Well, there’s a whole series of problems with that, that has me at a loss for where to begin. I’ll start with the whole Seeding 101 as to why it’s a good idea to begin with. This is based on real examples and experience; I ran tournaments basically the whole time MvC2 has been out when I lived back in Portland, and the manager of the Tilt in downtown Portland considered me the best operator to work with. Just as an FYI.
Say that you’ve got three or four players who are absolutely, unequivocally, stupidly better than everybody else. We’ll say there’s four, and we’ll call them A, B, C, and D. Player A is generally better than B, B is better than C, C is better than D, and you’ve got years upon years of tournament experience to draw upon to come to a very reasonable conclusion that they’re the best four players there.
Now take player E. He’s got a valid claim to being the “best of the rest”, but he can’t beat any of players A through D on his best day, when they’ve got three broken fingers on their button hand and are in a wheelchair with a bad brake so they can’t even trust they won’t go rolling away from the controller. Screw opinions: for the sake of argument here, we will stipulate this as a given fact. A through D don’t correspond to any real names here in the bay area, it’s a pure hypothetical.
Seed randomly. By luck of the draw (the chance is > 0, and something like it happens more often than you’d think), or just because player E is running the tournament so “accidents happen”, players A through D are all on the same half of the bracket, and player E is on the other half. Let’s say he’s playing very well and wins his half, so he effectively skates through the tournament all the way to the winners’ finals without having to play any of A through D until he gets there.
At this point, player E is guaranteed no worse than a third place finish, without having to actually beat any of the top four players. He can press the buttons with his nose from that point forward and get third, even though he knows deep down, and everybody else in the building knows, he cannot beat players A through D. He loses for free to player A in the winners’ finals, loses for free again to player B in the losers’ finals, and places ahead of C or D without actually having to beat them, without having actually proven anything.
This is why you have to have at least some seeding if you want your tournament results to sort out anything like an actual list of who’s better than who. You can easily obtain some sort of random matchups to make one tournament different from the next by randomizing everybody else from there. But if player A is in one quarter, B in another, C in a third, and D in a fourth, it’s evenly spread out, and the bar for anybody to make their way into the top five is even, everyone knows where it is, and everybody has to win the same number of tough matches to hurdle it.
The only real trick to this is identifying A, B, C, and D from the sample space of past tournaments. If all your players are from the same area and have played in tournaments against each other a long time, this is not hard to do at all. I could tell you who they are at least three deep from every place I’ve regularly played tournaments in off the top of my head, and in every one of those places I would have at least two years of sample data to back it up.
You sit here and tell me this isn’t fair to everyone else? If you actually believe that, I’m going to simply tell you that you have no idea what you’re talking about. Even if I guess wrong with this system, the only thing a player E has to do to get into the winners’ semis if he’s truly better than those guys is beat the seeded man in his quarter of the bracket, and failing that, he gets a shot at whoever of them wasn’t good enough to stay out of the losers’ bracket – either way, his final placing in the tournament is in direct proportion to how many of those guys he can beat. If you randomize, that guy might luck his way into the winners’ finals because everybody who can beat him isn’t in the right part of the brackets to stop him, or he might crap out early because he has to play all of them in succession. Either way, if you randomize seeding you have no intelligent way of knowing how good he really is compared to any of them, because you have no way of knowing whether how far anybody got in the brackets was because they were actually any good or if they just skated through an easy part of them.
If you spread out A through D, he’ll have an even challenge no matter where he falls. The tournament will sort itself out as to how many of those guys he can beat. They’ll be spread as evenly through the losers’ bracket if they don’t suffer any upsets as they are in the winners’ bracket, so any dark horses who want to make a name for themselves will do so in direct proportion to how many of them he can beat, and he’ll place in the tournament based on that, and that alone. That’s what good seeding does for you. As long as you can at least take an intelligent and educated guess as to who the four best players from a given area are, the tournament bracket will sift itself out such that the last four to six players left will always be the best four to six there who are playing the best on a given day – whether that’s the four seeded players plus two others, or somebody who was able to beat them. Whether the seeded players live up to expectations, as long as there aren’t too many upsets (i.e. your guesses turned out to be Tweedle, Dee, Dumb, and Dumber) then everyone who is still alive at that point will deserve to be there.
Yes, it requires an educated guess. As long as the guess isn’t too far off, the system will work. We have enough sample data from Fairfield tournaments that we know that Chunk, Crizzle, and Tinh are the most likely top three seeds unless one of them is having a very off day. Who’s #4 could be argued any which way, but that’s an easy one to resolve: if you’ve got two guys you’re not sure of who could belong there, you just conveniently seed them both as 4 and 5, and make them play to demonstrate which is which. (An expanded bracket that goes out to 8 seeds would have 1 vs 8, the winner of which plays the winner of 4 vs 5 in one half of the bracket, with the other half pitting 2 vs 7 and 3 vs 6. Just fill out however many of the top three-to-however-many seeds that you’ve got a clear estimate on, and let the rest go random.)
So I hear you asking, what happens if the people in the tournament are from lots of different areas? Easy. You go to regional seeding. Take the area you’re most familiar with, and see those guys. Then spread the out-of-towners amongst all four corners of the bracket so that they’ve got the broadest chance to prove themselves, probably with a little more randomness there unless you’ve got reliable information about which of them is the best. Once the tournament shakes itself out, you’ll have your answers based on who played the best.
Randomness does not improve tournament results or make them more reliable at all. In fact, no double elimination tournament that’s run randomly is reliable at all in determining anybody below the top two. One guy will win it all, one of them will beat everybody but the top guy. As the A-D with inferior E example above demonstrates, you have absolutely no intelligent information below second place; third place on down is utterly meaningless. It can be because somebody actually did good, and it can be because they completely lucked out. If you want your tournament to avoid being unfair to people, randomness is about the worst thing you can do short of deliberately loading up unbalanced brackets to benefit your buddies.
Yeah, your educated guesses about who should be seeded might not be perfect. Somebody may outplay somebody. But at least the error happens because somebody outplayed somebody, instead of somebody having to outplay three or four somebodies just to get out of a jam because the brackets screwed him sideways with a chainsaw.