3bv limits

[KamilSaper]

I have one idea. Something like 6 months ago Rodrigo told that he is going to make new Clone. What about removing 3bv limits?
1. We could keep the world ranking with 3bv limits scores, but everybody would be able to get better personal records.
2. This would be the only way to check what are the best possible times to get.
3. MSX is an official version and it doesnt have 3bv limits. If playing on no-3bv limits version is legit, what is the problem with removing 3bv limits on clone and keep having history file?
4. I talked with Damien and he claimed that people would be very jealous if somebody got very good score on very low 3bv board. Anyway I believe that with enough time spent playing stats connected with getting easy boards are the same for every player. What is more, even boards with sup2/30/100 3bv boards are sometimes that easy that others are jealous, so what's the problem?
5. Removing 3bv limits would the only chance for players for get new"the lowest 3bv board ever"
As far as the world ranking would keep 3bv limits rule, I don't see any problem with removing 3bv limits on new Clone I'd like to know your opinions.

[EWQMinesweeper]

IMO 3BV limits should stay - though i would like to know why 30 and 100 were chosen back then when these rules were made

@1. that would lead to a messy and unfair ranking system. also for me minesweeper is not about getting better records but about getting better at sweeping (i hope it is clear what i mean)

@2. i think that the best way to check that would be upk-ing boards that other sweepers got in normal mode (like manu's 25s (i think) exp game). also 'best' is not 'fastest'

@3. the point is that if you get a score on msx that has a 3bv below the limit it would simply not count for rankings and statistics - therefore these games should not be included in stats and history files (at least not on the same tables as normal scores)

@4. there is no real problem. if someone's jealous - so what? be jealous and take it as a motivation

@5. i am not sure if this is really sth that is worth aiming at it


i think it would be better if the limits remained in the normal mode - maybe they could be removed in density mode or elsewhere, like in another mode. as these scores are not even valid i see no point in adding them to any stats together with valid boards - maybe stats tables for these boards below the limits...

[LTsully]

I think there should be 3bv limits but they should be re-adjusted to 2-25-90, one explaination of 3bv limits I have heard it to prevent anything like a 1 click win board on int because it is possible in theory. A lot of times done on now illegal 3bv remain on the world ranking, Damiens best 11 is illegal, and someone has 3-11-65 because the 11 was on 25 3bv.

[Mucleus]

I suppose I disagree with having 3BV limits... really it only comes into effect on beginner (stopping instant win, but is this really a problem? It's beginner's fault for having just 10 mines ;P) and intermediate, as I don't think I ever saw <100 game when I used to play on winmine, and I've only seen perhaps 2 games <110 3BV (incidentally, completed one to break my record 51->46). After all, minesweeper is a fundamentally lucky game. And shouldn't we have a measure of 'board difficulty' that actually seems to correlate well to performance before disallowing certain boards?

[Cryslon]

I think we should keep 3bv limits, maybe even raise them to smth like 10-40-130 to ban extremely rare and easy boards. Also, i think we should better look at stupid 50-50s. Clone looks like hell after minesweeperlive because of swarm of no-brain forced guesses.

(edit) @Kamil: if you want to play low-3bv boards, cheat mode with 3bv restrictions is specially for you. Or, maybe, you want to have separate mode with 3bv restrictions, without other cheats, and with pop-ups?

[KamilSaper]

@Luke: 2-25-90 would definitely make too big mess. Both Chinese players who got that 10s must have been accepted, as well as my 8s on 29.
@[name removed]: sub30 and sub100 3bv boards wouldnt be counted for for any rankings as well as history file wouldnt mention it. Players would play them just for fun. When I get 9s on Clone, Im going to start playing on MSX, just to get for example sub8 on sub30 board And then u all will be jealous, while wasting time on Clone

[EWQMinesweeper]

lowering the limits would make a mess, i agree on that. i see no need to raise them either. i see sub130 3BV exp boards almost every day i play exp, same goes with sub40 on int.

3BV limit on beg is clear and i think the limit on exp is ok too - i would vote to keep the current limits and maybe we could ask rod if he can add a mode to the next clone whcih is like the normal mode just without 3BV limits

[LTsully]

Well, I looked through the top 100 and found 9 int games done on now illegal 3bv, and 2 in the top 12, my question is why are these games kept on the ranking if they are illegal?

[KamilSaper]

I dont really see the point of making totally new mode. What is somebody got a record on legit 3bv in this mode?
@Luke: long time ago players decided to leave old scores alone. U know, to much mess if the wanted to find all players legit records.

[EWQMinesweeper]

@kamil: it was just a suggestion, nothing more to answer your question i'd like to point at clone's current density mode. as far as i know highscores on legit boards in that mode would count for rankings too

[Cryslon]

I dont really see the point of making totally new mode.

I don't want to solve board, look at red timer, see a record time, and then notice that 3bv is invalid.

[KamilSaper]

Then Rodrigo could add an option "generate only legit 3bv boards". If somebody doesnt want to get a record and realise that 3bv is not enough, he or she could just turn on that option

[LTsully]

Or there could be a cheat mode option that gave only illegal value 3bv boards.

[KamilSaper]

I also thought about it, but it would suck - its not the same as getting a nice board in normal mode - u get nervous if sub30 board if for example 1 out of 1000 boards, and u dont get nervous if there are 100% sub30 boards. Try 1st version of MSX or old Arbiter and set it to generate only sub30 boards if u want that cheat mode . It's easy to get a 9 then.

[abiu]

I think 3BV has a Normal distribution.
So we need to talk 3bv limits by "σ".
1σ ? 2σ,3σ or ....?

http://en.wikipedia.org/wiki/Normal_distribution

[Rogen]

@abiu: 3BV doesn't have normal distribution (at least because it has big tail of high 3BV boards). But nevertheless you can find normal distribution, which could approximate 3BV distibution, you'll see that 3BV 30 limit on int corresponds exactly 3 sigma

[KamilSaper]

i dont understand the picture, can anybody explain?

[Rogen]

I guess the explanations except "Normal distribution" are useless. Study probability theory.
Or use this http://en.wikipedia.org/wiki/Normal_distribution

[Mike.You]

According to the probability of the emergence of 3BV in 10000 games.U can educe the value of "μ" &"σ" : Beg 16 , 5;Int 65,11 ;Exp 172, 19.

[KamilSaper]

Let me guess: the picture means that if average 3BV is X, then the ammount of X-Y 3bv boards is the same as X+Y 3bv boards?

[Mike.You]

"μ" is not the average 3BV. It appears With the largest probability.

[Tommy]

[EDIT: moved this and the following 2 posts from lucky mode to here - EWQMinesweeper, apr 20]

Thanks for your replies, especially EWQMinesweeper - I don't have much to add.

By contrast, imagine Ian blasting his current expert highscore on a 50:50 near the end. That would be even more frustrating than trying to get a new highscore now would be futile.

For me these two sentences don't make sense.

same for me. ian finished his 33. so what? ian was/is on the same level as manu, dion, me and maybe tk and pavel. single-digit-ranked players seemingly don't care so much for their position on the ranking anymore. at least i am yet to hear a top10 player complain about ian's 33. ian did not stop playing because of his 33.

What I meant: This is a situation where luck plays a large role.

This may not have been obvious, but I don't actually like the luck aspect of the game; I would like to compete with other sweepers on an as even footing as possible. Yes, I play mostly for fun, but that doesn't mean that I don't want the ranking to reflect skill as well as possible. I just think that removing luck in this fashion is not sensible, that it changes the game too much. And therefore, I proposed alternative ways to make luck count less, that don't change the nature of the game in a way that makes it easier, and that don't allow additional assumptions to be made ingame*.

And that means that the rules will have to state exactly what patterns will be recognized and what patterns won't. To explain this rule to a newbie, you need to go past the basic rules; You will need to explain details about the game that are best learnt through practice.

I don't see this as a big problem. Eg, we have 3BV limits; and you suggest ZiNi limits. These rules are much harder to explain, but you don't see this as a problem at all.

Actually, the additional complexity of ZiNi is a good argument against ZiNi limits in my eyes. 3bv newbie explanation:

There is a statistic called 3bv. To calculate it, you count one for each opening, and one for every number that isn't adjacent to an opening. There are 3bv limits of 2-30-100 for the respective levels, and games played on boards with less won't count for the rankings; Most clones will not even generate such boards.

I don't think that 3bv limits are bad. I think that we need the expert limit to be more strict/effective. And I'd consider ZiNi limits anyway, because they are a much better benchmark, and still explainable to newbies. On the other hand, ZiNi has edge cases too, especially path-wise. So, I'd say that ZiNi is close to the limit for acceptable complexity, and I'm not saying that it is necessarily still OK - I haven't really made up my mind yet.


*This is not entirely the case. There are edge cases where a 3bv limit could allow deducing things ingame. For example, "if there is an opening here, there is no way for this board to be above the limit". But we can safely assume that nobody will ever use this for competitive minesweeping, ever, and if anyone disagrees, please open a new thread.

[qqwref]

Speaking of difficulty limits, I don't like the idea of ZiNi limits because:
- The metric itself is not perfect. For instance most of its decisions are based on information that may not be possible to obtain during gameplay, and the ZiNi value of a board changes if the board is placed in a different orientation (this is IMO a big one). And if we add an official limit like this, it had better be a well-thought-out limit, because it'll be sitting there for a while.
- Even if we have a better metric, something like ZiNi is harder to calculate than 3BV, which means more lag when starting a board. I wouldn't want people to need good computers to get optimal speed in minesweeper.
- Having two separate board-difficulty limits would end up making boards of a certain 3BV (say, <110) even rarer - and since flagging metrics don't really matter very much to NF players, it would end up hurting NF players for no logical reason.
- I'd also be against simply replacing the 3BV limit, because that could potentially allow some boards that weren't allowed before (and again this would be pretty silly to NFers).

[Tommy]

The metric itself is not perfect. For instance most of its decisions are based on information that may not be possible to obtain during gameplay, and the ZiNi value of a board changes if the board is placed in a different orientation (this is IMO a big one). And if we add an official limit like this, it had better be a well-thought-out limit, because it'll be sitting there for a while.

The asymmetry thing is really bad, yes, but the rest is true for 3bv as well - you don't open openings first, either. ZiNi is really hard to beat - if we assume that relative efficiency fuckups are reasonably equally distributed (as we implicitly do for 3bv), that isn't that much of a problem.

Even if we have a better metric, something like ZiNi is harder to calculate than 3BV, which means more lag when starting a board. I wouldn't want people to need good computers to get optimal speed in minesweeper

Shouldn't ZiNi be O(n^2) in the size of the board? Every iteration calculates at most n premia, and there are at most n iterations. Note that that is a very rough upper bound, the amortized runtime may well be much less (maybe even O(n) if you handle premium calculation efficiently - after all, you don't need to recalculate all premia, and the more you recalculate, the more squares you just opened. If there is an upper bound for the number of times that a premium is recalculated, I'm actually almost certain, but I may just be tired).

Anyway, with fixed board sizes, that is probably not going to be the issue - and see below.

Having two separate board-difficulty limits would end up making boards of a certain 3BV (say, <110) even rarer - and since flagging metrics don't really matter very much to NF players, it would end up hurting NF players for no logical reason.

If more clones had an NF mode, this would be less of a problem. However, I'm for applying limits after a game was finished anyway - that would also solve the problem in the previous point completely.

I'd also be against simply replacing the 3BV limit, because that could potentially allow some boards that weren't allowed before (and again this would be pretty silly to NFers).

Since I would only want to add limits, I'd be for keeping 3bv limits anyway. We can have both, that isn't a problem - and make ZiNi only count for FL.

[Tjips]

I just want to throw in some defensive commentary on behalf of zini limits here. I first want to address the point that it will take too long to calculate and thus create lag in the play of the game. As knowing the difficulty of the board beforehand is UPK, it doesn't need to be calculated beforehand. I'm sure a smart coder would even be able to have it calculate in the idle time between clicks. I'm talking of the basic levels of course. Also, the standard mid-range CPUs today are easily capable of a few hundred zini calculations per second, so I really can't see much of a lag issue unless we need to think about pre-Pentium 4 CPUs...

On the point that zini is inconsistent due to it containing location choices, thereby being influenced by the shape of the board: This is definitely true, but can be accounted for relatively easily I think. One possible approach would be to calculate zini using all the relevant X-X-most (e.g. top-left-most) rules for breaking tied premia, and having the final value be the lowest of the 8. This is of course very tedious and time consuming. One could perhaps only do it if a sample one is within a certain region above the actual limit. This would save 1/8 th of the zini computing time on most boards. But this would only be a stop-cap, however. The *best* solution I can think of would of course be some form of deep zini where it explores every single choice in every tied premia situation. This would supremely eliminate the symmetry problem as the final result would be independent of the shape of the board, and only dependent on the model used to determine the premia.

I'm not trying to get into a discussion of how we should implement and improve zini here though, I'm merely trying to make the point that shape independent implementations are conceivable.

On the issue that zini limits would be slightly nonsensical when considered in an NF context: Well NF is imo a self-imposed handicap which many sweepers practised initially to challenge themselves for fun, and which has become a sub-sport within the community. As such I think it would be prudent to promote the implementation of so-called "NF mode"'s in the popular clones within which different limits can be applied. I would actually go even further if I where to write a clone. I would have the thing only show either 3bv or zini (and their relatives) in the counters window depending on whether or not the player actually used flags (i.e. opened a square using a dblClick/chord). But that's just me

Anyhow, that's my 2 cents...