Favorite OOTS Character Tournament: General Discussion

[alwaysbebatman]

Here's everybody who was in the Tournament ranked by the ratio of votes they received against the total ballots in the contests they were in.

This makes Roy look great because he had an easy win in the first round, and a near loss in the second round prevented him from going on to face tougher competition.

Whereas Belkar, who lost the whole tournament by only one vote, is way down in 6th figured this way, because all of his matches were extremely close.

All of the top 6 received more votes for them than for their opponent over the course of their run in the tournament. Which to me says maybe there is something to learn from an extra round.

However, a full round robin of 6 competitors is 15 matches, and only 6 of those are already done, leaving 9 that would be required. Seven of which would feature Roy and/or Elan, which would get pretty repetitive.

I'm open to suggestions. Maybe the Swiss System would be the way to go? Please advise!

	Round 1	Quarter finals	Semi finals	Finals	All Sum
Roy Greenhilt	26 / 39 , 66.7%	25 / 51 , 49.0%			51 / 90 , 56.7%
O-Chul	24 / 35 , 68.6%	22 / 37 , 59.5%	33 / 67 , 49.3%	26 / 49 , 53.1%	105 / 188 , 55.9%
Redcloak	29 / 36 , 80.6%	24 / 46 , 52.2%	34 / 67 , 50.7%	41 / 81 , 50.6%	128 / 230 , 55.7%
Vaarsuvius	23 / 36 , 63.9%	24 / 32 , 75.0%	36 / 74 , 48.6%	23 / 49 , 46.9%	106 / 191 , 55.5%
Elan	22 / 36 , 61.1%	22 / 46 , 47.8%			44 / 82 , 53.7%
Belkar Bitterleaf	21 / 36 , 58.3%	26 / 51 , 51.0%	38 / 74 , 51.4%	40 / 81 , 49.4%	125 / 242 , 51.7%
The Monster in the Darkness	21 / 39 , 53.8%	15 / 37 , 40.5%			36 / 76 , 47.4%
Tarquin	18 / 39 , 46.2%				18 / 39 , 46.2%
Durkon Thundershield	19 / 42 , 45.2%				19 / 42 , 45.2%
Haley Starshine	23 / 42 , 54.8%	8 / 32 , 25.0%			31 / 74 , 41.9%
Shojo	15 / 36 , 41.7%				15 / 36 , 41.7%
Xykon	14 / 36 , 38.9%				14 / 36 , 38.9%
Wrecan	13 / 36 , 36.1%				13 / 36 , 36.1%
Blackwing	13 / 39 , 33.3%				13 / 39 , 33.3%
Therkla	11 / 35 , 31.4%				11 / 35 , 31.4%
Tsukiko	7 / 36 , 19.4%				7 / 36 , 19.4%

[Darth Paul]

The only thing that comes to me at this point is what my old drill sergeant used to tell me when I was running the track for PT: "Strive not to come in dead last."

[Jasdoif]

If you ran a fresh Swiss System tournament for all 16, that'd take four rounds and eight matches per round, for a total of 32 matches....

But the top grouping in a Swiss System tournament is effectively the same as a single-elimination tournament, so you could use the current matchup results instead of running the top bracket, cutting out 15 matchups and the entire first round...so you'd be down to three rounds and 17 matches. Or possibly 16 matches, if you want to keep the O-suvian Incident and it comes up.

More importantly(?), a Swiss System would mean everyone has been in exactly four matches, making it a lot easier to do valid comparisons all around.


Alternatively, if you only ran one for the 8 quarter-finalists, that'd be three rounds and four matches per round, for 12 matches...minus the 7 if you recycle the current matchup results, for 5 matches across two rounds. And then everyone would have been in exactly three matches for comparison data.

[alwaysbebatman]

I like that last suggestion with one caveat: it requires picking 8 competitors worth further attention, and only the top 6 are obvious.

Roy and Elan were both knocked out in the quarterfinals in very close matches by competitors who then went all the way to the finals. I don't think we have enough information to call Roy #1 as average votes or average MoV would do. But it's clear that he and Elan still appear competitive.

But for 7 & 8, especially 8, it is less clear. Tarquin has a significantly better average margin stat than Haley. But if you were to incorporate Strength of Schedule, that would certainly help Haley. Enough to put her in the top 8? I'm not sure.

So, there are two possible routes to take, for which I hope someone can help with one of these two:

1. Is there a non-elimination tournament system along the lines of the Swiss System that has the flexibility to work with 6 competitors? Any suggestions?
2. If not, I'd like to do a ranking that incorporates both MoV and SoS to determine who should be the 8 to include for the Swiss System tournament, such as an Elo-style ranking, or an SRS. But I have no experience with doing Power Rankings. Volunteers?

If anybody has ideas for a good non-elim tournament style that's less extensive than round robin, but doesn't have to have powers of 2 numbers of competitors like the Swiss System, we'll take that under consideration.

If we are going to do Swiss System, I definitely want to see a good Power Ranking to decide who should be in the top 8 and what seeding to use first. If somebody else can do that, that'd be really cool.

If nobody else has experience with Elo or SRS, I'll figure it out myself and move forward from there. It might take a couple days is all.

[Jasdoif]

I like that last suggestion with one caveat: it requires picking 8 competitors worth further attention, and only the top 6 are obvious.

Really? I kinda thought gauging the effectiveness of all 8 quarter-finalists would be the obvious choice, since you already ran some matches with them (and introducing competitors who hadn't competed in the quarter-finalists increases how many matches you'd need, which I thought you were trying to minimize).

Is there a non-elimination tournament system along the lines of the Swiss System that has the flexibility to work with 6 competitors?

You can run Swiss System with six. You'd have to run at least one match between competitors in adjacent scoring groups, but that's not unusual.

If not, I'd like to do a ranking that incorporates both MoV and SoS to determine who should be the 8 to include for the Swiss System tournament, such as an Elo-style ranking, or an SRS. But I have no experience with doing Power Rankings. Volunteers?

Well, most such rankings involve measuring prior history of competition...of which we have exceedingly little.

So I'm going to make up rating criteria instead

• The number of points they got in the pre-tournament poll, reflecting their ability to draw approval on their own
• The total number of votes they pulled in the tournament, reflecting their viability as a competitor in 1-on-1 matches and hinting at how far they got in the single-elimination tournament. (Notably, this excludes the O-suvian Incident, which occurred outside the single-elimination tournament, to avoid biasing results here)
• The minimum number of votes the matches they were in pulled, reflecting my best estimate of how well they draw crowds to 1-on-1 matches.

Since these are widely different scales, I'll be averaging them using a geometric mean; and rounding to the nearest tenth.

Competitor	Original Points	Total Votes For	Min Vote Matches	Rating
Redcloak	305	128	36	112.0
Belkar	216	125	36	99.1
Vaarsuvius	310	83	32	93.7
O-Chul	284	79	35	92.3
Roy	237	51	39	77.8
MitD	213	36	37	65.7
Elan	178	44	36	65.6
Haley	165	31	32	54.7
Durkon	146	19	42	48.8
Tarquin	131	18	39	45.1
Xykon	134	14	36	40.7
Shojo	98	15	36	37.5
Blackwing	64	13	39	31.9
Therkla	41	11	35	25.1
Tsukiko	8	7	36	12.6
Wrecan	2	13	36	9.8

(This is one of the fancy sortable tables; you can click on column headings to reorder the table)

[Liquor Box]

Really? I kinda thought gauging the effectiveness of all 8 quarter-finalists would be the obvious choice, since you already ran some matches with them (and introducing competitors who hadn't competed in the quarter-finalists increases how many matches you'd need, which I thought you were trying to minimize).

You can run Swiss System with six. You'd have to run at least one match between competitors in adjacent scoring groups, but that's not unusual.

Well, most such rankings involve measuring prior history of competition...of which we have exceedingly little.

So I'm going to make up rating criteria instead

• The number of points they got in the pre-tournament poll, reflecting their ability to draw approval on their own
• The total number of votes they pulled in the tournament, reflecting their viability as a competitor in 1-on-1 matches and hinting at how far they got in the single-elimination tournament. (Notably, this excludes the O-suvian Incident, which occurred outside the single-elimination tournament, to avoid biasing results here)
• The minimum number of votes the matches they were in pulled, reflecting my best estimate of how well they draw crowds to 1-on-1 matches.

Since these are widely different scales, I'll be averaging them using a geometric mean; and rounding to the nearest tenth.

Competitor	Original Points	Total Votes For	Min Vote Matches	Rating
Redcloak	305	128	36	112.0
Belkar	216	125	36	99.1
Vaarsuvius	310	83	32	93.7
O-Chul	284	79	35	92.3
Roy	237	51	39	77.8
MitD	213	36	37	65.7
Elan	178	44	36	65.6
Haley	165	31	32	54.7
Durkon	146	19	42	48.8
Tarquin	131	18	39	45.1
Xykon	134	14	36	40.7
Shojo	98	15	36	37.5
Blackwing	64	13	39	31.9
Therkla	41	11	35	25.1
Tsukiko	8	7	36	12.6
Wrecan	2	13	36	9.8

(This is one of the fancy sortable tables; you can click on column headings to reorder the table)

Since this is a new tournament why wouldn't you rely on the current rankings from the original thread?

[Jasdoif]

Since this is a new tournament why wouldn't you rely on the current rankings from the original thread?

Is it a new tournament? I was under the impression alwaysbebatman wanted the effect of retroactively adding rounds to the just-ended tournament.

[alwaysbebatman]

Thank you, jasdoif, that's really cool. Interesting that with your ranking system the Monster overtakes Elan.

However, I think that before moving on, I'd like to see (or likely, create,) power rankings that takes into account Strength of Schedule, ie: how tough was the competition. That may really change our perspective on rankings and cutoffs.

The tough part with that is avoiding circular reference errors. I presume limits, matrix algebra, or iterative calculation could be used. But I need a few free hours to really sit down with it and work it out. I might have time tomorrow evening?

[Jasdoif]

However, I think that before moving on, I'd like to see (or likely, create,) power rankings that takes into account Strength of Schedule, ie: how tough was the competition. That may really change our perspective on rankings and cutoffs.

I kinda think getting the data to make that kind of comparison is the point of moving on.

I mean, here, we have a measure of the strength of the competition: How many votes total were gained by whoever performed the knockout, with a special case for Redcloak getting his own vote count since he wasn't knocked out:

Name	Original Points	Total Votes For Self	Total Votes For Defeating Competitor	Min Vote Matches	Rating
Redcloak	305	128	128	36	115.8
Belkar	216	125	128	36	105.6
Vaarsuvius	310	83	125	32	100.7
O-Chul	284	79	128	35	100.1
Roy	237	51	125	39	87.6
Elan	178	44	128	36	77.5
MitD	213	36	79	37	68.8
Haley	165	31	83	32	60.7
Shojo	98	15	125	36	50.7
Durkon	146	19	31	42	43.6
Tarquin	131	18	36	39	42.7
Xykon	134	14	44	36	41.5
Blackwing	64	13	51	39	35.9
Therkla	41	11	79	35	33.4
Tsukiko	8	7	128	36	22.5
Wrecan	2	13	83	36	16.7

There's a little bit of shifting going on, but ultimately analyzing just one single-elimination tournament is supremely unlikely to veer from the results of that tournament: The champion, followed by the runner up, followed by the other semi-finalists, followed by the other quarter-finalists, followed by everyone else.

Getting more information about the non-winning competitors would make any strength estimations a lot more reliable...I mean, Tsukiko ran face-first into the champion (again); how much information is really there to glean? I thought you wanted ratings for that purpose, to help decide the exact matchups or seeding or something of the sort.

[alwaysbebatman]

Okay, I've been fiddling around with different options for taking into account Strength of Schedule (taking Expected margin from actual margin, and averaging the differences, and adding that averages to the average margin for one. I also tried a similar thing with Votes For and Votes Against.)

What I've learned: different methods will scramble the order of the competitors in different ways. (Ranked by un-adjustes margin, Roy is 1st as seen above. Adjusting by Expected Margin puts Belkar in 1st, and adjusting by expected votes for and against puts Tournament winner Redcloak back at #1.)

But it what it doesn't do is change the sign: six characters have a positive margin, the others lose more than they win.

If we take the existing rounds as the majority of a smaller non-elimination tournament, it only makes sense to me to make this tournament between these six: Belkar, O-Chul, Redcloak, Vaarsuvius, Elan, and Roy.

And the exact order doesn't look like it matters as much as I had thought. The thing about the Swiss system, nobody is eliminated, so each competitor needs the same number of matches with one another. These six have already had 6 matches among one another. Both Belkar and Redcloak have already had three matches with others in this group, the others have had one or two.

If we were to do a big Swiss system tournament of four matches for each competitor, it would take 9 more matches. Maybe a bit much?

Doing 3 matches per would be a lot more manageable, only requiring three more matches. Only downside is: Belkar and Redcloak would already be done. I know some posters were hoping to see them take on the oponents that they missed in their tournament path.

If we go with the small 3-match Swiss tournament, there are only two possible configurations for the three matches: Elan v Roy, and either O-Chul v Elan and Vaarsuvius v Roy; or O-Chul v Roy and Vaarsuvius v Elan.

Let me know what you think.

[Jasdoif]

Okay, I've been fiddling around with different options for taking into account Strength of Schedule (taking Expected margin from actual margin, and averaging the differences, and adding that averages to the average margin for one. I also tried a similar thing with Votes For and Votes Against.)

What I've learned: different methods will scramble the order of the competitors in different ways. (Ranked by un-adjustes margin, Roy is 1st as seen above. Adjusting by Expected Margin puts Belkar in 1st, and adjusting by expected votes for and against puts Tournament winner Redcloak back at #1.)

But it what it doesn't do is change the sign: six characters have a positive margin, the others lose more than they win.

If we take the existing rounds as the majority of a smaller non-elimination tournament, it only makes sense to me to make this tournament between these six: Belkar, O-Chul, Redcloak, Vaarsuvius, Elan, and Roy.

So your criterion is "competed against a finalist". Okay.

If we were to do a big Swiss system tournament of four matches for each competitor, it would take 9 more matches. Maybe a bit much?

Doing 3 matches per would be a lot more manageable, only requiring three more matches. Only downside is: Belkar and Redcloak would already be done. I know some posters were hoping to see them take on the oponents that they missed in their tournament path.

With six (or eight) competitors in a Swiss system tournament, you'd end up with a decisive single winner after three rounds anyway.

[alwaysbebatman]

The thing is, swiss system scoring is designed for chess, where there is no score, just win lose or draw.

Here, every character has votes for, votes against, margin, etc.

Based on Swiss System, we already have a decisive winner: Redcloak has three wins, everyone else has at least one loss.

Maybe it doesn't make sense to move forward with more contests without at least giving Redcloak a new challenge. (Maybe Belkar also.)

Here's a proposal: not exactly the Swiss system, sort of an unwieldy hybrid. But the advantage: we'd get new information about who is preferred head-to-head in pairings that people have expressed interest in, and we wouldn't be running anticlimactic bouts with the champion already determined and benched.

Do one round with Elan v Vaarsuvius and Roy v O-Chul, then a second round with Belkar v O-Chul, Redcloak v Vaarsuvius and Roy v Elan.

The top 4 would end with 4 matches and the other 2 would have 3 matches. It would be a ways from a full round robin but the most relevant matches would happen...

[Jasdoif]

Here's a proposal: not exactly the Swiss system, sort of an unwieldy hybrid. But the advantage: we'd get new information about who is preferred head-to-head in pairings that people have expressed interest in, and we wouldn't be running anticlimactic bouts with the champion already determined and benched.

Do one round with Elan v Vaarsuvius and Roy v O-Chul, then a second round with Belkar v O-Chul, Redcloak v Vaarsuvius and Roy v Elan.

The top 4 would end with 4 matches and the other 2 would have 3 matches. It would be a ways from a full round robin but the most relevant matches would happen...

So with your intention that the champion is not already determined and benched, and Vaarsuvius being the only one given a matchup against the champion....I couldn't fault StreamOfTheSky for seeing favoritism on your part.

[alwaysbebatman]

So with your intention that the champion is not already determined and benched, and Vaarsuvius being the only one given a matchup against the champion....I couldn't fault StreamOfTheSky for seeing favoritism on your part.

Ha! But on the other hand, V is the only one among the six who already has two losses. V fans may look at that as a final chance at redemption, but 2-2 is the best record V could possibly end with.

Hm. Then again, I can definitely see the argument for also running Redcloak v Roy and Belkar v Elan. Then the original tournament finalists would end with 5 matches and all the rest would end with 4 apiece.

ETA: I am spacey, I just realized that what I just proposed would only be two matches short of a full round robin! if we were going that far we may as well finish it!

[alwaysbebatman]

Here's a proposed schedule for a Round Robin tournament of the six most competitive characters, the six who had more votes cast for them than for their opponents.

Round	Open Voting	Declare Winner	Matchups	WINNER	Votes for Winner	Votes for Opponent	Margin
Quarter	2-Sep	9-Sep	Redcloak v Elan	REDCLOAK	24	22	2
Quarter	5-Sep	12-Sep	Belkar v Roy	BELKAR	26	25	1
Semi	9-Sep	16-Sep	Redcloak v O-Chul	REDCLOAK	34	33	1
Semi	12-Sep	19-Sep	Belkar v Vaarsuvius	BELKAR	38	36	2
Final	19-Sep	26-Sep	Redcloak v Belkar	REDCLOAK	41	40	1
Bronze	20-Sep	26-Sep	O-Chul v Vaarsuvius	O-CHUL	26	23	3
RR1	30-Sep	7-Oct	O-Chul v Roy				0
RR1	30-Sep	7-Oct	Vaarsuvius v Elan				0
RR2	3-Oct	10-Oct	Redcloak v Roy				0
RR2	3-Oct	10-Oct	Belkar v Elan				0
RR3	7-Oct	14-Oct	Roy v Elan				0
RR3	7-Oct	14-Oct	Belkar v O-Chul				0
RR3	7-Oct	14-Oct	Redcloak v Vaarsuvius				0

[LadyEowyn]

Aren't you missing a set of matchups? The table doesn't show O-Chul vs Elan, or V vs Roy.

Also, neat that the top characters form an X on the alignment chart: we have Lawful Evil, Chaotic Evil, True Neutral, Chaotic Good, and 2 Lawful Good.

[alwaysbebatman]

I have no idea how I missed that, thank you for the correction.

Nine more matches is too much, right?

Hm.

[Jasdoif]

Nine more matches is too much, right?

I don't see why. More specifically, I don't see why running three rounds of three concurrent matches would be notably more "much" than two rounds of two followed by one run of three. You'd be running three rounds either way....

[alwaysbebatman]

I don't like the idea of starting the same player in two matches at once. Seems both confusing and like over saturation.

However, if interest is there for a Round Robin, I'm willing to do it. If people don't feel like this whole thing is totally overkill, then a few more days added to the schedule to throw on one more wave of two matches is no big deal.

If it turns out that interest has faded now that the single elimination tournament is complete, we can reconsider.

Here is a corrected Round Robin table:

Round	Open Voting	Declare Winner	Matchups	WINNER	Votes for Winner	Votes for Opponent	Margin
Quarter	2-Sep	9-Sep	Redcloak v Elan	REDCLOAK	24	22	2
Quarter	5-Sep	12-Sep	Belkar v Roy	BELKAR	26	25	1
Semi	9-Sep	16-Sep	Redcloak v O-Chul	REDCLOAK	34	33	1
Semi	12-Sep	19-Sep	Belkar v Vaarsuvius	BELKAR	38	36	2
Final	19-Sep	26-Sep	Redcloak v Belkar	REDCLOAK	41	40	1
Bronze	20-Sep	26-Sep	O-Chul v Vaarsuvius	O-CHUL	26	23	3
RR1	30-Sep	7-Oct	O-Chul v Elan				0
RR1	30-Sep	7-Oct	Vaarsuvius v Roy				0
RR2	3-Oct	10-Oct	Redcloak v Roy				0
RR2	3-Oct	10-Oct	Belkar v Elan				0
RR3	7-Oct	14-Oct	O-Chul v Roy				0
RR3	7-Oct	14-Oct	Vaarsuvius v Elan				0
RR4	10-Oct	17-Oct	Roy v Elan				0
RR4	10-Oct	17-Oct	Belkar v O-Chul				0
RR4	10-Oct	17-Oct	Redcloak v Vaarsuvius				0

[LadyEowyn]

I'd put the last three matches of RR3 into an RR4. It's only one more round.

[alwaysbebatman]

Oh, I agree completely. I had them starting after, just forgot to change the round acronym in that column. Corrected now, thank you for pointing out the error!