EMA Tournaments

Now that TACOS has finished the tournament post, I’ll continue with a few thoughts I’ve gathered through playing in EMA tournaments.

EMA’s tournament system is based on total score. You play a number of games – amount depending on the organizers – and the person with the most points at the end of the day(s) win. Before I go on with this post, you’ll probably need some background on me – because a random nobody with no experience running or playing in tournaments is most likely not the best guy ever to listen to.

Be prepared for another wall of text behind the cut.

I’ve played or taken part in a number of different tournaments. From Chess tournaments to Diplomacy (best board game ever? I dare say so.) tournaments to Risk and Settlers of Catan tournaments to Magic: the Gathering tournaments (where I also have a judge certification) to Counterstrike tournaments to soccer and ju-jutsu tournaments. I’ve also, perhaps more importantly, organized tournaments to a great degree as well. Obviously, mahjong tournaments are also part of my experience, but I have to say that I’ve not hosted one – yet.

Now, I’ll say this: I don’t like EMA’s system. I think it has too many flaws to be used in a serious tournament. It leaves too much to chance, and doesn’t reward the proper players. There are also issues concerning record keeping from tournaments, but I’ll leave that for later in the post.

Let’s look at a theoretical situation here:
A good player will win against a bad player – most of the time.
A good player will get more points when playing against a bad player – most of the time.

If these two sentences are correct, then the following is also true:
A good player will have a harder time winning against another good player – most of the time.
A good player will, if he wins against that same player, win with less points than if he had played a bad player – most of the time.

That is to say: An experienced player will get more points by playing inexperienced/new/bad players than he would if he played experienced/veteran/good players. Chance will obviously play a part in this, but we’ll assume that the effect of chance evens out after an arbitrarily chosen number of games (because it does). Indeed, you can even say that chance affects a good player less than it does a bad player, but that’s for another day.

Now, if we agree – or at least lend the thought some merit – that what I’ve just explained makes sense, then the following would also make sense:
A good player facing inexperienced/new/bad players in a tournament will gain more points than a good player facing other players of the same skill.

Let’s summarize:
Imagine 2 tables, table A and B. On table A plays 1 good player (player A1), and three bad players. On table B plays 4 good players. On table A, player A1 is bound to come away from the game with a MASSIVE points difference, simply because he got lucky and scored a lot off points from the bad players. On table B the winner, no matter who it is, will most likely have a really hard time getting any high scoring hands to score from, since all four players are good at not dealing into each others’ hands. Now, after the game is finished, the tables will look like this, despite (or including) uma:
Player A1: A lot of points
Players A2-4: A lot of points minus
Player B1 (finished first on table B): few points plus
Players B2-4: few points plus or minus
(The actual values doesn’t matter at this stage.)

So whereas player A1 played at an easier table, he gets an easier head start. If this should happen for two or more tables, the chances of any of the good players on Table B catching up to player A1 is diminishing by each time they play against good players.

What I’m asking is: is this fair? In my very personal opinion, I don’t think it is. It allows chance to play a lot bigger part of the tournament than it should. Chance is already present in the hands you draw during each game, it shouldn’t have as much to say at a higher “level” of the tournament.

So instead of just complaining – like some are wont to do – I decided to try and come up with a better system.

You wouldn’t believe the amount of tournament systems we have at our disposal: bracket systems, the Buchholz system (it’s used in chess tournaments, I’ll touch on this later), elimination (single-, double-, triple- and so on) tournaments, group tournaments and swiss tournaments and oh dear they are so many.

What I came up with on my own:
First place gets 10 points
Second place gets 7 points
Third place gets 4 points
Fourth place gets 1 point

So even IF player A-1 got +56,000 points and player B-1 only got +15,000 points, they’ll still be tied (so far) with 10 points each, instead of player B-1 having to score, at least, +41,000 points in the second game only to catch up – and that’s before the second game scores themselves are added to the pot. It eliminates a lot of chance – maybe player A-1 got lucky and got a kazoe yakuman (*cough* not thinking of anyone specific), maybe player B-1 was unlucky and only barely (as you can see – 9,000 uma means B-1 only won by 4k tops) won his game. But in this system a first place is a first place, not a “you basically placed second, looking at your scores… sucks to not win eh?” from the second the first game is over.

Now, now, I can hear your objections to this system. I know them all – there’s a chance that, after the tournament is over, two or more players have played a perfect game. Say that the tournament had four games, that would leave all of them at +40 points. How can you solve that, you crazy person?!

Easy. I can hear all you out there complaining that total score SHOULD TOO MATTER, and hear me out. A while back – after returning from Austria where I placed at a respectable 8th place despite playing in a drunken stupor throughout the day – I requested game records from the Hannover tournament (where I placed a respectable 44 out of 50 or something equally silly) to see if I could apply my system to their tournament. Luck had it that the organizers kept very good records of the game; recording the actual results between each game instead of how the Austrian tournament was recorded (end scores of each game). I was sitting on a veritable treasure when it came to applying it. Deconstructing the results was easy enough, so I could apply my +10, +7, +4, +1 system accordingly.

Now, before I reveal the “shocking” result – this is how the EMA result looked:
Player A: 1st, 1st, 1st, 2nd
Player B: 3rd, 1st, 1st, 4th

Which one do you think won the tournament? Hint: it’s not player A.

You know, personally I would say a system like that, that doesn’t reward the better player, to be complete shit actually. Yeah, not the most diplomatic of descriptions, but really? REALLY? Someone plays an almost perfect tournament and doesn’t win? Someone else plays a not so perfect tournament and wins? Seriously, can even the most staunch of defenders defend that system?

Now, since I’m a huge nerd that happen to like numbers I broke the results down so that I could watch positions change between each round, but let’s skip all that unless you guys request it and go for the final positions of the top 8 players. I see no problem in posting their real names, since it’s not the players’ fault that system blows, but I’ll name them player A to player H since it’s easier for me.

Player A: 1st, 2nd, 1st, 1st – placed 2nd in EMA’s system (34 points with drob system®)
Player B: 1st, 3rd, 2nd, 1st – placed 3rd in EMA’s system (31 points)
Player C: 3rd, 1st, 1st, 2nd – placed 4th in EMA’s system (31 points)
Player D: 1st, 2nd, 1st, 3rd – placed 5th in EMA’s system (31 points)
Player E: 2nd, 2nd, 2nd, 1st – placed 8th in EMA’s system (31 points)
Player F: 1st, 1st, 2nd, 3rd – placed 10th in EMA’s system (31 points)
Player G: 4th, 1st, 1st, 1st – placed 15th in EMA’s system (31 points)
Player H: 2nd, 1st, 1st, 4th – placed 1st in EMA’s system (28 points)

A few thoughts are made apparent here: someone who played very well – one fourth place and three first places – placed 15th using EMA’s system? Also, the drob system® need some way to deal with ties – because we have a LOT of those in this tournament (it was expected). The drob system® lends itself very well to adjustments as well – if you change the points difference to 15, 13, 11 and 9 instead the ranks change quite drastically (notably: Player H is 2 points away from place 2-7, instead of 3), especially with the tie-breaker I thought of.

So how do we deal with the ties? Well, I figure that scores should matter in some way. One idea is to use a Buchholz system – let me explain. In order to determine ranks in a Swiss system tournament (which is kind of what we’re using here) where players have the same score you sum up the score of the players’ opponents and thus favor those who have confronted better opponents.

Sounds PERFECT, doesn’t it? That means opponent skill suddenly matters! It is, however, excruciatingly boring to do on a Saturday morning such as this – it’s not hard, it’s just tedious until I’ve written the macro for it, which is tedious to do since I didn’t design the document to begin with. My initial idea (and indeed it’s still my thought) was that it fits Mahjong perfectly (it’s used in Chess, among others), and it does! But, let’s be honest here. It’s EMA we’re talking about – there’s not much chance that they’ll ever bother to do this amount of work, and they’re quite opposed to change as it is anyway.

Anyway, one idea I fiddled with as a tie-breaker was a percentage of the total score. That way how much points you get suddenly matters again, and there’s not a very big chance that two people will have identical scores in mahjong, now is there? Buchholz is still a better system by far, and one day I will sit down and run the numbers for that as well – but not tonight. The problem is that we’re talking quite BIG numbers in mahjong here: with an uma of +9,000/+3,000, we can’t take too much of it, or our carefully constructed system gets obliterated. How does 0.01% sound? It sounds just about right – that means if you end a tournament with +100,000 points, you get +10 points! Could it be more perfect? (Actually, yes – Buchholz.)

So what happens when we plug those in as a tie-breaker? Why, beautiful things happens:
Player A: 41.84 (1st – before 1st)
Player B: 38 (2nd – before tied 2nd)
Player C: 36.94 (3rd – before tied 2nd)
Player D: 36.78 (4th – before tied 2nd)
Player H: 36.75 (5th – before 8th)
Player E: 35.75 (6th – before tied 2nd)
Player F: 34.77 (7th – before tied 2nd)
Player G: 33.19 (8th – before tied 2nd)

Beautiful or what? Changing, as I mentioned earlier, the points to 15, 13, 11 and 9 shifts the positions so that player H places 3rd instead – so no major difference, but one that is interesting to note none the less. How volatile do you want your tournament rankings to be?

Is the system perfect? Hell no! Not even a Swiss (which this is, kind of) tournament with a Buchholz system is perfect. Is it better than EMA’s current system? Yes, I do think so. Personally, that is.

A few other thoughts: Leaving the seeding random is a good thing, I think – it’s even better with Buchholz tied in – because it will let new players play against better players and as such evolve as players and all that community-building stuff.

Now, to return to my comment on tournament and their record keeping: if more tournaments did it the way the Hannover one (Phoenix something) did, it would be absolutely wonderful, because that would mean that you could try this stuff out on more than one tournament and see how it goes. I’m going to contact clubs across Europe in the hopes that they’ll send me tournament results – the chaps in Austria were very pleasant to deal with, but they didn’t write down how people placed in each round, which makes it quite hard (if not impossible) to reconstruct the tournament such as what you’d need for this system. Of note is that this system doesn’t work very well with a small amount of players – 8 is, definitely, the minimum number of players (and let’s be honest, that’s not much of a tournament anyway), but it evens out at around 24 players or thereabouts, which is what most European mahjong tournaments end up at anyway.

When previewing this post someone mentioned that “just add a cutoff to last 4, or final 8/16 if possible, that would be even nicer”; or to explain it in another way: a finals table. Top 4 players, after Buchholz/percentage change, plays each other in one final game – even if they have played each other at some time during the tournament – to determine the ultimate winner. I like the idea, because it’s as TACOS said in the post before this one: it gives you a well-defined goal. You remove the world, the game is the ONLY thing that can decide how you place. That would be awesome in so many ways. But TACOS described it better.