52 and more) and a short number of matches (15 btw).
And a good system for 52 players will work also for 32 players!

@Firewaster : the system you speak have the same base that the SB and will have the same kind of matches programmed with most of your matches against people of your level. We go here further using the value of your opponent as a tie breaker. But not only a tie breaker cause we combine the two factors and the experience proved that was a very good thing to do to improve the ranking (easy to understand analysing the results of the mid-players zone)

But what is so bad about having the top 3 of a mode on the same side of a draw? It was in your own hands to not be there. If you dont perform on the day, you cant grumble when you get a less than favourable knockout position.

Personally Im in favour of smaller groups, but on the current setup. 4 groups of players split evenly. 2 seeds per group based on past CDM performance. Top 4 qualify for knockouts. But thats just me.

I love it when big names have to faceoff early in competitions ala the FA Cup. Its not like you dont have ample opportunity to win the group. Just live up to your skill level and win all your matches.

I get where youre coming from with your arguement, but I cant reconcile it beyond just you. I agree that in a 4 race/battle round its quite easy to get screwed over vs a lesser opponent. But its also just as easy to have luck on your side and have it happen to your opponent. Be they noob or pro.

You say its an issue of variance. But if that is the case, surely Sami (who is pretty much accepted as the best there is at MR) should have the odd occasion where he slips up and doesnt win his group. (granted that did happen in 2011) but for the most part he always wins his group. How can variance come into that when the best MR player always wins his MR group?

The groups in all modes pretty much all finish as expected on a consistent basis, so I cant see why you dont accept the "play the best on the day" arguement. It seems as if you want an easier ride to the knockouts so as to eliminate the luck factor in SMK (knowing you, every year you have a big complaint about it.) But thats the thing, you cant eliminate luck of items etc from SMK. Its literally part of the game. So to want to have a tourno which solely only ranks on skill level, rather than the game as a whole, surely is against the spirit of the game? Id find it extremely boring to have an already fairly predictable tourno made even more predictable. Which in the case of this new system, will be exactly what happens.

On another note, maybe the reason some sides of the draw are top heavy with "big names" is because they arnt as good as they think they are. Or arnt as good as they are expected to be. Duff is a prime example. He is still unanimously put forward as one of BMs best. Which yes he was, a few years ago. But of late many new people have caught him up. Namely Scoub, and a resurgent Harold.
Infact Duff has lost to Scoub in the knockouts at the last two CDMs.

What im trying to say is just let it playout, instead of trying to engineer the tourno to favour the so called best.

And with a 4 groups system, people from the top 8 play only one other person of the top 8
Completely absurd to said your rank is ok with only one match against another top 8

With a 2 groups system (work only until 32 players in 15 matches...), you play against 3 other top 8! That's here yet more relevant but far far away from the 7 matches vs other top 8 proposed by the SB!!!

Fant0m wrote on 01/30/14 at 02:47:21:

That's what I'm thinking everyday, actually: I know only 2-3 people who feel the absolute need to change the current system, mainly because they went very unhappy with their position in the final bracket. Some others just don't like to play a long groupstage with very few interesting matches of their level, which I can understand a bit more, but I still can't think this is a good reason enough to change everything.

Anyway, unless there are not many people enough OK with this, we can give it a try, CDM has no big economic concerns like international sport tournaments and it won't kill it if we change the groupstage system even if it turns out the majority experimenting it eventually dislikes it. But once again we won't see such a big change if there's not a majority of confirmed participants who clearly ask for it. Especially because this has never been Karel & co. asked the most for at first.

djo wrote on 01/30/14 at 03:09:48:

What I want the most is to make sure an undisputed majority of confirmed players want the system to be changed.

This said if it turns out to be a 45+ attendance we'll have to think about something different, because even if this is not a problem to me, I understand much more how long the groupstage it could be for 90% of us.

KVD wrote on 01/30/14 at 04:22:33:

Well it's also very easy to ask oneself if you would feel the need of a system evolution if your results had been better and if you had not face some specific people too early from time to time. ^^

Fant0m wrote on 01/30/14 at 02:47:21:


You just proved that you don't know much about this, because we actually used another sytem for the group stage in 2012


Fant0m wrote on 01/30/14 at 02:47:21:


Don't worry then: we're trying to make it better.

Without going too much over old ground (i.e. I prefer the old system but am not going to keep stating my reasons why) if a new system is to be implemented then I think it does need slightly more variance added to the current proposal where the top 6 or so players will likely have very little or no exposure to players outside the top 20ish players.

What I would suggest is that we engineer a system more along the lines of keeping a single large group but treating it almost as 4 separate groups (top quarter with best players down to bottom quarter with worst players) but make the match making part allocate matches such that:

The top quarter play 50% of their games against others in the top quarter, 25% against the 2nd top quarter and 12.5% against each of the bottom quarters

The 2nd quarter play 50% of their games against others in the 2nd quarter, 25% against the top quarter and 12.5% against each of the bottom quarters

Each of the bottom 2 quarters play 50% against their own quarter, 25% against the other bottom quarter and 12.5% against each of the top 2 quarters

So if we had 40 players and played 16 matches in the group stage (16 is a mathematically nice number since it divides exactly by 2, 4 and 8 for 12.5%, 25% and 50%) what we would have is

Top 10 players each play 8 of the other 9 players in the top 10, 4 of the players ranked 11-20, 2 of the players ranked 21-30 and 2 of the players ranked 31-40.

The system could easily be adapted to fit varying participation levels (if less than 36 players then reduce the number of games, if more than 40 then just make the quarters bigger or perhaps just make the bottom "quarter" larger than the others.

This would maintain what seems to be the current aim of making more matches of a balanced skill level, whist also maintaining guaranteed exposure to players of all skill levels (something that I personally believe essential to the continued success of cdm, if certain groups of players become very isolated from each other you might find that could cause some to stop coming).

The only issue I see is designing a system that would do this perfectly (due to players moving between quarters during mid-late progress of the stage, but likely to only really affect those on the cusps i.e 10th/11th in the 40 player scenario).

Anyway those are my thoughts and if anyone would like to see what it would look like I could set up an example of how the matchmaking could work, though would have to wait until next week as don't have my work laptop with me and don't really fancy doing in open office calc (I'm off until monday and bought too much food shopping during my lunchbreak today to also be able to carry my laptop home).