Throwing a Game

[manary]

Hypothetically, let's say a team wants to throw a game (lose on purpose). There are a few reasons for this - Maybe they want to leave early, so they want one more team to get in to the playoff division. Maybe they may lose to their B team so that both get in the top X of the field, and both qualify for a national tournament. Maybe they want a lower playoff bracketing for some complicated reason.

What do people think? Can TDs stop this? Should they? Should the second team qualify? Is this wrong? Immoral? Blasphemous?

Also, do any of your answers to the below questions depend on how they do it? What if they neg 20 times, or alternatively, what if they play pretty hard, getting a decent PPG/PPB, but just give up a few easy questions, or something in between?

What about HS vs College?

ALSO should teams be required to make their A teams their best, so that TDs can bracket correctly? What if they split their A team on to B so that both qualify for nationals? etc etc

[Auroni]

I don't know if throwing a game is technically against any established quizbowl rules, but to me, it's definitely a violation of the spirit of them. I'm opposed to any attempt to pull this off (and, usually, such attempts are pretty transparent). However, I want to stop just short to intentionally awarding points for intentional wrong answers or not negging teams that are intentionally negging and I don't know what the best way to handle such obvious shenanigans is.

[dtaylor4]

In bracketed RRs, this is why TDs should do their best to separate A/B/C teams as much as possible.

Ultimately, teams that do this do have to deal with being shamed in public, both at the tournament, and afterward. Usually, this is enough to disparage people from doing this.

[Sun Devil Student]

Is this something that has happened in California (or elsewhere) in the past?

It seems fairly simple to prevent conspiring between teams from the same school by making sure they play each other before they play "opposing" teams from different schools. If teams from the same school advance to the same playoff bracket, then within that bracket, reassign the team numbers such that they play their colleagues first.

While it's fairly easy to throw a game without being obvious, it doesn't seem like a big problem if it's a team wanting to leave early or be ranked lower for whatever reason. If a team doesn't want to win, then they won't win, period. (Unless both teams are trying to lose in the same round, which is very unlikely given that 99.99% of all teams are definitely trying to win at all times.)

There's no reason to require the A team to be the best one, nor any way to enforce such a requirement. If a school is willing to risk having neither team qualify (by weakening the A team in order to improve the B team), then it's their choice to gamble. If this school's teams are so good that they can afford to weaken their A team like this and still be nearly certain of victory, then that school probably deserves both qualification slots anyway.

[Mechanical Beasts]

There's no reason to require the A team to be the best one

Bracketing.

nor any way to enforce such a requirement.

Mostly true, except when you have decent knowledge of the teams attending.

[MahoningQuizBowler]

NAQT Rule K1 says "All players, coaches, institutional representatives, and other persons associated with a team are bound by an honor code to behave responsibly and ethically. This includes, but is not limited to: ... not colluding with another person to "fix" a match result, not intentionally "throwing" a match".

[Duncan Idaho]

To me, the scenario of a team attempting to throw a game seems to be most real at HSNCT. For a pretty good team which could make the top fifteen or twenty five teams in the country, but not the top eight, there's a very real incentive in a card system to lose an early game, like the third or fourth prelim match, so that such a team could spend more of their middle games playing easier or middling opponents rather than really hard ones. In this way, such a team could more easily guarantee that it goes at least 7-3 and gets the double-elimination benefit that 6-4 teams don't get.

One could argue that this is counter-productive, because this hypothetical team would eventually have to play those top teams anyway in playoffs, but this seems like a strong incentive to lose deliberately.

[Sun Devil Student]

There's no reason to require the A team to be the best one

Bracketing.

nor any way to enforce such a requirement.

Mostly true, except when you have decent knowledge of the teams attending.

But if you know the teams well enough to tell that their A team isn't their best possible A team (say, by having stats from earlier performance), then you surely also know enough to seed the weaker A team and stronger B team accordingly.

One possible way to bracket when sufficient knowledge about team abilities is lacking would be to have the top team (or two) from each preliminary bracket advance, followed by wildcards in order of bonus conversion or some other non-opposition-dependent statistic so that a stronger preliminary bracket sends more teams to playoffs and fewer to consolation. However, in the situation you're discussing, you have enough knowledge of the teams that you might not have to do that.

[cchiego]

But if you know the teams well enough to tell that their A team isn't their best possible A team (say, by having stats from earlier performance), then you surely also know enough to seed the weaker A team and stronger B team accordingly.

Ah, but what if they decide to switch as they go to their rooms on the morning of the tournament after they see the brackets? I'm afraid you're underestimating the potential deviousness/cleverness of high schoolers trying to get to Atlanta/Evanston. That way, the weaker B team would be seeded like a strong A team and thus not have to play as many strong opponents initially and possibly getting to a higher playoff bracket.

To me, the scenario of a team attempting to throw a game seems to be most real at HSNCT. For a pretty good team which could make the top fifteen or twenty five teams in the country, but not the top eight, there's a very real incentive in a card system to lose an early game, like the third or fourth prelim match, so that such a team could spend more of their middle games playing easier or middling opponents rather than really hard ones. In this way, such a team could more easily guarantee that it goes at least 7-3 and gets the double-elimination benefit that 6-4 teams don't get.

I wonder if teams have done this in the past (without anyone being able to tell, of course)?

NAQT Rule K1 says "All players, coaches, institutional representatives, and other persons associated with a team are bound by an honor code to behave responsibly and ethically. This includes, but is not limited to: ... not colluding with another person to "fix" a match result, not intentionally "throwing" a match".

Smart teams would do it, but not tell anyone about it. And if people found out about it or suspected it, what could be done? It's almost all hearsay.

[Sun Devil Student]

But if you know the teams well enough to tell that their A team isn't their best possible A team (say, by having stats from earlier performance), then you surely also know enough to seed the weaker A team and stronger B team accordingly.

Ah, but what if they decide to switch as they go to their rooms on the morning of the tournament after they see the brackets? I'm afraid you're underestimating the potential deviousness/cleverness of high schoolers trying to get to Atlanta/Evanston. That way, the weaker B team would be seeded like a strong A team and thus not have to play as many strong opponents initially and possibly getting to a higher playoff bracket.

Wow, I didn't even think of that. Has any team actually tried this kind of thing? It would be funny if it wasn't so unfair to the teams trying to qualify properly following the rules. I'd think the statkeeper can easily catch teams trying that trick though, since the B team's names will be on the scoresheet that's supposed to be from the A team vs. whoever their first-round opponent is, and then the statkeeper will see that the names don't match the ones he got from the A team at registration (and that were used for seeding). Then you fix it like you would fix any accidental out-of-bracket match (how do you fix those anyway... do you just replay the matches during lunch or something?) and you don't even have to prove it was intentional or accuse anyone. In fact, you could just assume it was a mistake unless you know for sure otherwise.

This doesn't stop people playing under false identities, but that's beyond the scope of this discussion.

[Carambola!]

Yes, during a recent tournament where we sent 5 teams, our A and B teams were bracketed together so B team played as D team in order to avoid that.

[cchiego]

Then you fix it like you would fix any accidental out-of-bracket match (how do you fix those anyway... do you just replay the matches during lunch or something?) and you don't even have to prove it was intentional or accuse anyone.

So basically, you have to add an entire extra round to the tournament to fix that. Too costly; not going to happen. It's a fait accompli by then.

Yes, during a recent tournament where we sent 5 teams, our A and B teams were bracketed together so B team played as D team in order to avoid that.

That makes sense if it's a TD who didn't realize that MSJ B was that good (and it's kind of understood that TDs will try to spread out A-n teams from the same school across brackets). It wouldn't make sense for the actual B team though if they were trying to underhandedly advance in the rankings since the D team's schedule would presumably be more difficult, so it sounds like y'all were fine with what you did.

[Crimson Rosella]

Yes, during a recent tournament where we sent 5 teams, our A and B teams were bracketed together so B team played as D team in order to avoid that.

Ah, that makes sense now.

[Steeve Ho You Fat]

But if you know the teams well enough to tell that their A team isn't their best possible A team (say, by having stats from earlier performance), then you surely also know enough to seed the weaker A team and stronger B team accordingly.

Ah, but what if they decide to switch as they go to their rooms on the morning of the tournament after they see the brackets? I'm afraid you're underestimating the potential deviousness/cleverness of high schoolers trying to get to Atlanta/Evanston. That way, the weaker B team would be seeded like a strong A team and thus not have to play as many strong opponents initially and possibly getting to a higher playoff bracket.

Wow, I didn't even think of that. Has any team actually tried this kind of thing? It would be funny if it wasn't so unfair to the teams trying to qualify properly following the rules. I'd think the statkeeper can easily catch teams trying that trick though, since the B team's names will be on the scoresheet that's supposed to be from the A team vs. whoever their first-round opponent is, and then the statkeeper will see that the names don't match the ones he got from the A team at registration (and that were used for seeding). Then you fix it like you would fix any accidental out-of-bracket match (how do you fix those anyway... do you just replay the matches during lunch or something?) and you don't even have to prove it was intentional or accuse anyone. In fact, you could just assume it was a mistake unless you know for sure otherwise.

This doesn't stop people playing under false identities, but that's beyond the scope of this discussion.

At an Ohio high school tournament last year, a team did this. It was in our local format, so they didn't qualify for HSNCT/NSC from it, but they did qualify for our local format's regionals. At the time, I remember the TD posted on our boards after someone pointed out how much better, statistically, the B team was doing than the A team, that they had indeed done this but he hadn't realized it until round 3, at which point it was too late for him to do anything about it. It's easier to do in our format though because individual statistics aren't kept there.

[Stained Diviner]

To me, the scenario of a team attempting to throw a game seems to be most real at HSNCT. For a pretty good team which could make the top fifteen or twenty five teams in the country, but not the top eight, there's a very real incentive in a card system to lose an early game, like the third or fourth prelim match, so that such a team could spend more of their middle games playing easier or middling opponents rather than really hard ones. In this way, such a team could more easily guarantee that it goes at least 7-3 and gets the double-elimination benefit that 6-4 teams don't get.

One could argue that this is counter-productive, because this hypothetical team would eventually have to play those top teams anyway in playoffs, but this seems like a strong incentive to lose deliberately.

If the card system is set up well, then it is always to your advantage to win, barring some upsets that you wouldn't know about while you were playing your match.

[Charbroil]

ALSO should teams be required to make their A teams their best, so that TDs can bracket correctly? What if they split their A team on to B so that both qualify for nationals? etc etc

Is there a problem with this as long as the TD knows what's happening and can bracket accordingly? We've done this a couple of times at WUSTL this semester to give everyone a better chance of both answering questions and doing reasonably well, and I don't really see anything particularly illegitimate about it.

[cvdwightw]

ALSO should teams be required to make their A teams their best, so that TDs can bracket correctly? What if they split their A team on to B so that both qualify for nationals? etc etc

Torrey Pines was a notable offender of this practice back in the early 2000s. The story I heard from their then-coach was that they just found however many players were interested, split them up into teams in the car, and decided which letters went with which teams before getting to the tournament. This led to at least two tournaments my senior year that had teams freaking out that "Torrey Pines B is actually their A team!"

And actually, it is very hard to detect a team "throwing" a game. All it takes is sitting on a few questions that they know the other team will eventually get (and then expressing frustration that they were "sitting"), negging a few questions with a logical but wrong answer (or even stalling, as if they're hard-pressed to come up with the answer), and/or intentionally picking the wrong one in bonus parts where a team with decent knowledge could have narrowed it down to two answers. All three of these things are legitimate brainfarts performed all the time, even in the same game, by reasonable players intending to win. Teams that "throw" a game by negging a bunch of times, or whatever, are usually teams that are frustrated with the questions and likely wouldn't win the game anyway.

[Duncan Idaho]

To me, the scenario of a team attempting to throw a game seems to be most real at HSNCT. For a pretty good team which could make the top fifteen or twenty five teams in the country, but not the top eight, there's a very real incentive in a card system to lose an early game, like the third or fourth prelim match, so that such a team could spend more of their middle games playing easier or middling opponents rather than really hard ones. In this way, such a team could more easily guarantee that it goes at least 7-3 and gets the double-elimination benefit that 6-4 teams don't get.

One could argue that this is counter-productive, because this hypothetical team would eventually have to play those top teams anyway in playoffs, but this seems like a strong incentive to lose deliberately.

If the card system is set up well, then it is always to your advantage to win, barring some upsets that you wouldn't know about while you were playing your match.

Perhaps I'm just being dense this afternoon, but this isn't immediately obvious to me that this is true at HSNCT, where the first three prelim losses don't carry on to playoffs. Is this because, by throwing a game in the prelims to face easier teams and accrue fewer losses, a team would wind up with a lower playoff card, and have to play tougher teams on Sunday, thus only delaying the inevitable?

One concrete example of a loss benefiting a team in a card system happened to me in 2009. At the second card system tournament I ever played, Dorman's Season Ender of my junior year, we played four or five prelim rounds with cards, then were rearranged into playofff brackets. After winning our first two or three games, we lost to James Island, who took our (higher) card. However, after the remaining prelim rounds, despite our loss to James Island, we controlled the number 1 card, and James Island possessed the number 2 card, even though they were undefeated.

After discovering this, Mr. Huff ranked us number two for whatever reason, probably legitimately. But had he not done that, that loss we took to James Island would have still benefited us by giving us the number 1 card. Is this a failure of the particular system he used, or an example of when a loss can be beneficial in a card system tournament?

[Stained Diviner]

Sorry for taking this conversation afield, but what probably happened in Ben's case is that you had four rounds with more than 16 teams or five rounds with more than 32 teams, and after losing you played a team with a perfect record, probably a 4-0 team vs your 3-1 team in the fifth round. There are some oddities in card systems, especially when the number of teams is not a power of two. By losing, your team did not get any advantage, since the card number did not matter in the end, and it is unlikely that you would have finished worse than 4-1 had you won the match. I do not know enough about that system to determine whether or not it is reasonable.

[grapesmoker]

A team intentionally losing a game in order to improve someone else's (say, their B team's) chances of making the playoffs/qualifying for something is a clear violation of ethics; if, as TD, I detected something like this happening and could prove it, that team would be in huge, huge trouble with me. As for dividing your club into A, B, and C teams, the labels don't matter but I would say you are informally obligated to commit to a particular line-up. That's what I'm asking people to do when registering for Nationals; if the line-up changes in a significant way, you'd better let the TD know.

Speaking as a player, I only once encountered a team that was intentionally playing to lose, and it was one of the most unpleasant quizbowl experiences I've ever had. I'm not sure I could come up with a better penalty in that situation than just being shamed by the community for what I think is seriously unsportsmanlike conduct.

[crimsonscholar]

I can speak of only one occasion where a team tried to throw a match, to echo Jerry's sentiments, it was unpleasant and quite annoying. The match was after lunch, it was against a high school team at a modified ACF tournament, and it was just plumb pathetic. There was also a time that I remember about a team wanting to throw a match against their institution's B team, since they had one national bid in the bank and wanted to give the B team a shot at a second team to make it. Thankfully, this did not happen.

It is better to make it clear about A, B, C, and so on teams. Better for stats, better for the TD. There are other reasons not to be willy-nilly with lineups, but in general, it is unsportsmanlike no matter what to throw a match, though sometimes it is not clear.

[Captain Sinico]

Let me respond to all the points I'm seeing here:
1. Like Jerry et al. are saying, a team throwing a game (tanking) is unethical and a team having been caught doing so should suffer a penalty assuming your tournament has published rules against tanking. If your tournament doesn't have such a rule, here's incentive to create one.
Coach Reinstein is right that tanking in a well-constructed, well-seeded card system is a bad strategy in general as it will cause a team to play a harder schedule on average. To carry the analysis a bit further, even when tanking appears to be a good strategy, be it because of a poorly designed or poorly seeded card system or because a team happens to have hit upon such a case in a good system, tanking is still in general a non-optimal strategy unless a team already knows which other teams are tanking, which is not in general the case. That is to say, tanking can easily be neutralized or even backfire very badly through other teams tanking.
I'd say in summary that tanking can be in general very difficult to detect but, because of its non-optimality and the rareness of cases in which it will reliably pay off even if a team assumes no other teams are tanking, it will always be rare and self-correcting, so it shouldn't be something you need to worry much about as a TD.
(You can add to all the claims here: "if any only if you've designed a good tournament format and seeded it well.")

2. Also like Jerry is saying, labels shouldn't matter, but rosters in a seeded tournament do and must matter. Seeding must be based on pre-submitted, fixed rosters that teams understand they can't change without permission.
A tempting seeming way around this issue is to enact rules along the lines of "The team with highest label must be the best team." However, such rules are ill-advised for a variety of reasons. To take the most obvious example, such rules are unenforceable in the sense that what comprises a violation is in general difficult even to define, much less to detect. That is, it's difficult-to-impossible to determine which team is best a priori, even assuming we define "best" in an easily calculable way and have access to relevant data. To be more specific, let's suppose a tournament has a rule stating "If a program's B team finishes higher than its A team, it forfeits $1000 and we will make fun of it on the internet." If the tournament is otherwise well designed (so that no B team will have a much easier path to a high finish its corresponding A team,) it's overwhelmingly likely that such a rule will punish only programs who happen to have A and B teams that are close in quality.

3. A well-designed seeding system should on average punish teams that switch labels (after seeding, I mean.) The easier schedule gained by the formerly lower seed is in general more than counterbalanced by the harder schedule taken on by the formerly higher seed. That's because a program in general should optimize its chances of winning the tournament, which means its best team winning in the overwhelming majority of cases.
Further, label switching is a non-optimal strategy in the same sense as above: in general, even in cases in which label switching would otherwise pay off, it can be undone or even backfire badly through other label switches. Label switching is only a good strategy in general if a team already knows whether teams are switching, which is not in general the case.
It is tempting to say, those things being the case, that is it not necessary to make rules against label switching. Unfortunately, that is definitely not the case - label switching happens in spite of those facts, so they do not comprise a sufficient deterrent.
It is also tempting to say that label switching only affects (and, indeed, hurts) the switching teams, so no rule against it should be necessary. This is manifestly untrue in bracketed systems - the other teams in the brackets containing the switched teams are likewise impacted.

Actually, that brings me to a related point. A good tournament system should be robust to small deviations in seed, since there will always exist not inconsiderable errors in estimates of team quality, even given copious data. I suspect that card systems are poor by this criterion, i.e. less robust to seeding errors than, say, snake-seeded pooled round robins. Unfortunately, I can't make any quantitative estimates at this point, so any further discussion is appreciated.

M

[Important Bird Area]

Actually, that brings me to a related point. A good tournament system should be robust to small deviations in seed, since there will always exist not inconsiderable errors in estimates of team quality, even given copious data. I suspect that card systems are poor by this criterion, i.e. less robust to seeding errors than, say, snake-seeded pooled round robins. Unfortunately, I can't make any quantitative estimates at this point, so any further discussion is appreciated.

M

Shouldn't the distinction between card systems and round robins depend on the number of games played and total number of teams involved? (It is NAQT's impression- likewise backed with no quantitative data at this point- that, say, balancing 20 10-team brackets for the HSNCT would in fact be much more vulnerable to seeding errors than the existing card system, in which the seeds have very little impact after the first 4 or so rounds.)

[Stained Diviner]

There are some cases where label switching can be advantageous. If you figure that your best team is going to make the top bracket no matter what, but your second best team is going to need an advantageous schedule to make the top bracket(, and you don't care about fairness because you're a jerk), then you should switch team labels. I don't think that this is a particularly rare scenario (except for the parenthetical part).

[manary]

Most tournaments I have run or staffed are of the following structure:

Field is divided into group (say 18 teams in 3, 6 team groups). If there are a total of say 6 really good teams, we try to put two good teams in each bracket. We just distribute the good teams evenly among the groups.
Each group of 6 plays a 5 game round robin, then the top two teams from each group advance to a new top group of 6, the middle to to a middle 6, and bottom to bottom 6. Then these groups play another full round robin, and are ranked in order in their group.

Thus, if we don't think a B team will be good, we will put it in a group with 2 other 'good' teams. And if the A team is not as good, then that group will be easier.

Is this just a terrible way to structure things? Should we be advancing based on some other number like PPB, rather than the top 2 teams from each group? It seems easily game-able.

Micah

[grapesmoker]

There are some cases where label switching can be advantageous. If you figure that your best team is going to make the top bracket no matter what, but your second best team is going to need an advantageous schedule to make the top bracket(, and you don't care about fairness because you're a jerk), then you should switch team labels. I don't think that this is a particularly rare scenario (except for the parenthetical part).

If I found a team doing this in a tournament I was directing, I would forcibly switch the labels on the schedules back. A team that further attempted to cause trouble would simply be expelled. I don't tolerate fools gladly, but I tolerate unethical behavior even less.

[Skepticism and Animal Feed]

Having been on a "Chicago F" team that won Michigan MLK (some guy named Yaphe was on it with me), I'm amused that anybody would put any value at all on what a team decides to call itself and does not instead push for rosters. People do wacky things with team names, especially in college.

Even if a team is honest and ranks its teams with A being the best, you have no way of knowing how good they are compared to other teams in the field. Random State B is worse then Random State A, but how does it compare to Random Tech A?

[cvdwightw]

Actually, that brings me to a related point. A good tournament system should be robust to small deviations in seed, since there will always exist not inconsiderable errors in estimates of team quality, even given copious data. I suspect that card systems are poor by this criterion, i.e. less robust to seeding errors than, say, snake-seeded pooled round robins. Unfortunately, I can't make any quantitative estimates at this point, so any further discussion is appreciated.

M

Shouldn't the distinction between card systems and round robins depend on the number of games played and total number of teams involved? (It is NAQT's impression- likewise backed with no quantitative data at this point- that, say, balancing 20 10-team brackets for the HSNCT would in fact be much more vulnerable to seeding errors than the existing card system, in which the seeds have very little impact after the first 4 or so rounds.)

As discussed somewhere in this thread, it's hard if not impossible to get any quantitative data, since we don't have the necessary controls (exact same teams on exact same packets, using two different systems, and varying the exact parameters differently). With a reasonable approximation of the probability that each team in the tournament beats each other team (and the appropriate schedules), I can run Monte Carlo simulations of the relevant card/bracket systems with the relevant tweaks. The problem is figuring out what is "reasonable" when making the probability matrix, especially given how little data exists and how noisy that data is.