3x3x3 Blind, changing to 3/5?

[kinch2002]

So here's my opinion

Speed is very very important. Especially in single 3bld, where it's not so difficult to get a success as in bigcubebld. I think the accuracy rate is about where it should be. I guess currently most of the fast blders get a success in most rounds on bld, which is good I think, because you don't want a podium of people who did 9 minute safety solves, and you also don't want people getting every solve success because that indicates that they're going too slow if the format is best of 3. So I don't have a problem with the system at the moment because I think it works well. But, I wouldn't be opposed to best of 5 at all as long as slow people have a cutoff of some sort (like sub-3 success in first 2 attempts or something).

 

[Olivér Perge]

But, I wouldn't be opposed to best of 5 at all as long as slow people have a cutoff of some sort (like sub-3 success in first 2 attempts or something).

In my opinion if the format for blind would be average of 5, our style would suit for it, and all the competitors would go a bit slower. I don't see why everyone is saying that slow people would come first... Fast people cannot go slower? Or slow solving means definitely 80% accuracy?

On the other hand, Thom has a very good point: Blind is not like speed solving. At all.

Honestly I don't see this happening, ever, but maybe that's just me.

 

[Stefan]

What about making blind a "how many successes can you get in xx amount of time" format?

Blows up the database format, you're gonna have to rewrite it then cause I don't want to

 

[Edward]

A good comparison can be made to the sports teams that aren't used to playing in cold conditions, real grass vs astroturf, etc. It's much harder to train in snow in Florida (american football) than it is to train in a very loud distracting area (no matter where you live).

Yeah we don't like the cold much.

 

[Pedro]

Actually the fatigue argument was headed more towards the 3/5 average proposal. We practice for speed events (which are 3/5) by doing many many solves in a row and rolling solves and stuff. I don't think that is doable for blindfolded, but maybe that's just me.

What I'm saying is that we shouldn't try to make it the same as speed events because it is, in nature, not the same. It's not like a system with feedback where you can spot a mistake and fix it quickly (or even a misalignment, which can make you confuse about wheter you did or did not do a move, or POPs, which are a disaster in bld, but can be fixed in speed events)

 

[qqwref]

@qqwref#34: Are "gods of consistencies" necessarily slower?

When I say a "god of consistency" I mean a fast BLD solver who can go at maximum speed and still get a very very high consistency. I don't think everyone is capable of this.

If media sees many DNF's they will think it is too difficult and they will not report on it (media doesn't report on the people dropping out of a marathon). If DNF's are too common media and the audience loose interest.

You keep mentioning the media. This proposal will mean fewer DNF singles, sure, but also fewer good solves as everyone who isn't a god of consistency slows down considerably. I think the media is more impressed by good solves than by consistency - they seem to be on 3x3 speed, too. The cubing community cares about averages and all the media wants to know is what your best single solve was.

And if you make 2 or 3 mistakes in a round of 5 I see no reason not to kick someone out.

OK, so let's make a rule that on 3x3 speed you get DQ'd if you make at least 2 or 3 mistakes of the form: (a) locking up, or (b) misrecognizing a case, or (c) popping, or (d) starting the timer incorrectly, or (e) doing something like U U'. This should make competitions more exciting for the media because the slow people wouldn't bother to compete. And hey, who cares if someone gets a DNF average on what might be their one competition this year? They just don't have what it takes and should be kicked out of the event.

@aronpm#30: Why is it too difficult? Because they don't practice for accuracy but for optimal single speed, which makes perfect sense under the current rules. The point of rankings in competitions is not to get the best times you can. It is to compare your skill to others. Again, the link between slower and more succesful isn't proven.

It's too difficult because it's too difficult. If you misremember a sticker/piece, forget an object, make a turn and aren't sure what you did, undo a setup move wrong, perform any cycles in the wrong direction, or anything similar, and don't know exactly what you did wrong, you will get a DNF with certainty. I will admit that I go for high speed in competition, but in practice I tend to try for accuracy and yet my success rate is not much better. It's been terrible ever since I started and probably always will be. And actually, I think the point of rankings *is* to get the best times you can, so that the competition can rank that. It's easier to rank times than skill, especially since a BLD avg5 will only tell you whether someone managed to get an average or not.

@tim-Pedro#35-37: So we agree that best of 5 would enhance the results of lucky cases and that results would be different depending on group. That seems like proof to me that best of 5 is BAD

Wouldn't it actually put LESS emphasis on lucky cases? If you get a very lucky solve, other people get 5 attempts in which to beat it, instead of 3. It's similar to how best of 3 involves less luck than best of 1.

So let me propose an idea that is very rough and that I really would like to see discussed as well:
What about making blind a "how many successes can you get in xx amount of time" format?
[...]
- You get a total time for the event of xx minutes (let's assume 10 for now)

I'll surprise you: objectively, I like this idea as an event. It combines speed and accuracy without making either unimportant, and because you don't DNF all results with less than 80% consistency it is much more yielding to slow people. However:
- It feels a lot like multi, considering the scoring system, and we have enough of that. Also the need to be able to do BLD solves in quick sequence means that probably the vast majority of people will get a best result of 1 or maybe 2 points, and only a few people will get beyond that (and just like multi some will go very far beyond). Rankings like that tell me that the event is too difficult for most competitors.
- As mentioned, it's really tough for spectators and for StefanDatabasemann. Do we even record single solves? If so, how? As a combined final type round with anywhere from 1 to 15+ solves in it?
- It would make it harder for a beginner to do solves. Since we need to limit the time globally - to 10 minutes, say - anyone over 5 minutes would likely only get one real attempt. This compares to competitions now where a nice organizer may let them try 3 times. And even if you are a bit slower, if you fail the first solve you must then get 2 more to get a positive result, which is annoying. By a "rolling average" type of argument I'm not sure someone who gets the first solve they try and then stops is any more skillful than someone who gets a DNF, then a solve, then stops.
- It definitely goes against the current feel of the event. 3BLD is based on performance in single solves; we don't measure 3speed by having people do as many solves as they can in under 60 seconds.

 

[MatsBergsten]

I don't think it has been explicitly mentioned that the difference in scrambles has a greater impact in bld than normal speedsolving. Compare a scramble with only one cycle each for corners and edges (and perhaps a few pieces correct) and no parity with one scramble with lots of cycles and twisted/flipped but correctly placed pieces and on top of that also parity.

I may underestimate (as I am a good bld-er but a very bad speedsolver) the bonus of an easy cross or already connected pairs in speedsolving, but one very lucky solve in speed does not affect a speed average at all. In bld it makes the whole result.

So I think it would be more fair with an avg also in bld. But I also see the backsides of the need for 80% accuracy. Why not take the best of both Arnauds and Michaels arguments: an avg 3/5 does not necessarily have to be the same in bld as in speedsolving. How about an avg of the three best of five, that would allow for 2 DNF:s but still require 60% accuracy so no one but the "gods" mentioned can go for maximum speed.

If you find that 5 solves take to much time, an avg 2/3 (that is avg of the two best solves of three) would remove the fluke/chance solves but still not require too much accuracy.

You may say that I personally would benefit of this (as being the slowest turner of all). Perhaps I would, though I thought more like Arnaud in the beginning of my bld career. I don't really see the need for change today, most often the best blindcuber (however we define that ) wins no matter what.

As for averages for big bld there's really no need for it. The very much greater number of moves needed to solve a big bld both in itself constitute an average and also lessens the effect of good/bad scrambles.

------------


Edit: post 900
Edit 2: perhaps this avg 3/5 does not suit the WCA DB/Stefan either, too bad

 

[riffz]

How about an avg of the three best of five, that would allow for 2 DNFs

You beat me to it. This is at least somewhat reasonable, although I still don't like the idea of taking BLD averages anyway.

Kirjava is basically summing up my viewpoint on this issue, so I won't bother repeating points that have already been mentioned.

 

[Pedro]

Another thing I just thought...

Let's say Feliks uses 55 moves in average. So his 5 solves take 275 moves. How many mistakes can he make and still get a sub-10 average? I'd say quite a bit...like 2 or 3 per solve? Let's say 3, he may be out of shape...10 out of 275 = 5.45%.

Let's say a very good blder uses something like 30 moves for corners and 45 for edges. In 5 solves, that's 375 moves. (And this is a top blder, not the average competitor who uses maybe Pochmann or long cycles with setups and possibly piece orientation). How many mistakes can he make to get a valid average, no matter what speed? 1! one! ONE! How much is that? 0.27%. Which is 20 times less than what Feliks can do. I don't see why we should be so harsh with someone, specially since 3x3 bld is already much harder than regular speedsolving...

Is like I said, the events are very different in nature. Regular speedsolving is about knowing a lot of algs and constantly looking ahead and updating the cube status in your mind, with the crucial advantage that you can see when you make a mistake or can make a change in your plan which may give you an advantage. Bld is about planning everything ahead and making every turn, algorithm, object conversion, memory browsing, everything correct. There's no room for the tiniest mistake.

Different events shouldn't have the same measurement/treatment, in my opinion.

 

[riffz]

I have to agree with AVG for the most part.

- 3/5 is the preferred format for most puzzles (now also for megaminx) and I long for the day that it will be available for even more events like 6x6x6, 7x7x7 and 3x3x3 blind (I am serious about that last one, but please don't fill up this already hard to maneuver topic with replies to this)

I know you asked specifically that we don't reply to this, but I just had to. Expecting competitors to get 4 successes out of 5 attempts in BLD is ridiculous IMO.

Yes, because people are totally incapable of solving 24 out of 24 cubes succesfully :fp

But let's continue that discussion here: http://www.speedsolving.com/forum/showthread.php?t=24079

That's entirely different. Divide the total time that attempt took by 24 and you'll notice that the indivudal times per cube are not comparable.

If I proposed that competitors not be allowed to finish an average unless the mean of their first 3 solves was sub-10, I could also point to people who could achieve that easily, but it doesn't make it reasonable or applicable to the community as a whole.

 

[Jani]

In here, the BLD competitors are so many. As you noticed in Indonesian Open 2009 perhaps.

At our last competition, we were running more than 1 hour late just because 3x3 BLD events. I just can't imagine if it becomes Ao5.

Maybe we should lower down the limit, to 3-5 mins?

 

[Gunnar]

In here, the BLD competitors are so many. As you noticed in Indonesian Open 2009 perhaps.

At our last competition, we were running more than 1 hour late just because 3x3 BLD events. I just can't imagine if it becomes Ao5.

Maybe we should lower down the limit, to 3-5 mins?

But this is a problem of how the competition was organised, not of the format itself. The format is "best of " e.g. you may give 1, 2 or 3 attempt per person, depending of how much time you are willing to use for the event. The same could go for a format "best of <5". The only difference is that the organiser MAY allow 4-5 attempts, if they find it appropriate.

Time limits are also already allowed to be set by the organiser.

On subject, I guess I agree with what Sakarie and others wrote. The event is very different from speed events. The times would most probably increase, if we start to use the average format, since you won't benefit (as much) by going fast. If you are very fast you know that by going safe you may still get an average faster than a slow solvers average, but the single times won't be as impressive.

Of course, one may say that singles aren't important, but in sighted events you may go for full, or close to full, speed and still have a very high chance of getting a good average. To me, singles are much more important in BLD events.

I don't know if one person's stats tells much of the blind events, but I have maybe 25% accuracy in 3x3BLD and 80% accuracy in multiBLD just because I memo a little slower in the latter event.

 

[macky]

I'm against average of 5. I won't repeat the argument that blind is inherently different from speed, but here's the objection from Pedro that I find most convincing, and which no one seems to have directly addressed:

And, seriously, I don't like the idea of a [(8:03), 9:04, 9:30, 9:50, (DNF)] beating [(48), 49, 50, DNF, DNF]...

But I'm not against average itself. Ignoring for now the question of format, how should the following results be ranked? Should C be ranked first?

A: 8:03, 9:04, 9:30, 9:50, DNF
B: DNF, 1:31, DNF, DNF, 1:50
C: DNF, DNF, DNF, 50, DNF

How about here?

A: 8:03, 9:04, 9:30, 9:50, DNF, 8:50, DNF, 9:31, 8:38, 9:01
B: DNF, 1:31, DNF, DNF, 1:50, DNF, 1:49, 1:33, DNF, DNF
C: DNF, DNF, DNF, 50, DNF, DNF, DNF, DNF, DNF, DNF

I think there's a good argument to be made for a success rate requirement. It's also true that scrambles vary considerably in difficulty (especially for certain methods), and so there's a good reason to rank by average, which effectively sets a success rate requirement. I'm fine with this as long as it doesn't significantly change the top blindfold cubers' competition strategy/approach to single solves.

For example, with a 3-attempt round, mean of best 2 (requires 2/3 success) is already too much; DNFing the first solve forces two safety solves. Compare this to speed events, where DNFs are rare and can be avoided with little loss in speed. Even after a first-solve DNF, a top 3x3 solver can do 4 more solves at full speed with little worry. If any success rate requirement is to be in the rules, it shouldn't be much higher than 50%. An organizer who feels strongly about this can set a cut-off after 2 attempts, without any rule change. With a 5-attempt round, 2 out of 5 (mean of best 2) seems acceptable, but probably not 3 out of 5.

So, a question for the top blindfold cubers: would requiring 2 successful solves out of 5 significantly change your competition approach to single solves? How about 3 out of 5?

 

[Kynit]

Would median of 5 be a bad idea?

You'd need a 60% success rate to get a valid 'average', which seems very possible by most competing BLDers, and having a 4th/5th success can only help you. Is it too much different from 3/5 to be a good comparison?

 

[syuhei222]

For example, with a 3-attempt round, mean of best 2 (requires 2/3 success) is already too much; DNFing the first solve forces two safety solves. Compare this to speed events, where DNFs are rare and can be avoided with little loss in speed. Even after a first-solve DNF, a top 3x3 solver can do 4 more solves at full speed with little worry. If any success rate requirement is to be in the rules, it shouldn't be much higher than 50%. An organizer who feels strongly about this can set a cut-off after 2 attempts, without any rule change. With a 5-attempt round, 2 out of 5 (mean of best 2) seems acceptable, but probably not 3 out of 5.

So, a question for the top blindfold cubers: would requiring 2 successful solves out of 5 significantly change your competition approach to single solves? How about 3 out of 5?

I agree with Macky in principle.
I think [2 success out of 3 attempts] is better than [2or3 out of 5].

For most competitor, on BLD event, they need longer times than regular speedsolving events.To time 5 attemps for ALL competitor may cost too much times. It will trouble with Organazation-team.
3 attemps at most is suitable.

 

[Pitzu]

In Euro 2010 final there was 15 successful solves among 45 tries. I invited 2 (non cuber) friends of mine and they saw 2/3 of the solves failed.

 

[RCTACameron]

I agree with this idea. How about scoring in this order:

Fast average
Not as fast average
3 successes, fast single
3 successes, not as fast single
2 successes, fast single,
2 successes, not as fast single

etc...

 

[yoruichi]

1. u get 3 solves
2. if one of the solves < 4 minutes then u get 2 more solves?


i didn't read the whole thread so this idea is probably in it somewhere

 

[Tim Major]

Would median of 5 be a bad idea?

You'd need a 60% success rate to get a valid 'average', which seems very possible by most competing BLDers, and having a 4th/5th success can only help you. Is it too much different from 3/5 to be a good comparison?

I like this idea. But why not 1/3. Take out best and worst and keep that as the "average", and then a best time can be the "single". 2 DNFs = DNF average, but this way, you can do a safety solve, a faster solve, then a flat out attempt

Edit: And then it would be presented in similar format, to other events.

 

[yoruichi]

does mean of best 3 solves out of 5 work?