Early losses in power-matching = advantage?

quizbowllee · Post by **quizbowllee** » Sun May 27, 2007 8:42 am

Brindlee Mountain A went 5-5 after starting out 4-0.

Brindlee Mountain B, however, went 6-4 and made the playoffs.

It is VERY obvious to me now that losing an early round makes it much easier to make the playoffs.

Small Schools Playoffs include:

Brindlee Mountain B
Brindlee Mountain A
Danville
George Mason
White Cloud

I think there are a couple of others as well.

btressler · Post by **btressler** » Mon May 28, 2007 11:43 am

quizbowllee wrote:It is VERY obvious to me now that losing an early round makes it much easier to make the playoffs.

I made a similar argument to R on Saturday night, and plan to back up my argument after we see some data. I felt our schedule caused us to be 2-4 spots lower than we were capable of, as we played several of the very top teams and only one of the teams within a few spots of us.

Most of the bracket is fair and gives you matches near what you deserve. But there are places in the standings that can help or hurt you. For example, let's say you probably deserve to be in the 6-4's. Get on a run and go 5-2 in your first seven, and you'll play the rest of the day against other teams who went 5-2, and somebody in that group has to sink to 5-5.

I'm curious as to the details of why a match had to be thrown out. The cards tell you exactly what two teams are supposed to be playing. Did somebody misread a card?

Byko · Post by **Byko** » Mon May 28, 2007 12:50 pm

So I've only glanced briefly at the statistics, and I think there's more credence to the idea of taking an early loss and having an advantage. We already saw heard how Brindlee Mountain A started 4-0 and ended up 5-5. Russell (KY) did the same thing. Following some quality early wins, the rest of the schedule was murderous:

Round 2: d. Hartland (4-6) 265-90
Round 3: d. Bloomington (7-3 playoff team) 225-210
Round 6: d. Saint Thomas (5-5) 255-175
Round 7: d. State College B (7-3 playoff team) 260-140
Round 8: l. Eden Prairie A (#6 seed) 255-285
Round 9: d. Danville (6-4 playoff team) 200-185
Round 10: l. Eden Prairie A (still #6 seed) 160-340
Round 12: l. Gonzaga (7-3 playoff team) 290-320
Round 13: l. Brookwood A (6-4 playoff team) 205-380
Round 14: l. Santa Monica A (#11 seed) 150-370

Now show me how this team was not worthy of a playoff spot. Compare this to (as an example) East Lansing B:

Round 1: l. Fred T. Foard (3-7) 95-120
Round 2: d. Dunham (4-6) 280-20
Round 4: l. Champlain Valley Union (5-5) 180-255
Round 5: l. Charter B (5-5) 120-160
Round 6: d. Orono B (3-7) 190-90
Round 8: d. Central Hardin (5-5) 145-125
Round 10: d. Mustang (4-6) 120-100
Round 11: d. Simon Kenton (5-5) 210-100
Round 13: d. DeLaSalle (6-4 playoff team) 275-165
Round 14: l. State College B (7-3 playoff team) 190-210

Again, I'm not picking on East Lansing B--after a rough start, they put together a nice run when they had to. But their schedule is clearly far easier than Russell's.

I'd hate to recommend to good-but-not-elite teams that want to make it into the playoffs at NAQT that they should intentionally try to take an earlier loss to help themselves out, but I think this should make it clear that perhaps this is not a fair system.

Gonzagapuma1 · Post by **Gonzagapuma1** » Mon May 28, 2007 1:23 pm

I also think that this is true and really hurt our playoff seeding. I could be wrong abvout this but i think that early byes also maek your schedule harder. Our schedule looked like this:
Simon Kenton
Auburn
DCC A
Dunbar A
MLK A
Santa Monica
MLK A(again)
Russel
Stuyvesant A
Maggie Walker A
I know a lot of schools had tough schedules but I thik that schedule is a killer especially because we beat DCC A and Santa Monica who ended up ranked higher than us and we lost to the 4 seed overall twice and the 1 seed overall once.

First Chairman · Post by **First Chairman** » Mon May 28, 2007 1:27 pm

I want to be careful when one says "advantage" because I want to know what you mean by that. Also, how early one takes a loss... I suppose if I had a huge spreadsheet I can figure out the big tree that gets someone into a final position.

Ideally though, powermatching should occur with a random first-round draw, and subsequent quasi-random placements afterwards among the teams still in contention. I have to lean on my colleague Andrew Feist who knows and is certified in the algorithm (and completed his math degree from Duke when I was there the last few years) for more details.

You also need to have the right comparator groups. Not that Bykowski hasn't done a good job, but if we were going to present a hypothesis that an early loss can help a borderline team get into the playoffs, we have to compare apples to apples when it comes to scoring productivity but differences in win-loss.

So, let's look at the top teams who were 5-5 and compare to similarly-productive teams who were 6-4. Let's also look at the results of their first four games and see if there is a pattern. Of course we're not looking at whom they played.

5-5
Novi (MI, 210.45): LWLW
Harker School (CA, 203.96): WLWL
Russell (KY, 202.23): WWWW
Brindlee Mtn A (AL, 195.95): WWWW
Central Gwinett (GA, 192.50): WWWL
Kellenberg Memorial (NY, 188.02): WLLW
Fisher Catholic (OH, 181.69): LWWW
Champlain Valley (CT, 177.46): LWWW
Trinity (FL, 175.88): WLWL
St. Thomas (MN, 174.77): WWLW
Henry Gunn (CA, 173.76): WWLL
St. Johns (TX, 173.25): LLWW

6-4
Shady Side (PA: 212.39): WLWW
Rancho Bernardo (CA: 212.22): WWLW
Madisonville North Hopkins (KY, 212.16): WWLW
St. Ignatius (IL, 211.26): WLWW
Bergen County (NJ, 206.22): LWWL
Maggie Walker B (VA, 200.00): WLLW
George Mason (VA, 200.00): LWWL
Livingston (NJ, 197.74): LWWW
Richard Montgomery B (MD, 195.50): WLLW
Wayzata (MN, 188.55): WLWL
Brindlee Mountain B (AL, 185.92): WWLW
Garfield Heights (OH, 184.02): LWWW
Parkview (GA, 183.19): LWLW
Pensacola (FL, 181.19): WWLL
Norcross (GA, 179.15): LWWL
Aiken (SC, 179.07): LWWL
Minnetonka (MN, 175.59): WWLL
Columbus Acad (OH, 174.06): WWLW
Ransom Everglades (FL, 173.54): WLWW
De La Salle (MN, 169.67): WLLW
East Lansing B (MI, 167.91): LWLW
Wayzata B (MN, 163.39): WLWL
Arcadia A (CA, 155.86): WWWL
Arcadia B (CA, 150.23): WLWW

I think what is most interesting is that the two teams that went 4-0 in this sample wound up out of the playoffs at 5-5. But overall, the vast majority seem to be performing near their anticipated record of 0.500 whenever they hit their loss. But yes, it's basically up to the selection criteria of who your opponents are that may set up the disparities, and not simply "when" you took a loss. For Russell, the repeat game is more punitive (and in my opinion is unfair given the number of teams in the field). For Brindlee Mountain, strength of matchup really changed between R4 and R5; I don't know how the pairings occurred after the first four rounds, but you can see that none of the other teams (other than perhaps Russell) equally suffered in the same manner but got into the playoffs.

So the algorithm error I think for those two anomalies is an issue of the way opponents for games 5-10 on their records if it is completely different from the way opponents were selected in games 1-4. Sure, you have to play everyone, but it is not the fault of the teams that the original four games may have boosted their points-for stats to a level that the placements for the next few rounds were so difficult.

quizbowllee · Post by **quizbowllee** » Mon May 28, 2007 2:22 pm

While I still think that there is some merit to this argument, I have to say that my A team had some type of fundamental change come over them after they played Maggie Walker A. We played an absolute barn burner against Maggie Walker. I thought we were actually going to beat them. We had them 170-85 at the half. But, MW rallied and won 300-200 after we fell apart in the second half. Their coach later told me that other than MLK, who did manage to beat them, that we were the toughest match they had.

However, a COMPLETELY different Brindlee Mountain A emerged from that match. We lost to TJ A 545-125 the next round, then beat Wilmington Charter B. After that, we couldn't buy a win, no matter who we played.

I KNOW that BMHS A is a much stronger team than BMHS B, but even in the small schools playoffs, BMHS A choked hard. They played a lackluster match and managed to beat White Cloud, then got the crap kicked out of them by George Mason. Then, when our A and B teams faced off, the B team knocked the A team out of contention by winning 315-60 (NAQT has this score wrong on their results).

In a way, I'm glad that others finally got to see what I have been saying about my team all this time. When they are on, they could beat ANYONE. We almost took down Maggie Walker. We're 4-1 against MLK Magnet this year. But, at every tournament, they get to a point where they just STOP. I know that it's psychological, but I don't know how to stop it! At least my B team was there to pick up the slack. They made it further into the playoffs than any BMHS team ever has (tied for 21st), and managed to at least come home with 2nd in small schools - which is where my A team was last year. I wish my A team (my real A team, not the lifeless zombies that emerged after Round 5) could've had a shot at Danville. But, there's always next year. I've got one more year with this group to figure out how to stop this from happening.

Also, maybe we'll be more consistent at PACE.

Lee Henry
Coach of the most WILDLY inconsistent team in quiz bowl history

First Chairman · Post by **First Chairman** » Mon May 28, 2007 3:08 pm

I doubt that you are the most inconsistent team ever, but I'm sure that the psychology of the Maggie Walker match has merit. The kids will find a way to work with it as a learning moment, though obviously they have to pick themselves up and forget games. Similar to sports psychology.

MLWMathStar · Post by **MLWMathStar** » Mon May 28, 2007 9:57 pm

In VHSL competition (which admittedly has questions much less pyramidal) Freeman, who managed to beat a nearly full-strength MW A in an NAQT match this year, narrowly lost to our B team - who we sent to VHSL regionals while our A team cleaned up at Gonzaga. The very next game their opponent, who we had no trouble beating, tripled their score.

At the George Mason tournament, Gonzaga came away from a close loss against State College to play against a less-than-full-strength MW B team. They ended up making several slipups during that game (e.g. "Glass Palace" for "Crystal Palace" or something like that) and lost by 5 points. There is no explanation except for losing a close game just before. Gonzaga is far better than us.

My point is that even the best teams can have this demoralizing effect from losing a tough game, or watching an upset slip away from them. If this lasts for 4 or 5 games, or even into the next day, obviously something is wrong. But it might not even be that having a one-game slump after a close game is a weakness; instead, being able to bounce back should be considered a strength.

centralhs · Post by **centralhs** » Tue May 29, 2007 11:37 am

I'm not sure exactly how I feel about this whole issue of "Early Losses=Advantage", but our school (Central Gwinnett) followed a similar path to Brindlee Mountain A. We started out a strong 4-1 and then lost 4 games in a row to end up 5-5. Our point total of 2310 was higher than 19 of the 6-4 teams that made playoffs. Certainly I believe that our team and the top few teams to end up 5-5 were stronger teams than some of the 6-4 teams.

Oh well... what can you do?? This was our first trip to Nationals and we had a great time anyway!

First Chairman · Post by **First Chairman** » Tue May 29, 2007 11:48 am

There is something that can be done, but it would be general sacrilege to many people. Basically guarantee all teams based not only on W-L but also on ppg until you fill a double elim field. If you stated a guarantee of all 7-3 teams and then all other teams who have PPG's within a certain range of the weakest of the 7-3 teams' PPG's, you might get a "fairer" answer. That means say 20 teams would get automatic tickets, and the remaining teams would be under consideration for the last 44 spots.

That means this year 27 teams would be automatically in the field as 7-3 teams, and so 37 teams get in on PPG. Granted, this assumes some level of expedient stats crunching... so this may be completely untenable.

rchschem · Post by **rchschem** » Tue May 29, 2007 12:06 pm

Now I'd like to see someone actually take a dive in the first round and see how it plays out. Put one's money where one's mouth is.

I don't know much about the math, but in my mind it seems that this system does little to discriminate well among the 6-4 and 5-5 teams. If you're good, you'll stay in that upper pool if you lose consistently, you'll likewise stay on the bottom. But if you alternate winning and losing then it's tough to get out of those doldrums. So how does the signal of losing early to get a better path compare to the noise of the system?

BobGHHS · Post by **BobGHHS** » Tue May 29, 2007 12:58 pm

We didn't intentionally lose to Stuy A in the first round, but after that we rattled off 6 in a row to qualify for the playoffs at 6-1 already... (we would later lose 3 in a row)... In those 6 in a row, we beat Dorman A and Shadyside both... so it really didn't give us much of an easier schedule after the first 2 or 3 rounds...

So I'm not sure if losing early benefits those teams or not...

rchschem · Post by **rchschem** » Tue May 29, 2007 1:08 pm

BobGHHS wrote:We didn't intentionally lose to Stuy A ...

I apologize if I implied that you did. My point was that despite these perceived advantages, I think there is too much integrity out there for anyone to really take advantage of such a weakness in a system.

Raleigh Charter did, however, take a dive against MW A in the second round of the playoffs. How else do you explain 710-30?

They offered us many Extreme Pizzas.

Eric

Sen. Estes Kefauver (D-TN) · Tue May 29, 2007 1:15 pm

God, I can't imagine how much TJ payed Walton then...

Anyway,
NKC on Saturday did definitely get an advantage by losing early to inferior teams (Pensacola and Shady Side), although neither loss was intentional by any means. We still had some very quality opponents, especially Santa Monica, but we weren't up against MLW etc.

I guess we wore out that karma on Saturday, because on Sunday I really felt like we got punished for winning early. We beat Ransom Everglades and State College B, and then did what I thought was unthinkable and edged out Dorman in round 3. Then we played Maggie Walker and obviously lost badly. Then we played Wilmington Charter A and got a loss that I wasn't satisfied with, as it seemed like they won on trash and a hose (I won't get more specific so no one cheats). Dorman meanwhile worked the loser bracket into 4th place, and Charter tied for 5th. I felt like we deserved a mid 10s trophy, and if we had lost to Dorman I think we would have gotten a higher rank. That's my gripe about this.

Lapego1 · Post by **Lapego1** » Tue May 29, 2007 2:15 pm

charlieDfromNKC wrote:Dorman meanwhile worked the loser bracket into 4th place, and Charter tied for 5th. I felt like we deserved a mid 10s trophy, and if we had lost to Dorman I think we would have gotten a higher rank. That's my gripe about this.

Keep in mind that Dorman didn't have that easy of a road to 4th. They did have to work their way in the playoffs through Santa Monica and Richard Montgomery, who was like the 3rd seed at the time.

Sen. Estes Kefauver (D-TN) · Tue May 29, 2007 2:18 pm

Yes, but we also beat Santa Monica the day before. My point is that I think we were about as competitive as most teams in the top 10.

standrews · Post by **standrews** » Tue May 29, 2007 2:54 pm

E.T. Chuck wrote:If you stated a guarantee of all 7-3 teams and then all other teams who have PPG's within a certain range of the weakest of the 7-3 teams' PPG's, you might get a "fairer" answer.
[snip]
That means this year 27 teams would be automatically in the field as 7-3 teams, and so 37 teams get in on PPG. Granted, this assumes some level of expedient stats crunching... so this may be completely untenable.

Dr. Chuck,

Looking at the aggregates for each round there was as much as a 102 point difference between round 8 (315 ppg) and round 13 (416.98 ppg). If there are no byes, this is not a problem. But a team with a bye in round 13 would quite conceivably have more trouble scoring points that a team with a bye in round 8. In the stat crunching, some consideration might need to be given to the difficulty of the questions in each round as reflected in mean ppg per round.

Bigfoot isn't the pr · Tue May 29, 2007 3:40 pm

I was on (Wilmington) Charter's A team and we lost our second (or maybe it was third) match on Sunday. The point is whoever had beaten us (I think it was a Michigan team) went on to play the Maggie Walkers (A and B if my information is right). That loss had actually saved us from being knocked out so early and allowed us to go farther in the tournament. While I don't know if this is fair or not, that loss completely saved us. Whether that was an intentional part of this scheduling system or not, I can't say, but I think that bit of information further supports the "taking a loss early helps" hypothesis

cvdwightw · Post by **cvdwightw** » Tue May 29, 2007 6:36 pm

rchschem wrote:Now I'd like to see someone actually take a dive in the first round and see how it plays out. Put one's money where one's mouth is.

I'm sure this is only marginally relevant, as the year I am referencing is 2003, when there were only 62 teams (64 signed up and 2 didn't show) and 24 playoff teams (exactly 3/8 of the expected with 64 signed up).

My senior year we (Capistrano Valley) took a close, arguably fluke loss, 310-300 to State College B, an eventual 5-5 team that didn't hit 300 the rest of the day. We worked our way back up to get to 6-2 and a match with top seed MLW A (who went 10-0 that year in the prelims) in our ninth game. That was the only team with a better record than us that we played, although we did play playoff teams in each of our last 6 matches (three of which ended up tied for 8th with us after Saturday). We finished Saturday at 7-3 with the 7 seed based on PP20H going into the Sunday playoffs.

Sunday morning, I wasn't feeling well at all, and that probably contributed to a close loss to Noah's Solon team in the first round. As the highest seed to fall, we then got the lowest seed in the first elimination match (Will Turner's Petoskey team), then beat St. Andrew's before falling in overtime to eventual third place St. John's.

So we managed to get a t-8th with an overall record of 2-3 against teams with the same or higher finish. Of course, that year employed a hockey/football playoff system (highest seed takes lowest seed regardless), not a basketball-like system (e.g. winner of 1-8 plays winner of 4-5 regardless of outcomes, in other words, a card-swapping system), and I'm not sure which of those NAQT currently uses.

Stained Diviner · Post by **Stained Diviner** » Tue May 29, 2007 7:14 pm

NAQT switched to card swapping in the playoffs this year. It made things more efficient because teams could go directly to their next room rather than wait around the bulletin board to find out where their next match was. It also made the seeds from Saturday a little less relevant because you changed your seed whenever you were involved in an upset on Sunday.

Byko · Post by **Byko** » Tue May 29, 2007 10:24 pm

ReinsteinD wrote:NAQT switched to card swapping in the playoffs this year. It made things more efficient because teams could go directly to their next room rather than wait around the bulletin board to find out where their next match was. It also made the seeds from Saturday a little less relevant because you changed your seed whenever you were involved in an upset on Sunday.

To add to this, in some ways, the card system was more fair to lower seeded teams. If you got into the playoffs as a low seed (e.g., Bentley last year as something like a #39 seed) and won your first game, you got to play someone near the #1 seed next. If you managed to beat them (like Bentley did last year in defeating Grosse Pointe North), you got to play the #2 seed next. So the end result is that it gives way too much of an advantage to the higher seeds and way too harshly penalizes the lower seeds.

I liked the card system this year because it was definitely more efficient time-wise than last year and it seemed to balance things out a little more. It mattered a little less this year whether you just barely got into the playoffs or made it in easily.

BuzzerZen · Post by **BuzzerZen** » Tue May 29, 2007 10:51 pm

Actually, it seemed to me that the playoff cards merely replicated a normal double elimination playoff bracket until there were 7 teams left.

As to the "early loss" hypothesis, my take is this: yes, an early loss could potentially make it easier for you to qualify. However, it could also potentially make it harder for you to qualify. Our loss to Eden Prarie in round 2 may well have made the difference for 5-5 vs. 6-4 to some team down the line. There is really no way of anticipating if a loss is an advantage of any kind since you can't predict when other teams will lose. It's still always better to win than to lose at any given time.

MLWMathStar · Post by **MLWMathStar** » Tue May 29, 2007 10:53 pm

This year's card system and the standard bracketing system have no technical difference, only a logistical one. As a team that made an upset in the first round, I can attest that Byko's example situation sort of happened to us this year as well. There was no real change.

First Chairman · Post by **First Chairman** » Wed May 30, 2007 12:48 am

standrews wrote:
E.T. Chuck wrote:If you stated a guarantee of all 7-3 teams and then all other teams who have PPG's within a certain range of the weakest of the 7-3 teams' PPG's, you might get a "fairer" answer.
[snip]
That means this year 27 teams would be automatically in the field as 7-3 teams, and so 37 teams get in on PPG. Granted, this assumes some level of expedient stats crunching... so this may be completely untenable.
Dr. Chuck,

Looking at the aggregates for each round there was as much as a 102 point difference between round 8 (315 ppg) and round 13 (416.98 ppg). If there are no byes, this is not a problem. But a team with a bye in round 13 would quite conceivably have more trouble scoring points that a team with a bye in round 8. In the stat crunching, some consideration might need to be given to the difficulty of the questions in each round as reflected in mean ppg per round.

Good point. The aggregate scoring however can also reflect the fact that more of the prolific offenses were also playing those rounds compared to the rest of the field, though that takes a lot more analysis. Especially given the larger number of games played per round, that explanation would be less likely.

Stained Diviner · Post by **Stained Diviner** » Wed May 30, 2007 6:06 am

This year's card system and the standard bracketing system have no technical difference, only a logistical one. As a team that made an upset in the first round, I can attest that Byko's example situation sort of happened to us this year as well. There was no real change.

There was a change. Your team was the lowest seed to win a first round match. Under the old system, that would have meant that your second match would have been against the highest seed--in other words, you would have played your A Team. Instead, you took over Bloomington's seed. You still had a tough 2nd round match-up, but it would have been tougher using last year's system.

Especially given the larger number of games played per round, that explanation would be less likely.

Their system purposely had teams that had won a lot of matches sit out during certain rounds and teams that had lost a lot of matches sit out during certain other rounds, so the law of large numbers does not apply. I personally thought that the final round (14) was difficult, but the aggregate score is above average. I think this is due to the fact that the 2-8 teams were not playing.