[Strategy] IDing in a swiss tournament.

[aTn]

Assume a (swiss) tournament structure where the number of rounds is given by the DCI recommendations:

Number of players / number of rounds

4-8 / 3 rounds of Swiss
9-16 / 4 rounds of Swiss
17-32 / 5 rounds of Swiss
33-64 / 6 rounds of Swiss.

Here's the question I'm trying to answer:

As a function of the number of players, the round being considered, a player's record, etc., when is it the best time to ID (intentionnaly draw) a match (i.e. when does it maximize your chances to go to Top8, if there is one) ?

I guess it might be easier to answer the following question:

When should I not ID ?

I've been playing in tournaments for a while now but I can't say I'm very comfortable with my decision process when it comes to IDing.

I don't think there's a deterministic answer to any of these questions since they depend on the expected outcome of the decision made (to ID or not to ID), but I guess by amassing enough "experimental data" we can come up with guidelines for IDing.

I started this thread because I think it can benefit a lot of players (to the mods: sorry if such a thread already exists or if it's not posted in the appropriate subsection of this forum).

[Nihil Credo]

Without information about matchups, the best time to ID is as late as possible, since it has the best effect on your tiebreakers.

With more information there can be more interesting choices. For example, if my deck is a total glass cannon tuned against control, a round 1 ID can become a worthwhile gamble.

[aTn]

Thanks for your answer. Assuming you're playing in a relatively large Legacy tournament (33-64 players) and that you're not playing a glass cannon (e.g. Affinity), as a function of your record, at what round is it reasonable (or good) to ID ?

For example, if my record after 3 rounds is 3-0, should I ID round 4 or risk losing the current round (with no a priori knowledge of my opponent's deck) ?

[Nihil Credo]

If you have no idea what you're going to face then you should only ID when you already have all the wins you need to make the Top 8.

At a six round tournament that means four victories; though if you're close enough to 64 players you can fail to make the top 8 with a 4-1-1 record if your tiebreakers are bad enough, and maybe even with a 4-0-2.

You'll have to look at the standings and do the math for each tournament; either way, round 4 is definitely too soon to ID.

[klaus]

Can somebody post a solid advanced approach to this. I feel there's a lot more to learn than what's been said sofar.

[Frenger]

Is there a surefire way (aside from playing the match) to avoid getting 'leapfrogged' when IDing into the top 8? ex: you ID and end up in 9th and somebody behind you in the standings wins and makes it.

I almost always end up playing the match out because I'm scared of this happening.

Another rulings related question. Can you ask a judge for counsel regarding IDs, and what is a judge allowed to say if you do? Like if my opponent asks me to ID saying we'll both top 8, am I allowed to have a judge verify this?

[wmagzoo7]

Honestly, its all about your Tiebreakers which can change every single round. As everyone should know, a win is 3 points, a draw 1 and a loss is 0. It really depends on the number of rounds as a whole. I personally only ID if i'm in the top3-4 in a 4-5 round tournament and thats always in the last round. I also don't really believe in drawing early to try and play control decks because unless there is some behemoth of a deck that is so slow it can't finish a round or there are a million people drawing won't help you out and could screw you in the long run especially since it makes the tournament almost a single elim. tourney because if you lose again you may not be able to make top 8 but that's just me.
Edit: Judges usually will tell you to sit down and play if you spend too much time looking at standings etc. Most times they won't tell you because then they have to do math and that takes time :(. Best bet is to ask a friend who you trust to be intelligent to check for you.

[cdr]

The more rounds there are in a tournament, the less predictable tiebreakers get. This should be obvious since the more previous opponents you have, the more variables there are.

There's been a couple famous cases of tiebreakers changing dramatically in GPs, which are ~15 rounds of swiss. One I was there for was (the real) Dave Price being in 8th before the last round of day 2, winning his match, and then ending up in 9th (http://www.wizards.com/sideboard/article.asp?x=GPDEN01\768pricestunned).

The swiss round system is designed to guarantee anyone with an X-1 record or better to make a top 8 cut. This means you are always safe to draw the last round of swiss if you are undefeated and almost always safe two draw the last two if you're undefeated, but again this should be obvious.

I only found one relevant article (http://www.starcitygames.com/magic/fundamentals/3746_Double_or_Nothing_Breaking_the_Tie.html) on SCG, but I know the topic has been covered more than once before (probably on other sites). It does have a good bit of advice I had forgotten though: tiebreakers will generally only change about ±5% in the last round of swiss.

[quicksilver]

The more rounds there are in a tournament, the less predictable tiebreakers get. This should be obvious since the more previous opponents you have, the more variables there are.

There's been a couple famous cases of tiebreakers changing dramatically in GPs, which are ~15 rounds of swiss. One I was there for was (the real) Dave Price being in 8th before the last round of day 2, winning his match, and then ending up in 9th (http://www.wizards.com/sideboard/article.asp?x=GPDEN01\768pricestunned).

The swiss round system is designed to guarentee anyone with an X-1 record or better to make a top 8 cut. This means you are always safe to draw the last two rounds of swiss if you are undefeated (or draw the last if you are x-0-1 etc), but again this should be obvious.

Am I fake or something?

Also x-0-2 is less than x-1, so I don't understand that reasoning.

[cdr]

You haven't won a PT yet. You can be legacy Dave Price for now.

You're right, I slipped up - I edited my post.

[Forbiddian]

I believe X-0-2 always makes it in as well no matter what.


I'll go write a computer program later to figure this out. But you can already pretty much with mental math eliminate the possibility that 8 people go X-1, (X+1)-0 or X-0-1 without anybody going (X-1)-0-2.

[quicksilver]

What about say 4 people going x-1 and 5 going x-0-2?

[KrzyMoose]

What about say 4 people going x-1 and 5 going x-0-2?

Assuming no one is undefeated or X-0-1, the four people who are X-1 are guaranteed to make it, and one X-0-2 is going to miss out (due to tiebreakers).

Tiebreakers only matter between people with the same number of points.

[quicksilver]

Assuming no one is undefeated or X-0-1, the four people who are X-1 are guaranteed to make it, and one X-0-2 is going to miss out (due to tiebreakers).

Tiebreakers only matter between people with the same number of points.

That was not the question I was asking, I was asking if that situation was possible since it was stated x-0-2 is guaranteed to make it in.

[pi4meterftw]

I believe X-0-2 always makes it in as well no matter what.


I'll go write a computer program later to figure this out. But you can already pretty much with mental math eliminate the possibility that 8 people go X-1, (X+1)-0 or X-0-1 without anybody going (X-1)-0-2.

Actually, one can show using the usual model that after 8 rounds, it is no longer the case that X-1 even gets in.

[Van Phanel]

Actually, one can show using the usual model that after 8 rounds, it is no longer the case that X-1 even gets in.

Pics or it didn't happen.



Eh... what I meant to say is: Do so, show it.

[quicksilver]

Actually, one can show using the usual model that after 8 rounds, it is no longer the case that X-1 even gets in.

I think it is impossible for X-1 to not make it in under any circumstances. Do you perhaps mean X-1-1? In any case I do not beleive the number of round effects it at all, only how close you are to a cutoff.

[Nihil Credo]

256
128 128
64 128 64
32 96 96 32
16 64 96 64 16
8 40 80 80 40 8
4 24 60 80 60 24 4
2 14 42 70 70 42 14 2
1 8 28 56 70 56 28 8 1
The above (sorry for the bad alignment) represents the evolution of the distribution of scores over eight rounds of play, assuming the maximum allowed number of players and that nobody draws (or drops). As you can see, at the end there will be eight players at 7-1 and one at 8-0.

More generally, in any graph like the above (starting from exactly 2^N and evolving N "rounds"), the bolded number, which here is 8, will be equal to the number of rounds played. This means that, starting from eight rounds, if the number of players is sufficiently close to the maximum you can fail to make Top 8 even with X-1. Exactly how close is a cool little arithmetics problem, but we don't really care since we're just providing a counter-example to the "X-1 always makes it" claim and not operating under anything close to realistic assumptions.

For the curious, the graph can be obtained simply by drawing a standard Pascal's triangle and multiplying each line by 2^(N-i), where i is the number of the 'round'.

[Nemcon]

I've been under the assumption it always depends on how many players, and how many other draws have occured during the day.
If a lot of draws occur near the top, it screws a lot with pairings and tiebreakers.
And I know I'll always get paired down when I need an ID into top 8, lose, and end in 9th. But that's my own luck.

Maybe it would be better if I IDed earlier, but I know a lot less players would be willing to do so.
I also know that I've come into tournaments with friends, and one time, we IDed 3rd round (out of 8 I think) just to not send either of us to the losers bracket (we were 2-0 each at the time). By IDing, we ended up with what we thought were a lot more control matchups than normal (but it's hard to tell what we would have otherwise).

I think it is possibly a valid strategy, if you have a deck that is really good against control, as you said. But in Legacy...is there decks that destroy control? Is there a lot of drawing with time running out? I haven't paid attention too much in tournaments to know these numbers. If you have decks that can destroy decks that are more likely to stall out and draw matches, it could work. But remember, you could just as likely get paired against another deck that got stalled out by that control deck.

[cdr]

Ok, I misremembered (again) - the player/round cutoff is not based a guarentee that x-1 or better makes it, but apparently rather on more complicated math based on getting the expected number of players at x-1 or better as close to 8 as possible.

The cutoff for 8 rounds would have to be 216 (rather than 226) for a guarentee. The expected number of people at 7-1 or better is with 226 is 7.94531, 227 is 7.98047, and 228 people it's 8.01563. 8-1 or better at 409 is 7.98828, 410 is 8.00781.

Why 226 rather than 227? I dunno, and I doubt there are many people that completely know the reasoning behind the system.

You are still guarenteed at x-1 for 3 to 7 rounds, and the chance of not making it at x-1 in 8 rounds is vanishingly small.

[quicksilver]

256
128 128
64 128 64
32 96 96 32
16 64 96 64 16
8 40 80 80 40 8
4 24 60 80 60 24 4
2 14 42 70 70 42 14 2
1 8 28 56 70 56 28 8 1
The above (sorry for the bad alignment) represents the evolution of the distribution of scores over eight rounds of play, assuming the maximum allowed number of players and that nobody draws (or drops). As you can see, at the end there will be eight players at 7-1 and one at 8-0.

More generally, in any graph like the above (starting from exactly 2^N and evolving N "rounds"), the bolded number, which here is 8, will be equal to the number of rounds played. This means that, starting from eight rounds, if the number of players is sufficiently close to the maximum you can fail to make Top 8 even with X-1. Exactly how close is a cool little arithmetics problem, but we don't really care since we're just providing a counter-example to the "X-1 always makes it" claim and not operating under anything close to realistic assumptions.

For the curious, the graph can be obtained simply by drawing a standard Pascal's triangle and multiplying each line by 2^(N-i), where i is the number of the 'round'.

Thanks for this post. So it is techincally possible to go X-1 and not make it. It however is very unlikely since this model does not take into account draws.

[Forbiddian]

I believe that 256 players play 9 rounds, due to exactly the reason stated above although I couldn't find the formula for cutoffs and such in either the Magic Floor rules or DCI UTR.

[cdr]

I had thought the player/round cutoffs were common knowledge or had maybe already been posted. Maybe my last post will make more sense now if it didn't.



Number of players Number of rounds
8 3 rounds of Swiss
9-16 4 rounds of Swiss
17-32 5 rounds of Swiss
33-64 6 rounds of Swiss
65-128 7 rounds of Swiss
129-226 8 rounds of Swiss
227-409 9 rounds of Swiss
410+ 10 rounds of Swiss

[Nihil Credo]

Ah, thanks for the chart. I thought that the maximum number of players was always equal to 2^(number of rounds). Apparently the DCI guys were smart enough to tweak the Swiss system accordingly.

[Skeggi]

Those are only recomendations, right? A couple of our local tournament organizers prefer to do about 5-6 rounds while this chart indicates there should be 3-5.

[frogboy]

Now that Regionals have been split and you don't have ten round monstrosities, for all practical purposes, x-1 and x-0-2 are locks, x-1-1 is usually but not always good enough, and you never want to ID before the penultimate round unless there is a specific reason concerning a matchup or teammate.

Once standings go up just look at the point thresholds and if there are say six people with sixteen points looking for four slots, look at how good your breakers are relative to everyone else. I'm not sure if you're allowed to write down breakers between standings but before pairings for consulting purposes before you sit down; consult the HJ.

In general, those who lost in earlier rounds will have worse breaks and be forced to play, and someone who lost round five of a seven round event but won round six will be safe to draw in round seven.

GPs/PTs are obviously a different matter.

[cdr]

Those are only recomendations, right? A couple of our local tournament organizers prefer to do about 5-6 rounds while this chart indicates there should be 3-5.

TOs are free to do as they wish for their own tournaments. The only requirement for a sanctioned tournament is that there be at minimum three rounds.

Premiere events like PTQs, GPTs, etc are different - you have to follow DCI policy for those, which is outlined in the event info sheet.

The more competitive an event, though, the more I would want to follow the standard structure.

[HdH_Cthulhu]

I played in Nats Qualifier and my standing was 4-1-1. Then round 7 my opponent wanted to ID but he couldnt say anything about the math so i didnt ID. That made my opp very angry but we played game 1 wich I won.
Now (!) he explaind me the math and i would be in the t8 no matter what would happen. So he offered me all his prices for IDing, wich i accepted.

We played the next game just casual, and I took a free mull back to 7 and after his first land drop a friend of him sayed that there is an error in his math. He would ID on place 9. LoL. Now he wanted to cancel the ID wich i also accepted (cuz i took a free mull and had a perfect hand XD). Then i smashed his face!

Why do i tell this confusing story:

There are to much things to consider for IDing. Do the math for you and your opponent. Do it right. Also dont try to force your opp to an ID. Be patient and explain your opp the math so he would accept.

[Nihil Credo]

So he offered me all his prices for IDing, wich i accepted.

I'm pretty sure this is bribery and would have gotten you both a DQ. Outside of the finals it's forbidden to offer anything in exchange for draws or concessions.

[cdr]

Yep.

A) That is bribery.

B) Aside from that, once you decide the match (either through playing or agreeing to draw), you technically can't change the result.

[HdH_Cthulhu]

I'm pretty sure this is bribery and would have gotten you both a DQ. Outside of the finals it's forbidden to offer anything in exchange for draws or concessions.

It was the final round. But i didnt know that.

[cdr]

It was not the final round. It was the last round of Swiss.

[Dark_Shakuras]

It was not the final round. It was the last round of Swiss.

Yeah, the only time you can ID for a split (ether 50/50 or otherwise) is in the Finals of the Top 8 (16, 32 etc.). Anyother time and it is considered bribery. Anyone know the MTG artical the pro players get DQed in for this?

[Nemcon]

That's just shady all around. Yes, never offer or accept prizes for a draw/concession. My opponent once offered it, and my reponse? "Judge". Got the win from that.

Besides that, make sure the math is right too. I only do ID if I'm certain that 8 players will not pass me, or I'm certain that my tiebreakers are high (Ex: there are like 4-1-1 going down to 10th place, but my breakers are so much higher than the other two, There's no way I will get below them if they have the same record as me)

[Big Ear]

Try this site for determining IDs. Top 8 predictor (http://www.limitedinformation.net)

[Griselpuff]

Correct me if I'm wrong, but I thought the idea was, if you're in the Top 4 before the final round you can draw into Top 8 for sure.

In addition, if you're 14th or something like that, a win can at most get you into your current rank/2 (7th, in this case).

[Malchar]

At my lgs they print out the standings, points, and even tiebreakers for all players after each round and post it near the next round's pairings. This way, everyone knows whether or not they can safely ID. I assumed every store did this.

[cdr]

I know no one actually looks at the thread before posting, but this one was necroed by a spammer.

[Richard Cheese]

Not to promote necros or spammers or anything, but this seems like the right place for this question: Does anyone know of an Android app that can keep track of tournament statistics like number of tourney stats like players/rounds and win/loss record? Magic TG Tracker and MTG Utilities, two of the more popular apps, keep track of wins/losses, but don't really let you apply that to the bigger picture of a tournament.

[Sintheros]

I too would love to find one of these, I'm personally awful at tournament math.

[Malchar]

I know no one actually looks at the thread before posting, but this one was necroed by a spammer.

I actually read the thread, but I didn't bother to look at the dates on the posts. To me, it doesn't really matter what the date of the last post was. It's still a relevant discussion.

[Big Ear]

Not to promote necros or spammers or anything, but this seems like the right place for this question: Does anyone know of an Android app that can keep track of tournament statistics like number of tourney stats like players/rounds and win/loss record? Magic TG Tracker and MTG Utilities, two of the more popular apps, keep track of wins/losses, but don't really let you apply that to the bigger picture of a tournament.

I too would love to find one of these, I'm personally awful at tournament math.

This is actually pretty easy, since it reduces nicely 99% of the time. Its simply

Your match points
------------------
3 * #of rounds

But before draws, its really just

# of Matches won
------------------
# of matches played

Also, there is really no need to divide it out into a decimal, the number is just as meaningful as a fraction.

[Griselpuff]

yeah i just noticed the necro...

but my questions still stand:

does being in the top 4 before the final round GUARANTEE an ID into top 8?
to get into the the top 8, your standing needs to be at least 16th before the final round (more likely you need to be something like 12th, but in theory 16th might get in assuming nobody drops?)

[cdr]

yeah i just noticed the necro...

but my questions still stand:

does being in the top 4 before the final round GUARANTEE an ID into top 8?
to get into the the top 8, your standing needs to be at least 16th before the final round (more likely you need to be something like 12th, but in theory 16th might get in assuming nobody drops?)

Please use capital letters.

In a normal tournament, the only thing that guarantees making top 8 is a record of X-1 or better. The swiss round recommendations are set up mathematically with that in mind.

If you're in the T4 going into the last round, chances are you can draw and be X-0-2 (which is better than X-1) or X-1-1, which is not guaranteed but usually safe. If the number of players was just under the number for an extra round, it's possible for an X-1-1 to be 9th if everyone draws.

The more rounds a tournament has, the less stable tiebreakers are. In Grand Prix Denver 2001, Dave Price was in 8th place on going into the last round, with the best X-2-1 tiebreakers. He won his round and when final standings went up he was in 9th - pretty much every person he had played in the last 12 rounds had lost.

[Griselpuff]

So, in other words, the only guarantee for Top 8 is having at least <total possible points> minus three.

[cdr]

So, in other words, the only guarantee for Top 8 is having at least <total possible points> minus three.

Yes - assuming the recommended, normal round structure.

And while that's good to know, you're best off taking into account the situation you're in at that particular event. Depending on various factors, a record as low as x-2 might be a lock.

[Irenicus]

The more rounds a tournament has, the less stable tiebreakers are. In Grand Prix Denver 2001, Dave Price was in 8th place on going into the last round, with the best X-2-1 tiebreakers. He won his round and when final standings went up he was in 9th - pretty much every person he had played in the last 12 rounds had lost.

Just read the quoted passage and it doesn't make any sense to me. Maybe I miss something but as I see it the tiebreakers should me more stable the more rounds a tournament has. The reason for this is that tiebreakers are specific numbers and therefore possible change decreases with every round.

An example:

Case A: 5 round tournament. After round 4 your first tiebreaker is 50%. In the fifth round each of your opponents loses. Therefore your new tiebreaker should be:

(200%*4/5+the match win-percentage of your last opponent)/5

Case B: 10 round tournament. After round 9 your first tiebreaker is 50%. In the tenth round each of your opponents loses. Therefore your new tiebreaker should be:

(450%*9/10+the match win-percentage of your last opponent)/10

I don't want to assume a tiebreaker for the last opponent but you can see that the first part of this calculation changes more in Case A.

Case A: (200%*4/5)/5 = 40,00%

Case B: (450%*9/10)/10 = 46,67%

I hope that I made my thinking understandable. If I made a mistake please let me know.

Tiebreakers by the DCI (http://www.wizards.com/dci/downloads/tiebreakers.pdf)

[Big Ear]

I'd have to agree, with the tie breaker being an average, i'd expect it to be more stable the more data points you have. Even at the end of a long tournament, if all your opponents lose, the impact to your points should be minimal.
This could still make the difference between 8th place and 9th though.

Went and looked up the data.


Round 11
1 Batarseh, Sammy 30 71.10%
2 Mandel, Daniel 29 75.08%
3 Borteh, Alex 29 73.92%
4 Price, David 28 72.46%
5 Buffi, Geno* 28 70.40%
6 Abraham, Mike* 28 63.87%
7 Knobloch, Aaron* 28 60.66%
8 Parker, Brock 27 72.29%
9 Kastle, Darwin 27 69.53%

Round 12
1 Batarseh, Sammy 31 71.9577%
2 Knobloch, Aaron* 31 61.3757%
3 Borteh, Alex 30 74.9206%
4 Mandel, Daniel 30 72.7778%
5 Kastle, Darwin 30 70.5908%
6 Parker, Brock 30 69.3122%
7 Abraham, Mike* 29 65.1786%
8 Price, David 28 71.9841%
9 Shears, Brett 28 71.1400%

Round 13
1 Batarseh, Sammy 32 71.6850%
2 Abraham, Mike* 32 64.1589%
3 Knobloch, Aaron* 32 62.1443%
4 Borteh, Alex 31 73.3933%
5 Mandel, Daniel 31 72.2611%
6 Kastle, Darwin 31 70.7326%
7 Shears, Brett 31 69.8413%
8 Parker, Brock 31 69.5604%
9 Price, David 31 69.2474%


BTW, decided to run the Denver GP information(402 players,13 rounds) into http://www.limitedinformation.net.
The top 8 results are nearly identical.

[mchainmail]

If you're in the T4 going into the last round, chances are you can draw and be X-0-2 (which is better than X-1) or X-1-1, which is not guaranteed but usually safe. If the number of players was just under the number for an extra round, it's possible for an X-1-1 to be 9th if everyone draws.


X-1 is better than X-0-2.

In a 5 round event, 4*3 points vs. 3*3+2 points.

[mchainmail]

If you're in the T4 going into the last round, chances are you can draw and be X-0-2 (which is better than X-1) or X-1-1, which is not guaranteed but usually safe. If the number of players was just under the number for an extra round, it's possible for an X-1-1 to be 9th if everyone draws.


X-1 is better than X-0-2.

In a 5 round event, 4*3 points vs. 3*3+2 points.