Replace join order in tie-breaker system

[D4 Damager]

Concise description:

• Replace join order from tie-breaker system with a better indicator of skill

Specifics/Details:

• In auto-tournaments at present, the third condition for resolving tie-break situations is "The order in which the player joined the tournament (first player to join wins)"
• This is not a condition that gives an advantage to more skilled players, which you would assume is the whole point of tie-breaker conditions
• A better alternative might be the average number of rounds taken to win in the players won games? But there are plenty of better alternatives.

How this will benefit the site and/or other comments:

• Tournament progression more based on skill

[DoomYoshi]

Yay.

[patrickaa317]

• A better alternative might be the average number of rounds taken to win in the players won games? But there are plenty of better alternatives.

Only on same map, otherwise those doodle assassin players would have a huge advantage over the large game players. My suggestion on tie breaker would be the person that conquered the most territories during the game. This would be a method to incentivize attacking, even if just a little, through out the game.

[DoomYoshi]

Another bad thing about join order is that it is deceptive. If I join a tournament seeded by score, I have no way to verify whether the tournament is indeed won by join order. It is a non-transparent system. It's like a TO assigning victory based on handing out cards from a deck he kept at home. Which was ruled illegal by the TOs, ftr.

[spiesr]

• A better alternative might be the average number of rounds taken to win in the players won games? But there are plenty of better alternatives.

Only on same map, otherwise those doodle assassin players would have a huge advantage over the large game players. My suggestion on tie breaker would be the person that conquered the most territories during the game. This would be a method to incentivize attacking, even if just a little, through out the game.

Doesn't that just create the reverse situation, where those playing on large maps have the advantage?

[DoomYoshi]

With access to the stats, it would be pretty easy to create normalized numbers, especially for 1v1 since 1v1 games are so popular.

[Gilligan]

the whole tiebreaker system could use a revamp. I don't really like "winning in the least number of rounds" either when people are playing on different maps.

A win on hive takes 20 more turns than doodle. Does it devalue the win because you won on a harder map? I don't think so.

[DoomYoshi]

the whole tiebreaker system could use a revamp. I don't really like "winning in the least number of rounds" either when people are playing on different maps.

A win on hive takes 20 more turns than doodle. Does it devalue the win because you won on a harder map? I don't think so.

For that to work, you calculate the normal time taken to conquer the map and then you determine who won with the least standard deviations below the average. Not too tough, assuming the stats can be dug up automatically.

[Metsfanmax]

the whole tiebreaker system could use a revamp. I don't really like "winning in the least number of rounds" either when people are playing on different maps.

A win on hive takes 20 more turns than doodle. Does it devalue the win because you won on a harder map? I don't think so.

For that to work, you calculate the normal time taken to conquer the map and then you determine who won with the least standard deviations below the average. Not too tough, assuming the stats can be dug up automatically.

It would obviously be too computationally intensive to try to calculate those historical stats on the fly. We would have to make a table of the results just once, and compare based on that.

But I don't like it. It is similarly non-transparent.

[DoomYoshi]

If you made the table of results public, anyone could refer to it. Right now, nobody can refer to any chart which lists join order (for seeded by score autotournaments).

[Metsfanmax]

If you made the table of results public, anyone could refer to it. Right now, nobody can refer to any chart which lists join order (for seeded by score autotournaments).

It would require a non-trivial set of calculations to verify that this was working properly, especially for a large tournament. It's not really feasible for that purpose.

[D4 Damager]

the whole tiebreaker system could use a revamp. I don't really like "winning in the least number of rounds" either when people are playing on different maps.

A win on hive takes 20 more turns than doodle. Does it devalue the win because you won on a harder map? I don't think so.

For that to work, you calculate the normal time taken to conquer the map and then you determine who won with the least standard deviations below the average. Not too tough, assuming the stats can be dug up automatically.

It would obviously be too computationally intensive to try to calculate those historical stats on the fly. We would have to make a table of the results just once, and compare based on that.

But I don't like it. It is similarly non-transparent.

Agreed. Also, there is some element of statistical uncertainty on the estimation of standard deviation, depending on completed games sample size. If you had a very new map, you might get weird results. Or no result at all if it is VERY new

I was wondering about simply having the cumulative number of rounds in all won games. I think it wouldn't work though, because even if everyone in the tournament had the same proportion of small maps and big maps, this would favour players that specialise in small maps (ie.. more likely to win on small maps so a smaller total of rounds for the same number of wins). Humph.

[agentcom]

Anyone ever been in a community-run tournament on a random map? How did they do tiebreakers there? I've spent a bit of time thinking about the tiebreakers and auto-tourney format, and also discussed with BW the same (as one of the beta testers). Just like the conversation above demonstrates, there seems to be a downside to every option.

You might be interested though in knowing that BW basically agrees/recognizes that the system has some problems. Re-evaluating the tiebreaker system is at least on his radar, but it sounded to me like he's got other major projects to do before it makes it on to the to-do list.

[Metsfanmax]

Anyone ever been in a community-run tournament on a random map? How did they do tiebreakers there? I've spent a bit of time thinking about the tiebreakers and auto-tourney format, and also discussed with BW the same (as one of the beta testers). Just like the conversation above demonstrates, there seems to be a downside to every option.

You might be interested though in knowing that BW basically agrees/recognizes that the system has some problems. Re-evaluating the tiebreaker system is at least on his radar, but it sounded to me like he's got other major projects to do before it makes it on to the to-do list.

In general the best tiebreakers are based on the dominance of your opponents. That is, did you beat people who went on to do well in the tournament, or did you beat people who went on to do poorly? Presumably the former is a greater accomplishment.

[DoomYoshi]

Anyone ever been in a community-run tournament on a random map? How did they do tiebreakers there? I've spent a bit of time thinking about the tiebreakers and auto-tourney format, and also discussed with BW the same (as one of the beta testers). Just like the conversation above demonstrates, there seems to be a downside to every option.

You might be interested though in knowing that BW basically agrees/recognizes that the system has some problems. Re-evaluating the tiebreaker system is at least on his radar, but it sounded to me like he's got other major projects to do before it makes it on to the to-do list.

Generally, community-run tournaments are only trying to figure out first-place. The problem (and the need for so many tiebreakers) in auto-tourneys is that they are trying to place the top 5 (actually they order the entire tournament pool).

Community run tiebreakers are almost always another random game.

I have suggested, although the thread is now buried in Tournament Discussion, that just having a random tiebreaker would be just as good (i.e. a random pick of the list of players).