PDA

View Full Version : Scheduled Tournaments and Swiss Pairings



Shaqattaq
06-23-2014, 03:58 PM
by Alex Charsky

Scheduled tournaments are coming in this patch and I wanted to communicate our internal roadmap for this feature. I also want to touch on the philosophy with which we approach Swiss pairings.

https://hextcg.com/scheduled-tournaments-and-swiss-pairings/

mach
06-23-2014, 04:37 PM
It seems you're slightly deviating from the standard Swiss algorithm. While I don't quite agree with this (I think giving a first bye to someone in the top 3rd is preferable to giving second bye to someone) this should happen so infrequently that it really doesn't matter.

The last principle could use some clarification, I think. In the case where you can't match people with the same records, you get as close as possible. So if you have 3 3-0s, 2 play each other and 1 plays a 2-1 (not a 1-2 or 0-3 unless things are very weird). I'm sure you are already doing this, but it's not clear in the article.

The most interesting difference is that once the principles are satisfied the pairings are random. I'm not certain of this but IIRC Chess uses tiebreak-based pairings all rounds and Magic uses them in the last round. In other words, if there are 3 3-0s going into the last round, the one with the worst tiebreaks will get paired down. However, it seems you're making that choice randomly. I think that decision could use some explanation.

oncewasblind
06-23-2014, 04:45 PM
Orange wins!!!!

Also, Auction House?

Chark
06-23-2014, 05:31 PM
It seems you're slightly deviating from the standard Swiss algorithm.

What standard are you using? Chess uses 4 different ones (http://www.fide.com/fide/handbook.html?id=170&view=article). Other TCGs adopted Chess's and omitted a few things (color preference for example).



While I don't quite agree with this (I think giving a first bye to someone in the top 3rd is preferable to giving second bye to someone) this should happen so infrequently that it really doesn't matter.

Giving a bye into a T8 for someone at the expense of avoiding a bye for someone who is not in contention seems worse. If you operate under the premise that pairings are a method to determine the best group of players (usually top 8) in the tournament, giving someone a free ride into that group didn't sit well with us.



The last principle could use some clarification, I think. In the case where you can't match people with the same records, you get as close as possible. So if you have 3 3-0s, 2 play each other and 1 plays a 2-1 (not a 1-2 or 0-3 unless things are very weird). I'm sure you are already doing this, but it's not clear in the article.

Yeah, we try for homogenous groups at the top. If that's not possible, we start pairing people down and try to place them into the group right below them. (i.e. with 3x 3-0s we put one of them into the 2-1 group). If that doesn't produce valid pairings for the tournament, we start moving people further down. I agree that it wasn't super clear in the article.



The most interesting difference is that once the principles are satisfied the pairings are random. I'm not certain of this but IIRC Chess uses tiebreak-based pairings all rounds and Magic uses them in the last round. In other words, if there are 3 3-0s going into the last round, the one with the worst tiebreaks will get paired down. However, it seems you're making that choice randomly. I think that decision could use some explanation.

I could be reading the chess algorithm (http://www.fide.com/fide/handbook.html?id=167&view=article) wrong, but to me it seems like they do not care about tiebreakers in groups.

It looks like there are reports of some versions of DCIr doing non-random last-round pairings based off of tiebreakers, but I don't think WER does this any more. Unclear what MODO does. When I was using DCIr extensively (about 10 years ago), I never knew of that feature (and I was fairly proficient with it). This feature was also never documented in any tournament policies that I recall during that time frame. I'll dig around, but to me random seems cleaner not to break more stuff (i.e. if this is a guideline we have to code, we would have to prioritize it somewhere among the 5 guidelines we have).

Chark
06-23-2014, 05:33 PM
Orange wins!!!!

Also, Auction House?

AH is coming, but I didn't really have much to share on the design philosophy. There was a reasonable number of misconceptions for Swiss, so that's why I wrote that up. Also, telling people what to expect with large tournaments seemed like a reasonable idea.

cavench
06-23-2014, 05:39 PM
"Use the links below to tell us your thoughts."

The link is missing at the bottom of the article.

Werlix
06-23-2014, 05:59 PM
Great article! Thanks :)

Are you able to share your tiebreaker method(s)? The most common I've seen is the sum of your opponent's Wins - Losses, then tiebreaker 2 is the sum of your opponent's tiebreaker 1s.

And also will you show tiebreakers on the tournament standings? Tiebreakers can become really important in large-scale tournaments in the late rounds...

Werlix
06-23-2014, 06:01 PM
It looks like there are reports of some versions of DCIr doing non-random last-round pairings based off of tiebreakers, but I don't think WER does this any more. Unclear what MODO does. When I was using DCIr extensively (about 10 years ago), I never knew of that feature (and I was fairly proficient with it). This feature was also never documented in any tournament policies that I recall during that time frame. I'll dig around, but to me random seems cleaner not to break more stuff (i.e. if this is a guideline we have to code, we would have to prioritize it somewhere among the 5 guidelines we have).

I thought that Mantis always used tiebreakers when determining pairings. Eg if 6 people are tied on X-1, they will be paired based on tiebreakers (while also following the guideline of not playing each other more than once)

mach
06-23-2014, 06:05 PM
What standard are you using? Chess uses 4 different ones (http://www.fide.com/fide/handbook.html?id=170&view=article). Other TCGs adopted Chess's and omitted a few things (color preference for example).


Good point. There are things that they have in common though.

The color preference rules are interesting. Have you considered something similar to balance the number of times each player gets to go first? Depending on the format, that can be a significant advantage.



Giving a bye into a T8 for someone at the expense of avoiding a bye for someone who is not in contention seems worse. If you operate under the premise that pairings are a method to determine the best group of players (usually top 8) in the tournament, giving someone a free ride into that group didn't sit well with us.


That would only be true for tournaments with T8 (in other words, extremely competitive tournaments). For other tournaments, I think giving everyone at least rounds-1 games is more important.



I could be reading the chess algorithm (http://www.fide.com/fide/handbook.html?id=167&view=article) wrong, but to me it seems like they do not care about tiebreakers in groups.


That one does not use tiebreakers, but it isn't random either. It uses ratings to rank people within score groups. Other systems (such as the Burstein one explained there) do use tiebreakers. I don't think any of the chess ones use RNG at all.

Chark
06-23-2014, 06:28 PM
The link is missing at the bottom of the article.

Discuss this article in our forums! (http://forums.cryptozoic.com/showthread.php?t=36599&p=385169#post385169) :)

You can also post in the beta forums. I troll all of OP related threads.


Great article! Thanks :)

Are you able to share your tiebreaker method(s)? The most common I've seen is the sum of your opponent's Wins - Losses, then tiebreaker 2 is the sum of your opponent's tiebreaker 1s.

And also will you show tiebreakers on the tournament standings? Tiebreakers can become really important in large-scale tournaments in the late rounds...

Currently finalizing tiebreakers. It'll likely be Opponent's Match Win%, Opponent's Opponent's Match Win%, Game Win% and finally the 4th hidden tiebreaker will likely be order of tournament enrollment.

We'll show tiebreaks in tournaments, since you can't ID in our game.


I thought that Mantis always used tiebreakers when determining pairings. Eg if 6 people are tied on X-1, they will be paired based on tiebreakers (while also following the guideline of not playing each other more than once)

Not that I knew. It's possible Erik stealth added it at some point, but I would feel like I would know.

mach
06-23-2014, 06:34 PM
Currently finalizing tiebreakers. It'll likely be Match Win%, Opponent's Match Win%, Opponent's Opponent's Match Win% and finally the 4th hidden tiebreaker will likely be order of tournament enrollment.


Do you mean game win for the first tiebreaker? Match win seems useless as a tiebreaker since if it's different there isn't a tie in the first place. :)

I do find it interesting that you're not using OMW as the first breaker. I thought that was the standard.

I think it would be better just to use RNG as the final tiebreaker. No point in encouraging people to sign up the second the server goes up just for that 0.0000001% edge.

Chark
06-23-2014, 06:37 PM
Good point. There are things that they have in common though.

The color preference rules are interesting. Have you considered something similar to balance the number of times each player gets to go first? Depending on the format, that can be a significant advantage.


I toyed with adapting the color preference into TCG going first back when I ran WoWTCG Organized Play. Ultimately it never came to fruition because I didn't have someone who could look at 10-year old code and make changes. We were also pretty happy with the pairing algorithm and didn't want to add a new code into it. It's always scary to make changes to something like that and then having a Worlds tournament blow up when you move from constructed to draft pods.
[/quote]



That would only be true for tournaments with T8 (in other words, extremely competitive tournaments). For other tournaments, I think giving everyone at least rounds-1 games is more important.


Fair point. I think the end state is having all Swiss tournaments being large enough where 2/3 of the people dropping is not going to be practical. In this state 8-person Swiss queues are replaced by asynchronous play.



That one does not use tiebreakers, but it isn't random either. It uses ratings to rank people within score groups. Other systems (such as the Burstein one explained there) do use tiebreakers. I don't think any of the chess ones use RNG at all.

Burstein one is interesting. I'll see if this is something we want to spend engineering resources.

Chark
06-23-2014, 06:47 PM
Do you mean game win for the first tiebreaker? Match win seems useless as a tiebreaker since if it's different there isn't a tie in the first place. :)

I do find it interesting that you're not using OMW as the first breaker. I thought that was the standard.


Yeah, I am not sure what I was thinking there. Updated the original post.



I think it would be better just to use RNG as the final tiebreaker. No point in encouraging people to sign up the second the server goes up just for that 0.0000001% edge.

That's exactly the point. There is an observable hesitancy for players to jump into a registration queue if they are first. Nobody ever wants to be the first to register for a tournament for some reason.

You can also observe average times a queue takes to go from zero to 4 and from 4 to 8 and see the difference.

This slightly combats that.

Xenavire
06-23-2014, 06:52 PM
Yeah, I am not sure what I was thinking there. Updated the original post.



That's exactly the point. There is an observable hesitancy for players to jump into a registration queue if they are first. Nobody ever wants to be the first to register for a tournament for some reason.

You can also observe average times a queue takes to go from zero to 4 and from 4 to 8 and see the difference.

This slightly combats that.

I gotta say, this makes me feel a whole lot better. I tend to turn up just as other tournaments fire, and have to wait far too long to even start seeing other players. Any way to speed things up is good news. :D

Chiany
06-23-2014, 08:01 PM
So, even if you have a VIP ticket (which you get by paying a subscription fee), you still have to pay a tournament entry fee?

The_Wine_Gnat
06-23-2014, 08:03 PM
So is the Swiss Draft "rule and 5 guidelines" already in place or will they be after this patch? If they are already in place, I need to dig up my most recent tournament for you to review. There were multiple byes, people playing the same person twice, and a massive difference in records (first hand experience). I'm not playing another one until I hear if the changes are live.

sukebe
06-23-2014, 08:42 PM
Thank you so much for this article :-)

Everything looks great, I am excited to finally get to play some competitive constructed tournaments.

Could you maybe tell us the planned entry fees for the various tournaments (scheduled constructed, scheduled sealed, and VIP) once you are satisfied they work? I do not have a huge amount of spending money so the longer I have to plan the better my budget will look :-)

trapline
06-23-2014, 09:01 PM
keep the tournaments free and I might bother patching the client. The lack of any free ranked pvp is short sighted, while I understand many are happily throwing money into this nickle and dime system, I have serious doubts that HEX will have a large healthy population in the long run.

Problem with only catering to whales in a small pond is when they start to get bored its too late.

AswanJaguar
06-23-2014, 09:41 PM
Could you maybe tell us the planned entry fees for the various tournaments (scheduled constructed, scheduled sealed, and VIP) once you are satisfied they work? I do not have a huge amount of spending money so the longer I have to plan the better my budget will look :-)

Look no further. (http://forums.cryptozoic.com/showthread.php?t=36600&p=385209&viewfull=1#post385209) No mention of VIP costs yet.

Avignon
06-23-2014, 10:01 PM
How often are tournaments going to run while they are free? Every 4 hours until trial is over or will it be limited to a few a day until they sort out stability? The article suggests that it will eventually be each 4 hours but not initially.

cferejohn
06-23-2014, 10:25 PM
I'm nearly certain one of the advertised perks of VIP was "entry to a free VIP tournament once a month" or words to that effect. Has that changed?

AswanJaguar
06-23-2014, 10:30 PM
I'm nearly certain one of the advertised perks of VIP was "entry to a free VIP tournament once a month" or words to that effect. Has that changed?

I believe they said that VIP tournaments will be added once these scheduled tournaments are stable.

*edit* I may have made that up in my head. This is the last info (http://forums.cryptozoic.com/showthread.php?t=35898&p=373442&viewfull=1#post373442) I can find on VIP tournaments.

cferejohn
06-23-2014, 10:34 PM
I believe they said that VIP tournaments will be added once these scheduled tournaments are stable.

*edit* I may have made that up in my head. This is the last info (http://forums.cryptozoic.com/showthread.php?t=35898&p=373817#post373817) I can find on VIP tournaments.

Well the article says "VIP tournaments will require a VIP entry ticket and an entry fee to participate."

Gwaer
06-23-2014, 10:38 PM
From the kickstarter: 540K - Add VIP Program at launch. HEX is still free to play. This is just a value add for people who want to opt-in. $4 a month gets you a free booster pack per week, access to a monthly tournament just for subscribers that features juiced prizes

Access to a tournament is not free entry entry into a tournament. I always understood the VIP tournaments to have some cost as to allow for the special juiced prizes.

cferejohn
06-23-2014, 10:38 PM
From the kickstarter: 540K - Add VIP Program at launch. HEX is still free to play. This is just a value add for people who want to opt-in. $4 a month gets you a free booster pack per week, access to a monthly tournament just for subscribers that features juiced prizes

Access to a tournament is not free entry entry into a tournament. I always understood the VIP tournaments to have some cost as

Huh, must have just conflated "free pack" in my brain.

Makizushi
06-23-2014, 11:16 PM
Yeah, ditto. I'd read is getting access to a free tourney as well, mostly because of the other free access tokens we'd received as part of the Kickstarter.

But thinking about it now, it does make sense to have to pay for the VIP tourney. Otherwise the VIP would gives us 1400 plat value for 400, instead of 800 plat value (as in 7 boosters instead of 4). Darn... Oh well, time to start grabbing cheap boosters from the AH :)

AstaSyneri
06-24-2014, 01:27 AM
When are the free Kickstarter VIP codes going to happen?

Khendral
06-24-2014, 01:30 AM
Is the second picture the blood mentor?

MUST-HAVE-SPOILERS!!! :P

benczi
06-24-2014, 03:28 AM
Awesome article.

Zaxian
06-24-2014, 04:31 AM
With the release of this patch, we’ll be changing the way we award prizes based on the number of match wins (instead of standing).
Yaaaaaaaay! :D

Thanks for the article, it clears things up nicely :)

negativeZer0
06-24-2014, 07:37 AM
DO NOT MAKE THESE FREE

Why deviate from the same ticket reimbursement we have now for draft?

locust
06-24-2014, 11:25 AM
We should be able to auction more than 1 card at a time, like selling 4 oracle songs for 10 plat or something like that

Chark
06-24-2014, 12:23 PM
So is the Swiss Draft "rule and 5 guidelines" already in place or will they be after this patch? If they are already in place, I need to dig up my most recent tournament for you to review. There were multiple byes, people playing the same person twice, and a massive difference in records (first hand experience). I'm not playing another one until I hear if the changes are live.

Live with this patch.

Gwaer
06-25-2014, 12:08 AM
DO NOT MAKE THESE FREE

Why deviate from the same ticket reimbursement we have now for draft?

Because the VIP program is already too good. It's worth paying for entry for better prize support.

stiii
06-25-2014, 09:24 AM
So where exactly are these events in the client? I don't see anything other than 8 man events under tournaments.

Shaqattaq
06-25-2014, 10:40 AM
Large tournaments are currently down as we look at some pairing issues that came up causing tournament progression to halt. We'll inform players when large tournament queues are live again.

AstaSyneri
07-02-2014, 06:57 AM
Large tournaments are currently down as we look at some pairing issues that came up causing tournament progression to halt. We'll inform players when large tournament queues are live again.

And up again!?

Where can I find the schedule for them, outside the client?

Yoss
07-02-2014, 01:36 PM
Where can I find the schedule for them, outside the client?

Or even in client? I couldn't find the schedule anywhere I looked.