Machine Politics vs L'Équipe, August 2010. Photo by Steph Simcox
Required reading before you comment in this thread ;-)
Would be cool to hear the pros/cons of how tie-breaks are handled by Podium, Seattleswiss, and Challonge.
Challonge apparently uses Median-Buchholz.
By the way, i don't think that the swiss rounds algorithm is the only thing that makes a tourney management tool the right choice, it's just one factor.
First, an important point. Polo isn't Chess. Chess is black and white, there is no score. Podium uses the fact that Polo has much more information built into the result than just W/L. When a game ends 5-0, that says more about the skill difference between those two teams than if the game ends 4-3. Systems using Buchholz flatten each result into 1 or 0, and ignore the meaningful data inherent in the game itself. I think this is a major flaw
Podium's ranking criteria goes like this: [points, differential, total_goals] (points: W3/L0/T1)). This is easy to understand, and it is packed with information. Anyone can look at the results table and see why their team is listed where it is, and what they have to do to move up. With Buchholz, you never know what is going to change after the round. Some team you played first round might win, giving you more tie-break points, or some team your opponent played in some round might win and you get bumped down and have no idea why. There is no easy way to figure out why your tie-break score is what it is.
Another advantage is that Podium can stage games before a round is over - something that Buchholz simply can't do. That means you wind up with courts empty for long periods of time at the end of each round - waiting for the last game that was delayed for whatever reason, wasting precious daylight. Podium only needs the results from a few games to make a proper match, because the tie-breaks aren't dependent on knowing the results of every other team in the session. As long as teams have played the same number of games, they can be paired.
To sum it up: sure, Buchholz sounds smart, but it 1) flattens results to W/L. 2) is not human readable. and 3) wastes court time every round.
Podium also has a more sophisticated staging algorithm than just straight swiss (1v2, 3v4, etc). The current system was developed over the course of a year at a handful of tourneys, with valuable input from the organizers and especially from kevin walsh and jonny hunter.
-The first round is what I call a "suicide" round. For example, in a session with 10 teams: 1v10, 2v9, 3v8, 4v7, 5v6. With a decent pre-rank, that means we get the most meaningful information. The top few teams are gonna beat almost anyone anyway, and the bottom are going to loose most games. I pair top v bottom to get those teams started in the right trajectory. The middle teams will have meaningful matchups, and that helps start sorting out the middle.
-The second round I call a "group suicide" round. I pair top v bottom, but inside groups of matching records. For example, say 4 teams are 1-0... inside that, the top team plays the bottom team, the two middle teams play... Same goal: create the most meaningful matchups in the middle, thats where you don't have as much information yet.
Also, this method prevents the top teams from playing too early. You want to hold that off for later in the rounds, when the result of the game will give you more refined information about skill difference.
Starting in Round 3 and on, pairings are done in ranking order. The only reason a potential matchup is skipped is if teams have played already.
Which brings up a final point. By round 5 in an appropriately sized session, a team will have played most of the similarly ranked teams, and the teams around them will have played each other as well. This becomes a more significant pairing criteria than the rankings themselves. Finding a set of pairings that avoids any duplicates means you'll have matchups between teams that aren't right next to each other in the rankings list every single time.
This basically means that the staging algorithm is only so important. There are many other factors that are of equal or greater concern: ease of use, understandablility, flexibility, etc. Podium provides a simple info-packed leader board display. Podium lets you enter scores from the court on a phone. Podium manages multiple courts for you. Podium lets you manually stage and re-arrange games. Podium handles the DE bracket, and seeds it with a single click. I could go on, but thats not the goal of the thread.
OK, that was fun:)
As I said in Phoenix, the suicide round is too extreme. You could start in the middle (1v5, 2v6,...) just as easily and have fewer pointless matchups. Pitting the top team vs. some newbies in their first tourney is fun for no one.
Using a computer is fine if someone else is focused on the bullhorn and getting people on deck, on court. My last two tourneys with Podium were a good contrast. In Phoenix, there was a PA and good communication. In PDX there was just Podium. No one had a clue what was happening all day long in PDX.
Running Swiss rounds need not be more complicated than dealing poker. Force sudden death (no ties) and you have simplified by one order of magnitude. A whiteboard with some hash marks will keep a count of wins and a goal differential. Every time I've set up a Swiss by hand, I've found a person who wanted to learn how it works, and in no time they are running the show and I'm fixing my bike or grilling some food. Every time.
Every computerized event I've been to, the organizer is locked to their machine and stressed out when there's an error of any sort. No one is able to help because it's dependent on the organizer and the machine. There's never a need to call the developer.
I feel like I'm telling people at work that I ride my bicycle to/from work. They think I'm compeletely nuts. Why would anyone ride a bike? Don't you have a car?
I think i agree with you about the first round. It's nice to have some ranking going in, having the top quarter of teams play the third quarter may make more sense and produce better games. That said, the lower ranked teams tend to be less known. At the northside qualifier, i think the bottom quarter and second-bottom quarter were pretty evenly matched.
Regarding forcing sudden death (no ties), i think this is a tough question. I don't have a strong opinion about it, but it does risk making swiss rounds take a lot longer. If it's hot out, players end up taking breaks before their overtime etc.
Regarding everything else you said, it's irrelevant to this thread. This is about algorithms (which can be very simple and not need computers)... not whether you're using pencil, pen, or sharpie.
This is a good point... if we had better pre-rankings, it might be worth trying the top v middle approach. But as it is, the only accurate pre-ranks are the very top, teams that are obvious, and the very bottom, teams that haven't played in any tourneys before. The rest is a big blur
Now that we have a full season of qualifier data, it might be interesting to go back and re-pre-rank the players according to zachs calculation, and see how well that matches the actual results.
Being on a team who regularly participates in the lopsided first round games my impression is that teams who are at their first tournament are usually pretty excited to play us, even if the end result is often 5-0.
I never have a problem with these games either as I have become confident in podium that the next games will be exactly as they should be and everything will work itself out.
I sure had fun at DPI4 when I got to play the Guardians for my first game, even if it was 5-0! it was a great warmup, and my second game was a great matchup. I'm of the opinion that the first rounds are important for creating a trajectory for a team - top teams should go way up, bottom teams goway down, and the middle teams should battle each other, not just get beat up by the top teams, who are often significantly better than almost all but 2 or 3 other teams.
And are you referring to pdx III, where we were drinking whiskey before the first game?
Forcing sudden death could be a huge time waster, I've seen games go a full 10 mins without a goal.
How do you prevent duplicate matchups in your whiteboard system? With say 24 teams, by round 5 are you going to look over a list of 40 games each time you make a match?
You are arguing against automated systems altogether, I'm hoping this turns into a good discussion about staging algorithms.
if time runs out with a tie score, keep going until someone scores a winner. No break. Lots of things can cause a tourney to run long. I don't think this is one.
Prevent re-matches by going backwards and verifying matchup didn't happen in a prior round. If you find a match, look for another team to pair. If there were 200 teams then you've made this easier by adding degrees of freedom (reduced likelihood of prior pairing). Granted, 200 teams would have bigger challenges. Swiss was "invented" to run large chess tourneys -- way before computers existed.
Lots of people will have ideas about how to make it more complicated (e.g., count shots or saves). Accomodate what you will. If everything is on a big board and visible, people see how wins and goal differential is all you need to set up good matches
Podium is perfect. Everyone who deserves to make it into Sunday will, and with a 24 team double elim, a lot of teams that don't-really-deserve-it will also make it in.
No system will give you a perfect match every round. There's no such thing as a perfect match. It's a ranking system, and goal diff is the only tie-breaker you need. With a pre-seed, its the PERFECT tie-breaker. How many goals you put up and how many you defend against is a rock solid indicator of tournament performance. Basically, see Vince's explanation above.
I don't appreciate competition to Podium. All you're doing is decreasing viewership by people who aren't at the tournament by offering an inferior follow-along experience and making it harder to sell sponsorship and make the programs a sustainable venture.
hair-splitting for the most part, but an interesting topic.
you could take the data you already collect in each game, and depending on how you weight it, you may be able to produce slightly different rankings.
basic goal differential is probably a reliable indicator. i'd be curious to know if you took scoring efficiency (goals scored or defended / min. played) on those first rounds into account, if the results produced different rankings.
do any of these existing systems take the duration of 5-x games into account? is game duration even recorded at all?
I agree with Kruse. PODIUM is clearly where it's at.
Tie breakers are not needed but at midwests there was something weird that happened, if you look at the morning bracket the team that placed 2nd (stragglers) didn't play the 4th (fluffy ducks) and lost to 3rd(mitten polo). Mitten had two ties which were to 1st and 4th place teams. And fluffy ducks had a loss to 1st and a tie with 3rd.
So I feel there still needs to be another level of complication to placing teams. Some how. This team slipped through the cracks the get ranked high just because the had 4 wins only over lesser ranked teams.
I think that if we can trust the preserving enough that before goal differential takes affect in the seeding I think there needs to be a value assigned for "toughness of schedule" which for teams with a ties or even a loss could get ranked above another team with a single win more.
I think what happened is that stragglers lost their first game and then won their second game by a low margin of victory. Then Mittens went through and tied everyone and the stragglers goal diff. was pretty low so they ended up sneaking up into 2nd because they were getting pitted against easier teams right until the end.
This happened at Ladies Army III in Austin. #1 Cunning Stunts played #2 C.L.A.P. and tied. #3 Nasty Ass Honey Badgers won against the #4 team. #3 Nasty Ass Honey Badgers became #1 without playing the #1 and #2 teams.
hardcourtbikepolo dot org
1) Bourbonic Politics 4/0/1 +15 17
2) the Stragglers 4/1/0 +3 13
3) Mitten Polo 3/0/2 +10 15
4) Fluffy Ducks 3/1/1 +12 17
So here are the final standings. One thing to note is that the order is determined by W/L/T only. Goal Differential doesn't even come into play here... Even if a strength of schedule was used instead of goal differential, the order would be the same. You'd have to elevate the tie-break to be more important than the actual results, which I don't think is a good idea...
The Stragglers wouldn't have gotten that high if Mitten didn't have so many ties. They seem a little out of place, but they did beat 6, 7, 9, and 10. Luckily, it's not meant to be perfect. They were promptly taken care of in the DE Bracket.
It wouldnt hurt it any to order teams by wins and then teams that are +/- 1 win of each other would be compared by strength of schedule. If there strength of schedule is considerably different then it would place the team with the hard schedule above the team with one more win. This is just nit picky but I think would solve the issue of ranking due to ties. But like Vince said it doesn't make a difference they were taken care of early. But it does make a difference when the top 8 after Swiss get a first round bye.
Preseeding not preserving but otherwise besides the explanation of who beat who and tied who... The rest is pretty clear.
I would be interested to hear from Vince what impact unlimited goals would have on Podium.
I think this would be a fascinating alternative to try at a tournament!
So right now, I think the 5 goal limit is one of the reasons my suicide round technique works. Without it, it's possible we could try doing away with the suicide first round, and pair teams using the top v middle technique that has been mentioned. The reason being that the wider spread of goal differential will do the job that the suicide pairing did.
I'd love to try it. It would mean we have much more information embedded into each game, and that seems like it could only help
i'll see if saskatoon will do this at our prairie province region tournament this weekend. we did unlimited goals last year, so the only change would be using podium. it's a small tournament (8-10 teams) but would offer some insight at least
I would also be interested in doing this in Saskatoon this weekend. Provided Saskatoon is down to do it and Vince is down to help us with Podium. The only real downside is that our tournament this weekend is going to be small. Swiss rounds for small groups are sometimes unnecessary. For example people got mad last year because we ran something like 7 rounds and middle ranked teams were forced to play top ranked teams.
actually this raises another question, do swiss rounds lose relevance once everyone plays everyone?
what i'm really asking is what is better.. full round robin, or repeats of close matches?
I am of the opinion that Swiss rounds can give us an accurate ranking of a large number of teams where full round robin is not possible. If full round robin is possible then that is the more accurate method in my opinion.
Even a relatively small tournament, a full set of round robins isn't possible. I'll email out what everyone thinks of using Podium this weekend...
Podium should be able to handle this as it stands. The goal sliders only go to 5, but you can enter any score in using the keyboard...
I like Podium. Keep up the good work, Vince!
Bike Polo is Our Language,
Good Will is Our Motto,
Friendship is Our Goal.
Play More Bike Polo Together!
About challonge, i don't understand this statment very well, can someone explain it to me?
Is it a mix with Median-B and other Tie-breaking system?
CHALLONGE! primarily uses the Median-Buchholz system for breaking ties. The "tie-break score" displayed in the results table is the sum of a player's opponents' scores, with the best and worst scores discarded. Should players be tied by score and tie-break score, they are sorted based on their records relative to one another. Seed breaks up any remaining ties."
Loves podium, (Vince, Johny and Kev i will contact you soon about the worlds).
By the way, are you planing in some near future (1 year?) to "release" it as a public website/app?
I'm not sure what you mean by "release" it... It currently is a web app.
By "release" do you mean open source it?
I mean that anybody can go to a website on click on a box called "start a new tournament" and start playing a bit with it without having "bother" you in person.
Clement: hope you like pepperoni-brother pizza :(
edit: Kev told me it's pretty simple to handle now? What are the requirement for not being dependent on someone else? Ruby?
i'm not sure what you're referring to but it's not necessarily easy to set up on a different server. I believe the current code is not on github anyway.
i think what i might have been saying is that it'll be no problem in Geneva to use Podium, there will be lots of people with experience using it for tourneys on site.
Anyone can try it out at http://hardcourtpodium.com/sandbox There's help text, and the data can be reset under the manage link
Setting up a tourney isn't a bother, it actually helps to open a line of communication for support and recommendations.
Didn't see this sandbox before, sorry about that and thanks for the information.
i think he means when will it be available on the itunes store
By "Johny" you mean "Jonny"... Gosh i hope it's not true. Last people who didn't spell his name right finished in "salami".
Another important improvement would be to add in the tie breaking games for 9th-12th place. There have been many tournaments where one or two of the top 8 teams have already qualified so 9th-12th had to be played out to determine which teams would make the qualification cut.
He means off topic. you're right, 9th through 12th needed to be played out this year at NAHBPC qualifiers (and it was, i believe, wherever it was needed). but this has nothing to do with swiss rounds tie-breakers.
No but it is something to add to podium which this thread is about essentially.
GD is a pretty poor way of tie-breaking teams that have played different teams (unsimilar paths through the rankings).
I'd argue that a W/L ranking (weighted with the final ranking of the teams) such as Buchholz is far superior to GD, who cares if you 5-0 the weakest team? You could argue that not using the current/final ranking of the teams during your tie break process is also ignoring important data?
Buchholz is easily readable once explained, it's just not as predictable (as teams move dynamically each round and affect your score). GD's predictability shows the flat nature of that data in my opinion (with a few early miss matches sending you clear of the pack).
The odd rankings come about in Challonge (and Podium perhaps?) because the system searches for the most similar pair of teams first and then the next most similar teams in turn until you are left with teams that have been marginalised by the system (they are already "well known" for that given round).
We'd love to use Podium in Europe but it has crashed and burned at many European tournaments with no-one able to pick up the pieces... a localised version with a bit of documentation and we'd be away, Challonge is the most widely used tournament software in Europe.
In Podium, GD is only used for W/L ties, though, right?
Vince mentions up the thread that it's Points (Win:3, D:1, L:0), then GD, then total number of goals scored (GF).
This would assume that a team that is pushed down to mid-table is inferior than a team that hits the bottom but works their way up... in my experience the opposite is usually true. Buchholz will show this (if true) by awarding merit to whichever of those two teams defeated the highest ranked teams, etc.
With RR (or equivalent grouped format where teams play similar competition) then GD is bang on, but for Swiss I don't think it's adequate.
Yes, I've had a few bizarre errors pop up when it has been used in Europe.
Some have been mis-understandings of how the interface works, which I attribute to the challenge of explaining technical ideas to folks who don't speak English natively (and I only speak american and computer, unfortunately). Also, I'm always asleep when the tourneys are happening, and can't provide live support.
A time or two there were issues I haven't been able to replicate, which makes me think something weird is happening crossing the atlantic, but thinking about that makes my brain hurt.
That said, there have been a number of tourneys in Europe that have worked out just fine.
Also, if people in Europe want to use it, it'd be great to have someone locally that could provide the support that I do here in the states...
As we are planning to use it for Worlds, we pretty have to be ready to handle any shit who happens.
What kind of knowledge do you need to provide support? Is a K.Walsh in the area enough?
Because an overseas hotline isn't a solution at all.
What about issue Jono pointed about meaningless goal difference against week team? Can podium handle something as a 5-0 against a team who loose all their game 5-0 is less important than beating a good team 5-0?
By the way, thanks for every minutes you put in.
having a Kev around is certainly enough support. he's been using it since the beginning.
The drift into setting up double elim is interesting. Double elim is actually a more challenging system to work out by hand. Afaict, printyourbrackets is what people copy, and once drawn out then you're done. I've found the order of games from PYB can be odd 'long wait, followed by back-to-back).
I did it by hand with 16 teams over Memorial Day. Came upon a re-match in the losers bracket. We just did a quick substitution though, no problem. If it were 17 teams, I guess you just have a dangler bracket down in the losers'? The way PYB sets things up feels odd, but it's harder for me to grok. It works. If you ever do a random double then you see the value in seeding. It can really suck to be on the wrong side of an unbalanced double.
You're substituting rematches in double elimination? Uhh... WHY?
Podium is great, it has done what it needs to do. I've been happy with it at every tournament I've been to where it was used, (except for saturday round 6 at ESPI, but that is another topic). But is it perfect? No.
Personally I am all for Buchholz. Buchholz fucking rule. Congrats to The Growlers for being Buchholz champions @ LACOED 2012. It is so much fun chasing not just wins but also Buchholz at a tourney.
I think that every polo tournament here on out should use Buchholz and each player should keep track of how many Buchholz they've got throughout their career so when you and your bro both have the same goal count and shooting percentage you can know who is actually better because of your Buchholz count.
But realistically the Buchholz system is a legitimate option and shouldn't be ruled out.
Basically Vince's only arguments against Buchholz is a)wasting time and b)not being predictable.
A) As tournament venues get more legitimate they will have lights and this won't be a problem.
B) That is the f*cking beauty of it!!! Knowing is blowing it! Who cares if you can't predict who you will play, or how many buchholz you'll get, it will just make you that much more suprised/stoked.
Also, where is warrior mike in this conversation?
I had to google buchholz and not sure i got the right answer. Can you describe?
sorry should've cited the buchholz i speak of -
that sounds like it would slow down tournament play, creating big unnecessary lags between rounds;
that means less games for each team, right?
Ok, so I'm one of the main organisers of tournaments in the UK, and I use Challonge every time.
I'm not happy with Challonge for a number of reasons, for one it's bizarre pairings at times (even when you take MB into account).
I would love to use Podium, but I've heard horror stories about it, and as an organiser, that scares me off.
With Challonge, I can enter the teams 5 minutes before a tournaments, change teams, add teams, remove teams, change the number of rounds during a tournament, and lots of other things that make it very flexible.
Vince, please correct me if I'm wrong, but can I set up a Podium tournament entirely on my own without any help from you. Can I do those things above? Until that's the case I'll stick to using Challonge, though I would rather support Podium.
Basically I'm asking for a "create tournament here" page, and a page to edit the tournament while it's ongoing.
Also, I would need it to be able to support round robin, single elim, double elim, and ideally group stages (rr or swiss), something challonge doesn't do either.
It supports multiple Swiss sessions, and the results automatically seed the bracket (SE or DE). It can even do multiple stages of Swiss... ex: play two sessions of Swiss day 1, use those results to play two more Swiss sessions on day 2, and use those results to seed the bracket. I think that you can see how complex it can get, which is why I prefer to create them myself
I don't have direct support for round robin but I could make an option to turn off auto staging, and you could manually stage as if it was rr. That kinda defeats the purpose of the whole thing though
Last time I used podium I was able to alter teams, add/remove teams, change the number of seeding rounds during the tournament, and change the number of teams in the bracket. However, I could not go from a double to single elimination bracket mid tournament.
Ok, well that sounds good.
But yeah, a complete online tournament setup, that I can do remotely would be required, for me to use Podium.
You can switch between SE and DE now also
Can you explain why that is so important? For me its simply much easier to do it from my end, and I like to have the line of communication open to help out.
Just because, at least with tournaments here, things are so unpredictable. We've had teams drop out minutes before, new teams arrive after the first round has started, the number of swiss rounds change during the day, and various other things.
And I'm just not comfortable being reliant on someone on the other side of the world, if something goes wrong.
It's the main reason I don't use Podium right now.
A "create a tournament page" very similar to the Challonge setup would be so useful, just select the type of tournament, number of teams, number of rounds, enter the data, and off you go.
Add in your better seeding, and the automatic seeding of elim from the rounds, and Podium would be so much better than Challonge.
I've been thinking about it, I'm going to spend some time in the near future to go ahead and build this...
During the EHBPC last summer in Barcelona, we had the intention to use Podium for the wildcard, which was basically swiss rounds. Everything was set up correctly but just two teams not showing up and a new team that finally did it, and we ended up using Challonge, because we weren't able to edit the teams list minutes before the tournament started. I don't know if that has changed.
This has been changed.
that and a whole bunch of other options
to try it click the manage link at http://hardcourtpodium.com/sandbox
Vince, can you switch the # of courts mid-tournament during the elimination rounds? Also, could a person potentially set up a tourney without your help other then initially http:// creation for the tourney?
Yes, you can alter the number of courts at any time during a Bracket... There's a description in the section that mentions this. For swiss, you can do it before you begin.
So far I prefer to keep the tourney creation to my side. That is so I can handle creating and connecting the varied number of sessions/brackets, as well as import the player data, which comes in number of different formats.
If this is something people really want, I may go ahead and do it, perhaps I'll put together a default tourney setup, and if people need to customize it further, they can contact me.
....so this thread turned into a podium help session
West and East squash the beef
That shit 's legit as fuck!
I did some comparisons by adding two columns to the typical Podium results
Sp: Strength in points, sum of opponents points
Sd: Strength in differential, sum of opponents differentials
The colored areas are teams that had the same record (those are the ones where the tie breaker was in effect)
Reading the dots from L->R, you get the order if [Sp,Sd] were used instead of [D,G]. Matching outcomes would be a line at an -45deg angle. Opposite outcomes would be form a line at a +45deg angle
Do you have an opinion about which is most appropriate in tourney play?
Yeah I'd be interested to hear your thoughts Vince.
In your opinion does the difference in GD come from consistently scoring an additional goal (or two) in each game, or by achieving a 5-0 (or two) when paired against a weak team?
This is cool! Worth noting that in these two tournaments, no one was 13th (just missed getting into Sunday) with a more difficult schedule (higher SP/SD) than the 12th place team.
But I bet that has happened.
I think that strength of schedule would matter more in round match ups then in placing teams after the SD was used to match up.
You sure? What about Old Fat & Angry v. Eat Some Chips in Eastside PM?
But yeah, what Jonny said -- this would be much more influential through use in picking the round 4, 5, 6 match-ups.
With additional rounds you have more variation in record and less need for secondary and tertiarry measures of team strength.
after seeing swiss rounds used for 4 rounds and 5 rounds at NAHBPC I think there might need to be some discussion around how rounds are staged. First, is that there are incentives set up to loose. There was a match that we tied with slab of bacon. By tying that game we ended up dropping a few spots to the last qualification spot for that bracket. But if we lost that game we would have ended up with a very easy game and ensured qualification into the saturday afternoon bracket. Instead, we ended up playing a much better game (which was great, not complaining about playing good teams) which we won and ended up in a pretty decent position for saturday. What i'm trying to point out is that the qualification into the last few spots of swiss round is very subjective and teams could game the system by throwing one game and ending up in a position where they play an easy team and then gain qualification. I think that it makes sense to find out who the best team is but is a poor system in terms of determining the lower ends of qualification if we don't take strength schedule into account.
This is an huge issue in my opinion. How can we avoid stuff like that?
Yes, draws/ties tend to be a bad thing in Swiss, in any tournament I organise we always do golden goal during Swiss, to avoid this.
As you say, if you drawing with the game almost over, it can tactically be much better to lose, than keep the draw (depends which round you are in though, and where other teams are around you).
What is also a bad effect of Swiss, is that it's far better to lose early, than lose late. This is because in the early rounds a loss can send you much further down the table (or a win much further up), than at the end.
So if you think you are a mid table team (and are likely to lose some games at some point), tactically it would be better to lose your first game or two, and then have 3 easy games to get to the middle.
Obviously it would still take a very brave team to lose a game on purpose.
Wise words, per usual.
Ties simply aren't very helpful when it comes to rank-ordering teams.
Sandbagging will always be a risky strategic choice in qualify rounds. That's why a proper tourney finishes with elimination. Elimination puts an end to sandbagging. Intentionally risking elimination can definitely backfire.
I think sandbagging is a Monday morning strategy..."yes we meant to lose that game because we're so good and strategic." (But not good enough to win every game?)
i like "Monday morning strategy"
Ok, so Podium made some decisions I can't explain, and again puts me off using it, vs Challonge (and I mean this in a constructive way, not a whiney way, I really want Podium to be the best solution out there).
There were some strange pairings in the Swiss rounds at the Euros today.
For example, Morning sessions:
Cosmic played Schlammbeiser twice.
Edisons, after winning three games, got drawn the bottom team, Passing Wind.
It would be interesting to see why it chose these things (in my mind neither should have happened).
I've administered Podium at about 6 tournaments and "watched" it about 10 times and i've never seen any errors like that. Maybe Vince has insight. I'm assuming the issues in the afternoon were more to do with lack of internet on site?
The problem with not many teams in swiss rounds is by round 5, if records go a certain way, you've played the teams that have similar records so it runs out of teams with similar records to play. It is a swiss problem, not a podium problem I think.
this shouldn't normally be a problem actually. What happened Friday morning at EHBPC is either a bug with podium or human error, or a weird combo.
Sure, but this was round 4, in groups of 16.
4 is the mathematically correct number of rounds with 16 teams, and at that point, two teams with 3 wins should be playing each other (though it is slightly complicated by ties).
But either way, in this case both Edisons and Passing win had much more logical partners (and as far as I could tell, it wasn't caused by other teams having better partners than them).
Those certainly look off, thats what I thought too when reading your post.
I took a look at that specific round in Podium, and for some reason, the game between Edisons and Cosmic in Round 3, which was the first game of the round, was entered into Podium last, after the other two games on the same court. By the time it was entered, all the other possible pairings were taken, and Podium's algorithm had no other choices available to it.
The order of games is intentional, with higher ranked teams playing first. If the game was entered immediately (assuming it was played in the correct order), then it would have had similarly ranked teams for it to pair with. But it was entered near the end, when the staging for the bottom ranked teams takes place...
Another thing I noticed is that it looks like one of the courts might have gotten backed up a few games, maybe a long mechanical or something similar. Since the games are meant to be played in order, this can cause odd pairings since the teams that will wind up with similar records are dispersed, and not where Podium things they will be...
The reason I use the order of games as part of the system is to avoid the potential long delay at the end of each round, waiting for all the results to come in before staging the next round. For example, if one games goes 10 minutes long for whatever reason, that would be 10 minutes lost per court, and when trying to play hundreds of games in a day, thats a big loss.
Ok, the Edisons vs Cosmic was postponed to end of round, due to an injury in the 1st 30 seconds to Snoops of Cosmic.
Ok, I now understand how this happened.
But, in that case, I'd say that's a flaw with Podium, or at the very least a lack of an alternative feature.
I appreciate what you are trying to do, starting the next round early can be an advantage, but as a organiser, I've already factored that into my timings, I don't need to be able to do that.
But I do need to be able to be flexible with the order of my games, accidents do happen, games need to be delayed, as was shown in this case,
What I need to be sure of, is that the pairings are as good as possible, and if you are predicting the next round before the end of a round, you simply won't get as good pairings.
So in summary, I think you've taken the wrong approach. It's a nice feature to have, but I wouldn't want it to be the only option. We use Challonge now, and I very rarely have a problem with more than a few minutes downtime anyway.
If an organiser really wants to do that, they can manually decide the first game(s) of the round, and then let Podium work out the rest when the previous round is done.
yeah that 's exactly how you should use it... set up the first phase of each round "manually" to get games going (you need a basic understanding of the pairing system). then let Podium do the rest.
Yep, that's fine, and yeah, I understand the pairing system well enough to be able to do that, I just don't want podium to do the automatic ones until I'm ready for it to do so (at the end of the round).
Having had a look at the sandbox, I can see what I have to do, to get it to work the way I want.
I now see I can manually unstage automatically selected games, and replace them with manual (more accurate) ones.
But I don't want to have to do that for every game, then I might as well do it on paper.
You can see on the example I've posted, that some games have been chosen, but depending on the result of the last game, that might change those pairings if they were worked out after all games. I'd prefer to have the more accurate pairings, after the end of the round.
So would it be possible to add the option to not automatically stage games, until the end of the round. It doesn't need to be the default option, but as long as there is a checkbox somewhere to turn it off. Or maybe a button which says "stage next round" and doesn't fill up the games until you are happy with it.
Then an organiser can add any manual games they want, before the end of the previous round if they desire, and then fill up the rest at the end of the round.
I would be much happier using Podium, if this was the case.
I'll definitely implement this.
In the early days of Podium, there was no automatic staging, and there was a button for it. Turned out to be confusing to most people, because that simple interaction actually meant they had to understand a lot of what was going on, and thats just too much to expect from most organizers.
This will be a sort of "Pro" mode for people who want the most control...
Vince that's super cool of you to be so receptive and accommodating. Lots of people would get defensive instead. Big ups dude!
Yeah, awesome, thanks Vince.
Just been testing it out on the Sandbox, and I love the "Pro mode" changes. The default is still the current method, which is great for users who don't want to get too involved.
It's so good we have access to someone like Vince, this alone makes this so much better than Challonge.
I will now be recommending this for European tournaments.
What about for Hell's Belles..?
(Answer my email)
Those are errors. Why not just manual stage games when that happens?
There's no "right" way to throw a tourney. If the King says "you two go joust" then ya get out there and joust. When there's no king, the peons have to debate over what's best.
It is a characteristic of the Swiss system that in a small tourney (< 16 teams), mis-matched games are forced in later rounds. Imagine an 8 team tourney with 1 A-class team, 2 B and 5 C. Chances are the A team will beat the two B teams by rounds 3 or 4. If there's another round, the A team will need to play a C team. If you had a 50-team tourney then there are a lot of degrees of freedom...likely lots of teams of comparable skill, or lots of options to choose from with pairings.
Now, I would never intentionally re-match teams in a round robin. Round robin means everyone plays everyone else, so a rematch is like a bent round robin. Vince, it's hard for me to reconcile your statement about not having equally matched pairings with your decision to set up the first round as top-bottom (i.e., starkly unequal to begin with).
Manually working out Swiss, while possible, is not really a realistic solution when you are under time pressure at a tournament.
You should be able to stage games manually within podium. Not sure if this feature is available but in the versions i have used then it is very easy.
This is exactly what I'd recommend when weird pairings show up for various reasons (I explained one just a few comments above)
Organizers are only a click away from un-staging the automatic choices that podium makes, and selecting any pairing they find appropriate. (This is explained directly in the interface of Podium, but perhaps I need a few translations, since it might be harder to understand to someone who isn't as fluent in English as native speakers.)
Ok, in that case, that's a lack of familiarity with Podium from the organisers.
Due to the fact that I've organised a lot of tournaments, I was called in when the problems happened, to try and help out. As I'm not familiar with Podium, I wasn't aware of that feature (and clearly the organisers weren't either).
So to some extent that's our/my fault. And I apologise to the teams affected for that.
I've now had a good look at your sandbox, and I feel a lot more comfortable with how it works, though I still don't like how it does the early pairings.
The rankings produced by Podium for NAHBPC double elim pretty closely resemble the final results of double elim. The top ranked team won, the top 6 all finished top 6, and 13 of the top 16 after swiss rounds finished in top 16....
Is someone (Greg/Vince/Etc.) able to chime in with what actually happened with Podium at Euros?
• TORONTO • OTTAWA • VALLEY BOYS •
Yeah, I was playing in Passing Wind at the tourney. I had my phone with me, with Podium on it. I was checking it regularly.
I looked to see what was going on after we had lost 3rd game (0/3), and saw that we were in the hole, and drawn against Schlammbeißer, which made sense, as they were ranked around us (i.e. last or nearly last).
We were called to the court, and then suddenly Edisons appeared. I thought, and said, "this isn't right, we're supposed to be playing Schlammbeißer". I was told by the officials that, no, we were playing Edisons. After a bit of grumbling (no profanity!), I got on with the game.
After the game, I was still utterly mystified as to what happened. I checked my phone, and indeed, Podium had shown Passing Wind vs Schlammbeißer. Then the page auto-refreshed, and was gone.
I saw few "connection lost" (due to poor wifi, like a semi-public one where you need to enter the account information every x minutes) and many brut-refresh of the Podium webpage this weekend.
I saw some matchup being announced (saw them on the webpage) and then, after entering scores from other games those matchup changed (what happend to you apparently Bill).
My guess is that it is a mix of hardware issue (poor connection) and human mistake (brutal refresh, like just after entering the score).
So, i'm not charging Podium for theses issues. That being said, the fact that it (Podium) seems so sensible to human and hardware issue makes me very scared and skeptical about using it for the Worlds.
As far as i can remember i don't recall any euro tournament started and finished with Podium, maybe i'm wrong.
i don't know why people in europe seems to have so many problems running a tournament with podium. all you need is an internet connection and someone who isn't too drunk. are those 2 things that hard to come by?
I agree. I ran a tournament from my phone here in the US and had no problems at all.
Didn't someone have to wake up Vince because it wouldn't stage the next round?
Plus you add the fact that we can't wake up Vince at 3 in the night to ask for help.
All the experiences you guys points about tourney run well in NAH with podium won't change the fact that we are all (i mean tourney organizers) really scared about using it right now. Swiss champs in 2011, Padova and EHBPC 2012 were really bad experiences.
I also really want something that can get run if you loose the internet connection.
I can't explain why we've had such a hard time using Podium in Europe.
There might me some issue with the latency of sending every single request back and forth across the atlantic (The server a Linode VPS, somewhere on the east coast).
There might be some difficulty understanding my instructions and the help text if the organizer isn't fluent in English, since it's a somewhat technical subject matter.
It's as frustrating to me as it is everyone else...
I would like to point out that Kevin and I will be at worlds and will be able to help with running podium. It isn't perfect but from the person who is interested in seeing what happens it is a million times better then challonge.
I second this, as a competitor, would love to see podium at worlds
while our experience at ESPI was 99 percent great, the other 1 percent could've been prevented by a little better preparation on our part. the one big thing that we didn't practice enough was restaging games.
i think the software instructions are pretty clear, but a longer FAQ or whatever might be useful. something that gives a list of things you should know how to do before tournament day, lays out what to do in certain "problem" situations and gives details how to recover from common podium user fuck-ups. i'd be happy to contribute or to compile one myself, if it'd be useful.
I've made a huge amount of changes in the last few weeks...
1) Major speed increases - fancy memcached use, background processes, lots of optimizing. Zero delay when navigating through the sessions, teams and brackets when following along
2) Re-written follow/homepage: http://followpodium.com always has the latest tourney, good for bookmarking or adding to your phone's home screen.
3) Trickery to prevent odd behavior when used on slow connections. Mutex locks to prevent race condition errors, etc
4) Option to select Auto-Pilot or Manual Staging
5) Other stuff I can't remember
Coming soon: Create your own tourney
Some thoughts on the suicide first round. I use Top v Bot, and some have suggested Top v Mid. JohnH and I have been talking about it and he brought up a good point: If we used Top v Mid, it creates a pretty bad situation for the middle teams. For example, the mid+1 team will play the top ranked team, and the mid-1 team will play the bottom ranked team, so that means two teams ranked essentially the same have hugely different pairings. With Top v Bot, those middle teams will play eachother.
For me, the point of doing suicide is to get meaningful info out of as many first round games as possible, so mid v mid is a meaningful result, and for top teams, they'd probably win anyway, and bottom teams will probably loose - which means all results are useful information.
I made this chart to examine this: the Y axis is pre-rank, top teams are at the top, bottom teams at the bottom. The X axis is the result of that first game for that team. The blue curve fits nicely. The red lines are what I predict would happen if we used Top v Bottom.
Top-middle to me...an example from a 16 team tourney is 1 vs 9, 2vs10, 3vs11, etc.... That is, top vs middle, 2nd vs middle minus one, 3rd vs middle minus two. I believe you're describing something else.
My opinion is first round pairings matter little in typical bike polo rounds. Pure random draw first round would be acceptable. Suicide is least optimal, for reasons I've already stated.
If you had 100 teams, the suicide algorithm would be horribly imbalanced for top and bottom teams, while the games would become progressively more balanced as the round worked through to the middle teams.
don't just stop... 1v9, 2v10, 3v11, 4v12, 5v13, 6v14, 7v15, 8v16
so 8 plays 16, 9 plays 1. sucks for 9, awesome for 8. thats what i'm saying. the randomness of assigning a team either 8th or 9th seed has a very huge impact.
TvM is unbalanced for the middle. TvB is unbalanced for the top and bottom. I'm saying lets accept that top teams will win their first game, and bottom teams lose their first game. And lets get the novel info from the middle pairings.
1 v 16 is more better than 8 v 16?
TvM is smooth. Random/inaccurate initial seeds are balanced.
Another way to do it would be 1v2, 3v4...15v16. While the first round may be optimally matched pairings, it's a crappy solution for another reason...one typically wants to see the true two top teams paired off in the final round, not the initial one. Forcing a loss on one of the top teams could lead to really inaccurate results if there are few rounds.
Curious...how are you pairing 2nd round teams? Do you stick with the TvB as a default?
Yes, because no matter what the pairing, 1 is likely to win, and 16 likely to loose. So we rig that to happen anyway, since having middle teams play each other gives us more meaningful info than the result of, for example: 8v16
This is exactly why I use the suicide rounds in the first place.
Round 2: TvB inside point groups (ie: winners play winners, losers play losers). This is the round where I'm considering using TvM instead of TvB
Round 3+: Closest record
Why give the top teams the easiest matches ("rig" is your word) and make the middle teams fight tougher bouts? In my opinion, the natural balance of TvM is more fair for all.
Initial seeds are highly fallible. With TvB, the tourney director's initial seeds have far more influence on the first round results. The highest seeds get the easiest games. One could bump the local team up to a 3 or 4 initial seed and practically guarantee a match against a weak team. One never knows with middle-seeds, that's where dark horses often start.
Pure random initial seeds would be most fair since it takes human judgment completely out of the equation. In Swiss, that makes a lot of sense. If I used a computer I'd likely do that. By hand, it would make setting up the first round look like getting a game of D&D started. I may be a complete nerd but I have my limits.
None of the options are perfect.
1 vs 2 messes up your seeding, for the first few rounds you'll have strange match ups, and results will probably be less accurate (IMO). After the first few rounds it's good though.
T vs B produces some very one-sided games for a few rounds, but at least the results should be accurate (assuming you get your seeding right). Within a few rounds the sets of teams on the same score are pretty small, and then you get good games for all teams.
T vs M produces a set of games which have equal difficulty, which is cool, but it does mess up the middle teams for a few rounds. Now that inconsistency should balance itself out after a few rounds (9 should then get a much easier run of games, compared to 8, assuming 9 loses, and 8 wins).
I work on the basis that the person in charge of the seeding won't cheat, I certainly wouldn't, ideally in the next few years we'll be able to use stats, such as hardcourtbikepolo.org to objectively seed teams based on past results.
I'm undecided whether I prefer T vs B or T vs M. I've mainly used T vs M, through using Challonge in the past, but I'm looking forward to using Vince's custom pairing (T vs B for two rounds, then traditional closest available pair).
I don't think there is a magic formula for those first 2 rounds, whatever you do won't be perfect. But what I really don't like is random. If you get your seeding and pairing right, it's like an extra round or two of swiss, compared to random, in terms of result accuracy. Random shouldn't be used for anything except round robin.
In my opinion it shouldn't be having balanced first games, or trying to work out the path of any given team (or any other tournament preference).
Statistically speaking it should be about what information you want to prioritise in the opening round. With TvB you are stating that the mid-table range is a more important area to find disparity within. With TvM you are stating that the extremes (top and bottom) have to re-prove themselves as being placed correctly before you start to find accurate disparity between the mid range teams.
If you have good seeding, then TvB should always be used as otherwise you are looking for somewhat duplicated information because you want to abandon the predictability of the opening games (which shows that your system is working). Any other approach places individual preferences on how a tournament should play out over the most accurate way of ranking all the teams?
Perhaps there is an argument for trying to create an interesting route to the final for the top teams, or putting lower priority on teams of a lesser ability (if you don't care about a ranking for all teams), but this gets pretty complicated and goes over my head.
I would have to ask Jake if he did any rearranging, but I am almost positive that we used strictly the hardcourtbikepolo.org stats to blindly seed North Americans.
Swiss is very efficient, ranks begin to stabilize after round 2. I bet one would be surprised at how the efficiency makes random initial seeds a viable option. I've never actually run a random initial seed Swiss, but I think it's worth trying. I realized dice are actually common at polo events. Each team rolls for starting position is easy.
I think TvB is the extreme in 'rigging' bias, I can't conceive a system more unbalanced, and I seem to be the only person who's ever even noticed. How bad could random be? The bottom line is matches got to be set up and put on court. All we want to do is play. There is no right answer.
Bias via seeding is a reality of human intervention. Nothing is perfect. A tourney director has more info about some teams than others, human judgment is what it is.
It'd be fun to set up a simulation engine, in my free time. Hah. It would provide empirical weight to some of these assertions. But I know you ultimately end up with a problem of 'what does +.5 phi coefficient mean in the real world?'