For tournaments, each player begins with a score of 10000 and each game updates this score according to the formula +(loser score/winner score)*200 for the winner, minus that for the loser.  Similar to global ranking.  In the event of a tie, this score is used to break ties.

There is one definite problem with this and another problem that is merely a potential problem.

The definite problem is that when the wins occur matters quite a bit.  In the tournament

http://www.wargear.net/tournaments/view/133

CiscoKid and Viper both posted 6-1 records.  It's a round robin tournament, and among the 5 opponents they commonly defeated, Ciscokid raked due to early wins, beating many of them at scores around 10000.  Viper, upon defeating the same 5 people, beat most of them at lower scores than CK did.  For example, Alpha was 1-0 when CK beat him, and 1-3 when Viper beat him.  Alpha finished 1-5, so CK got him at his maximum score.  

The remaining 1-1 in their records occurred as CK beats Hugh, Viper beats CK, Hugh beats Viper.   The second potential problem with the score approach is that the current formula punishes losing to bad players more than rewarding defeating good ones in the following way:  even if the scores were calculated "simultaneously", this 3-way 1-1 favors CK.  This can be "fixed" with a slight tweak of the formula (that would make the scores even), but it calls into question of which performance is better in this 3-way.    

Sports usually sides with using head-to-head as the tiebreak, ie Viper should win according to the NFL.  For round-robin duels, one could use the MLB 1-game playoff approach, but I don't know how you'd deal with a 2-way tie in a round robin 3-player tournament.

I have some suggestions and fixes, none of them fully comprehensive or satisfactory.  For now, I will just say that I think the tiebreak system needs some addressing.

Does this mean, the player who plays fast, often, and wins scores the greatest possible point counts?  One of the advantages of playing slow is the ability to study the history of games won by the leader.  If the scoring balance is set, then all histories in tournament games must be shielded from view until the tournament is over.

Now for personal satisfaction, Viper and I are doing a private head to, just so he and I will know what a tie breaker would reveal.

You'd rather play a high-scoring player later in the tourney (when they are more above 10000) and a low-scoring player earlier (when they are closer to 10000). My point is that the timing shouldn't matter so much. (With normal ranked games, this bias is negligible b/c people keep playing, but tournaments are finite and small differences determine tournament winners.)

I agree with your history comment. I scout sometimes, and yes yes, history blocking is a good option. That is a different conversation though! Good luck on the tiebreak - I prefer that over a scoring method (any college football fans out there?)

One problem with the tie breaker is, how many boards out there are "fair" to the player who sits in the second seat?

If you give me a formula I can use that's fairer that I can implement without losing my sanity I'm all ears Hugh ;)

The only way to do this is to enforce flat values.

Each match you can only win the same value for each round.

Example: 10000 base is 500 always, then there needs to be a means to award kill points for eliminations.

That means the tiebreaker becomes completley ineffectual - a 6-1 player will always end up with the same score as another 6-1 player.

Eliminations could be another tiebreaker but there will still be cases where two players have the same number of eliminations. Maybe turns taken as the final tie-breaker? Least turns taken wins? Not sure that is fair though.

Hugh wrote: You'd rather play a high-scoring player later in the tourney (when they are more above 10000) and a low-scoring player earlier (when they are closer to 10000). My point is that the timing shouldn't matter so much. (With normal ranked games, this bias is negligible b/c people keep playing, but tournaments are finite and small differences determine tournament winners.)

I agree with your history comment. I scout sometimes, and yes yes, history blocking is a good option. That is a different conversation though! Good luck on the tiebreak - I prefer that over a scoring method (any college football fans out there?)

Over time, in a RR format, don't you theoretically play just as many over 1000 as under?  The system should work for that.. Swiss is a different story..

How about Normal Board Ranking considerations (not tournament rankings)?

Speaking of.. I'm a little late to that party.  Can someone explain to me why Standard and Tournament Rankings are kept separate?  The only thing I can think of is that there's the consideration that players play differently in tournament play, perhaps having a tendency to team up on the leader of a tournament?

M, that is not the case. In a RR, one player can get a disproportionate amount of games against players when they were higher rated by being in the right place at the right time. This happened in the tournament mentioned in my original post. One can demonstrate this possibility by hand using small examples as well.

I can only find people referring to the tournament vs normal ranking discussion, but not the actual discussion.  In the "top x move on" format of WF, clearly moving on is a very different skill than winning a game.  As you said, one might imagine it to be in the best interests of non-leaders to gang up on leaders, and so forth... it is, at least potentially, a different beast than normal ranked game play, which I believe was the essence of that side of the argument.

Well I got my hat handed to me , but I am not giving up the trophy  !!!

Cheers,

The Kid

tom wrote: If you give me a formula I can use that's fairer that I can implement without losing my sanity I'm all ears Hugh ;)

For simplicity, I might side with various idiosyncratic sports solutions:  tiebreaker goes to who won the head-to-head (Viper won two head-to-heads, so CK should feel extra shame in accepting the trophy!!) ; or, tiebreaking matches when possible.

I did have a fairer (order-independent) approach to points, similar to CK's flat points.  It is as follows:  All players begin with 10000 as before.  For the sake of time-independence, the first "ranking update" applies the ranking formula to each game played as though everyone had a 10000 ranking.  So, record completely determines the first iteration.  For the next iteration, do the same, where everyone has their "updated ranking" used for the entire iteration.  So, CK/Viper would have a first iteration of 11000, but in the next iteration, their wins against Alpha would be calculated as 11000 vs 9200.  So, strength of opponents can influence the numbers, but it won't matter whether you played Alpha when he was 1-0 versus when he was 1-3 or 1-5.  

My 2nd comment was that even if you did this, the system punishes losing to a lower ranking player more than it rewards beating a higher ranking player.  My fix to this will appear strange:  Instead of 200*(L/W) as a ranking update, use 400*(L / L+W).  The 400 is b/c L+W in the denominator is nearly twice W in the denominator.  Using L+W in the denominator has many remarkable properties, one of which is that a 3-way 1-1 calculated in the "simultaneous" way suggested above, will treat each of the higher ranked players involved the same.

Note that L=ranking of the losing player and W=ranking of the winning player.

To self-criticize: In RR's, this will not distinguish between two players performing 6-1 (my belief is that this is the correct result, hence the need for tiebreaks or other things, like using head-to-head data). In Swiss, the system above should hand the tiebreak in a fair way to the person who played the better opponents in the tournament. In bracketed tournaments, none of this is an issue :)

Hugh wrote:

tom wrote: If you give me a formula I can use that's fairer that I can implement without losing my sanity I'm all ears Hugh ;)

For simplicity, I might side with various idiosyncratic sports solutions:  tiebreaker goes to who won the head-to-head (Viper won two head-to-heads, so CK should feel extra shame in accepting the trophy!!) ; or, tiebreaking matches when possible.

I won fair and square under rules in play.  I am also big enough to offer a private final match, and public discloser of my loss.  Viper and I know who is the better player, but that does not effect the rules for which the tournament was set up and executed.

The moral of the story, play fast and get points early.

Another radical solution: When a tournament ties, split the Win in the Wins column among the tied and give each player a trophy. ie Viper/CK both get 0.5 tournament wins and the trophy icon. This is against all that is good and competitive in the world, but it would work.

Radical indeed! Sounds like a communist plot to me.. Let's all share the trophy and that way no one will have their feelings hurt.

I think this tie-breaking system works fine. I mean, if someone wants to win the tournament he/she should worry about winning as many games as possible, starting from the first rounds to the last. Stated in another way, everyone plays with this system, so he/she knows that winning games in the early rounds increase the chances of winning a tie-break in the end. That's part of the tournament as well.

But I think Hugh is right, partially. Probably what should be introduced in Round Robin tournaments to solve the tie-breaking question (and I'm referring to the tournament Hugh cited and those that end up the same way), is the "direct-match" score (don't know how it's called in english). Basically if two players end up tied at the end of the tournament, the one who won the direct match against the other, wins. If games have more than one player, than the game where they're both in is considered and if this such game doesn't exist or it does but no one of the tied players won it, than use the tournament score as it is now.

About other scores, like "eliminations count" or "turns taken", I don't think it's a good idea. In this game it counts to win in the end, not how you do that, and that's the beauty of using different strategies. I mean, I could start hunting down my opponents one by one from the beginning, or simply wait that they kill each other waiting the right moment to strike. And personally I think that what strategy to use it depends on the map and the opponents I play against, so it's not even something fixed for every player.

That's why I think the actual system works fine. Probably those other scores could be used to measure some others attributes, like "aggressiveness" or things like that, but not to determine the winner of a game.

Cheers! (=

No, strongly disagree. It is whimsical and capricious. I should be careful how I say it - the system rewards you for playing early against those who will go on to have low rankings, because other high scorers won't get as many points against them. However, it is also rewards playing the high scorers later in the tournament when they have high scores.  You can try to reason why this sort of timing matters if you want, or even try to devise strategies around it, but it makes little sense to me as a tiebreaker.   

Again, ranking systems like the one we use work because games continue to be played.  In the long run, such time asymmetries are smoothed out.  Tournaments stop, and that is why we should worry about the time sensitivity of the system.

So here is how you game the system: Set your account so that you play more than one tournament game at a time. Look at the games that you get set up in. Play the noobs as fast as you can and take 1 day and 23 hour turns against top 5 players. It will annoy the top 5 players so much that they will quit the tournament and perhaps WarGear altogether. Then you win!
I definitely think this needs to be fixed. I finished second in a 1v1 tournament where it was awarded correctly as far as the head to head was concerned http://www.wargear.net/tournaments/view/54. But had we played in a different order it could have easily ended up just like the tournament that started this thread.
The problem that I am having is wrapping my head around a good solution.

Hugh wrote: (With normal ranked games, this bias is negligible b/c people keep playing, but tournaments are finite and small differences determine tournament winners.)

I'm not so sure it is really all that different between Tournaments and Ranked games.  I would much rather lose early in CP hunting, when I'm at 1000 score, than when I'm at 1400.  I'd also much rather win against someone with a 1400 score when I'm at 1400 than beat someone with a 900 score. 

For the most part, I would say that Ranked games are finite, we all know that there is a "ceiling" that we strive for, the 20 CPs at 1500 which, for the most part, indicates a finite system (the being passed and losing CPs does create more of an un-finite system, but a lot of maps the top player will be the top player for a while).

So it seems like a very similar system to me...which I'm not sure I have a problem with.

That said, I do think the idea of Ties goes to the winner between heads-up play (and score if no heads-up play) has merit.

Raptor wrote: But had we played in a different order it could have easily ended up just like the tournament that started this thread.

If my Ranked games were played in a different order where I played all my losses on Ranked games first, then all my wins I would have a lot more CPs :p