|
On April 12 2013 08:24 FallDownMarigold wrote:
I have a question about the QXC widow mine stuff, based on some of the posts here. So apparently it's really good in TVZ to do what QXC does, based on his games vs Idra + Stephano, in which he crushes them. My question is if it is simply an amazing strat, and not the result of some other factors at play, then why did I not see this strat employed in high level games, such as GSL quals and other Korean pro games? Do the Terrans at that level simply know less than QXC, or is QXC's success only possible vs. players like Idra + Stephano?
Excuse my ignorance, I'm just kinda lost. Should probably start watching more HOTS in general.
If its possible against Stephano is possible against anyone Stephano is one of the very best zergs on the planet and has been for a long time. But the thing with Stephano it might work 1 time only. He is very good at coming up with strategies against tactics like this but we will see
|
Widow Mines are too good for their cost.
|
On April 12 2013 08:29 Benjamin99 wrote:Show nested quote +On April 12 2013 08:24 FallDownMarigold wrote:
I have a question about the QXC widow mine stuff, based on some of the posts here. So apparently it's really good in TVZ to do what QXC does, based on his games vs Idra + Stephano, in which he crushes them. My question is if it is simply an amazing strat, and not the result of some other factors at play, then why did I not see this strat employed in high level games, such as GSL quals and other Korean pro games? Do the Terrans at that level simply know less than QXC, or is QXC's success only possible vs. players like Idra + Stephano?
Excuse my ignorance, I'm just kinda lost. Should probably start watching more HOTS in general. If its possible against Stephano is possible against anyone Stephano is one of the very best zergs on the planet and has been for a long time. But the thing with Stephano it might work 1 time only. He is very good at coming up with strategies against tactics like this but we will see
To be honest, the only thing I took away from that qxc/Stephano game, is that Stephano plays with 5 hotkeys, and will send his whole army back and forth in a clump all over the map to kill 8 window mines.
Also: "I don't really need to be aggressive much more". Turns out letting your opponent macro without any pressure or harassment is a bad idea?
A lot of it is the same situation as it always is. People watching someone play against a strategy they have never seen before, in the way that the strategy is designed to exploit, and then freaking out about it.
You don't see it a lot in high level games because it's new. It's POSSIBLE you don't see it because it's actually really terrible if the player you are using it against knows about it. It's POSSIBLE that all it would have taken to beat it was for Stephano to park his army outside of qxc's base and never let the mines or a CC out. There is no way of knowing yet.
|
Zergs are really good at dealing with widow mines right now in the GSL. Still, if a player isn't careful they can suddenly lose their whole army in one second.
I don't particularly mind strong abilities in the game. They allow players to come back from a deficit, something we don't see too often in SC2. I personally wish there were more OP abilities in the game overall, which would make a much more dynamic, back-and-forth battle for the spectator.
|
On April 12 2013 08:29 Benjamin99 wrote:Show nested quote +On April 12 2013 08:24 FallDownMarigold wrote:
I have a question about the QXC widow mine stuff, based on some of the posts here. So apparently it's really good in TVZ to do what QXC does, based on his games vs Idra + Stephano, in which he crushes them. My question is if it is simply an amazing strat, and not the result of some other factors at play, then why did I not see this strat employed in high level games, such as GSL quals and other Korean pro games? Do the Terrans at that level simply know less than QXC, or is QXC's success only possible vs. players like Idra + Stephano?
Excuse my ignorance, I'm just kinda lost. Should probably start watching more HOTS in general. If its possible against Stephano is possible against anyone Stephano is one of the very best zergs on the planet and has been for a long time. But the thing with Stephano it might work 1 time only. He is very good at coming up with strategies against tactics like this but we will see
Stephano hasn't won anything in months, and when he was dominant KESPA players hadn't doubled the amount of code S caliber koreans yet. Furthermore, he chose to live out his current life of playing the same 5 players on the NA ladder every day instead of training in Korea. Like that other guy said, he doesn't even feel the need to use more than 5 control groups. Look instead at all the GSL groups that play this week. Real top zergs are doing just fine against top terrans.
True vs Fantasy
Curious vs Marine King Prime
Roro vs Bomber
and in GSTL Scarlett vs MVP
All these games feature widow mine and medivac turbo use yet the terrans are still defeated.
The standards for micro have changed, and creative players like true have innovated by reviving use of burrowed banelings and nydus worms. It's too early to cry OP when so many new strategies are still developing.
|
We should wait for more numbers. I have looked win/loss ratio's on anguliac website. T seems favored against Z (55%) and P (58%). So yeah, they do seem a bit overpowered. My guess is it is not the mines, but the speedvac. Mines in tvp are not used so much as against Z, still winrates favor terrans.
If those percentages are correct : Leave mine as it is, change speedvac. Mines are not a gamechanger.
|
this sounds like a QQ thread to me. And to the OP, can I ask what race you play? I agree with some posters saying that this thread maybe a fear mongering thread and a little biased.
Based on watching GSL and SPL, widow mines are strong although as long as you are prepared, its easy to counter by just sending in a zergling and make sure you always have an overseer.
Widow mines are strong and yet the terran performance in GSL or SPL, it's a little disappointing actually.
And so what yeah whats the point of Widow mines being effective against zerg? is the OP saying it should not be effective?
|
On April 12 2013 09:30 lowercase wrote:
Zergs are really good at dealing with widow mines right now in the GSL. Still, if a player isn't careful they can suddenly lose their whole army in one second.
I don't particularly mind strong abilities in the game. They allow players to come back from a deficit, something we don't see too often in SC2. I personally wish there were more OP abilities in the game overall, which would make a much more dynamic, back-and-forth battle for the spectator.
Hey great points. Same here I agree with you that OP abilities are great in the game overall. And I noticed too that zergs in GSL are good at dealing with widow mines.
Fantasy destroyed by True. Life destroyed Flash.
I just wish people stop complaining or hint about abilities being too strong or too weak. LETS JUST PLAY THE FREAKIN GAME!
|
The most important fact that a lot of the lowest common denominator player base seems to forget is that there is actually a very wide realm between a unit being "too weak" and "too strong." They mistakenly believe every unit walks this razor's edge between underpowered and overpowered. In fact, there is a huge gap between those two possibilities.
The key to avoiding a unit being underpowered is to make the unit useful. If there is another unit available which is strictly better, the unit will never be used. And the key to avoiding a unit being overpowered is to make the opponent able to mitigate the unit through counterplay. Its actual effectiveness is not actually that important. If mines are better then they will be used more frequently, and consequently being skilled at mine counterplay becomes relatively more significant for the opponent.
The widow mine is an excellently designed unit in this regard. It can be useful, but the opponent can mitigate it through counterplay. Even if you buffed or nerfed mines significantly they would still fall within those parameters, but it would change how often and in what capacity mines would be used.
However a unit like the Marauder is a completely different animal. Its behavior is entirely determined by its stats, with no real counterplay possible except to have your own stat-based units to fight them. If the enemy has 50 marauders, there isn't really counterplay except to already have (or swiftly acquire) units to counter them. If you gave a Marauder a bunch more HP and damage, the result would be obviously overpowered since an A-move with a group would not be counterable by the opponent. 50 mines are a completely different matter; you can split units, micro detection and a ranged unit to chip away at them, etc. There are widely different possible outcomes for the same battle. Marauders... not so much. This is why a lot of players hate Marauders, hate Thors, hate Immortals and Colossi, hate Roaches and Infestors, and want back many Brood War units which resulted in vastly less deterministic battles. Deterministic battles are controlled by numbers and composition, both of which force players to keep their forces together to maximize their absolute strength, and act against skilled micro and multitasking of many small groups across the map.
The mine is a well-designed unit in terms of creating non-deterministic battles based on positional play. But honestly this should be the Siege Tank's province. Compare the tank and the mine. For all practical purposes, the mine is a strictly superior unit at the same roles as the tank. It can protect marines from banelings, it deals more damage to Protoss armies more quickly. It can stop the enemy from attacking into a position far more effectively. It costs less, it deals more damage with more splash, it burrows, it sieges quicker, it can be reactored, and it hits air units. Because mines don't make the game unwinnable for zerg and protoss (counterplay possible) then we can infer tanks are underpowered. At any point when you would get tanks, you should get mines instead, and this should be changed.
|
On April 13 2013 20:33 ishmoks wrote:
this sounds like a QQ thread to me. And to the OP, can I ask what race you play?
I play Protoss, gold league. I have no preconceived notion about widow mine or anything else. I want Starcraft to be as awesome as possible. I expect there is some optimal army % of widow mines for T to have vs Z, and I'm interested to know what their win% will be when have that %. I find the 65% winrate interesting, but I agree with those who point out there's not enough data to draw any drastic conclusions from that.
My points, and my only points, were those I made in my original post, in my followups, and here. I am serious about getting peer review before making future stat posts, which one person has been kind enough to volunteer for so far. There are a few hundred thousand replays in GGTracker, and I'm sure the aggregated stats there have interesting things to tell us. I've done statistics professionally for automated trading for several years so I know what I'm doing with stats.
I concede that my handling of the widow mine stats was too casual; someone asked a question and I just answered it without preparing a doctoral defense level intellectual infrastructure around it, nor do I do randomized double-blind studies. I'll definitely be more careful in the future with stats, and again would love to get some folks lined up for pre-publication review so that the community discussion is as constructive as possible.
Also, I'd be happy to expose data and tools to others if they are interested.
|
On April 12 2013 07:17 SC2Frozen wrote:OK, this thread has had enough of the "this is not significant" BS. Let us pull back the foreskin of ignorance and apply the steel wool of enlightenment. For those of you needing a little refresher on their stats (or would like to learn something new): http://en.wikipedia.org/wiki/Binomial_testSo, we have, at masters level, 345 games with more than 10 widow mines made @ 10 minutes, with a win rate of 60% (where we would expect 50%). Fires up R, dum-de-dum: Show nested quote +
> binom.test(0.6*345,345,0.5,alternative="greater")
Exact binomial test
data: 0.6 * 345 and 345
number of successes = 207, number of trials = 345, p-value = 0.0001203
alternative hypothesis: true probability of success is greater than 0.5
95 percent confidence interval:
0.5546467 1.0000000
sample estimates:
probability of success
0.6
So, given the sample, the probability that there is actually a 50% winrate if more than 10 widow mines have been produced @ 10 minutes is 0.000123 (translation: really fuckin' small). There is a 95% probability that the winrate is between 0.5546 and 1.000. Note, this does not say that the widow mine production caused the win, but it pretty definitely shows that in games where the WMs were produced in these volumes, the winrate was very likely higher than 55%.
You sound pretty confident here, so it's really bizarre to me that you seem to have forgotten that a simple binomial test assumes random sampling - which is violated if there's significant selection bias at work. It also doesn't rule out the possibility of confounding factors, both of which are the primary objections to the analysis performed in this thread.
I think you'd be well-served by qualifying your statements more fully in the future. It's tempting to throw yourself out there and pretend to be super confident in your conclusions -- and I totally understand that sometimes you may "have" to do so to be taken seriously -- but you do yourself and the entire field of statistics a disservice by understating your potential errors.
|
On April 14 2013 08:45 shaldengeki wrote:Show nested quote +On April 12 2013 07:17 SC2Frozen wrote:On April 09 2013 23:17 mau5mat wrote:
Terrible statistical testing OK, this thread has had enough of the "this is not significant" BS. Let us pull back the foreskin of ignorance and apply the steel wool of enlightenment. For those of you needing a little refresher on their stats (or would like to learn something new): http://en.wikipedia.org/wiki/Binomial_testSo, we have, at masters level, 345 games with more than 10 widow mines made @ 10 minutes, with a win rate of 60% (where we would expect 50%). Fires up R, dum-de-dum:
> binom.test(0.6*345,345,0.5,alternative="greater")
Exact binomial test
data: 0.6 * 345 and 345
number of successes = 207, number of trials = 345, p-value = 0.0001203
alternative hypothesis: true probability of success is greater than 0.5
95 percent confidence interval:
0.5546467 1.0000000
sample estimates:
probability of success
0.6
So, given the sample, the probability that there is actually a 50% winrate if more than 10 widow mines have been produced @ 10 minutes is 0.000123 (translation: really fuckin' small). There is a 95% probability that the winrate is between 0.5546 and 1.000. Note, this does not say that the widow mine production caused the win, but it pretty definitely shows that in games where the WMs were produced in these volumes, the winrate was very likely higher than 55%. You sound pretty confident here, so it's really bizarre to me that you seem to have forgotten that a simple binomial test assumes random sampling - which is violated if there's significant selection bias at work. It also doesn't rule out the possibility of confounding factors, both of which are the primary objections to the analysis performed in this thread. I think you'd be well-served by qualifying your statements more fully in the future. It's tempting to throw yourself out there and pretend to be super confident in your conclusions -- and I totally understand that sometimes you may "have" to do so to be taken seriously -- but you do yourself and the entire field of statistics a disservice by understating your potential errors.
Thanks for the reasoned insight. Given the size of the sample and the way ggtracker works (primarily auto-uploading), I would think the selection bias would be low, however that could be wrong. Further data analysis could test this hypothesis, and that was obviously not done.
Confounding factors (in my understanding) are only relevant if you are trying to show cause, which I'm explicitly not doing. These winrates could be produced because every one of these games had banshees (or speedvacs, or lagged zergs), it would not change the facts as stated above ("in games where the WMs were produced in these volumes, the winrate was very likely higher than 55%.").
|
On April 12 2013 09:30 lowercase wrote:
Zergs are really good at dealing with widow mines right now in the GSL. Still, if a player isn't careful they can suddenly lose their whole army in one second.
I don't particularly mind strong abilities in the game. They allow players to come back from a deficit, something we don't see too often in SC2. I personally wish there were more OP abilities in the game overall, which would make a much more dynamic, back-and-forth battle for the spectator.
/\ This
Part of what made BW so great was that so many things were "OP". Old Psi Storm was insane, one big reaver shot in a mineral line could change the game. Lurkers could wipe out an entire Marine and Medic force in seconds. All of these required significant effort to counter and play against, which made games more exciting, and games could change dramatically in a moment (and not because of an anti climatic death ball engagement but because of lethal force applied at the right moment).
If widow mines are "OP" that's not a bad thing. Making the game more challenging and dynamic and making comebacks possible is good for everyone. What is needed is not a nerf to these mechanics but ways for other players to micro against them.
|
Korea (South)1306 Posts
Something nobody has addressed is that when you bring up the argument that BW had OP things, you have to note that ALL of them were manned skill-shot abilities as were the dodging splits necessary to micro against them. Mines are not a skill shot. They are automated killing machines available incredibly quick with no fixed answer. They're just plain far too powerful for their cost and mobility. There's no downside, no "catch" to them.
|
I think one of the biggest problems of mines is the fact, that they don't do overkill when you don't command them to.
This allows terrans to bring a lot of widow mines to a fight, burrow all of them at the same place and watch magic happen.
If you want to break a siege tank line as Zerg, you can throw one IT egg to tank the tank shots, since ALL tanks will fire at it, resulting in an overkill and a short period of time in which Zerg can push forward.
Against widow mines, Zerg can't do such a thing. Throwing one IT egg will trigger one mine, that's it. Bringing in 10 zerglings will trigger 10 mines, but 10 zerglings will die instantly against bio/mech support behind the mines.
You just can't avoid the mines if they lie in a path you have to take to get to your enemy. Either you try kill the mines with hydras/fungals, possibly resulting in high losses of resources/energy, or you can't fight, that's it.
|
TLADT24920 Posts
Widow mines remind me of spider mines in the sense that you can pick them off from a distance. In TvP, the terran player would lay a ton of spider mines out and then the protoss will clean them up with goons+observer. I figure eventually zerg players will start using hydras+overseer to pick them off. Keep in mind that widow mines take up 2 supply and once they fire once, need time to recharge unlike spider mines. Do I think widow mines are OP? Maybe a bit and to a degree, I still prefer the siege tanks to deal with units. The fact that I don't have to wait a whole minute for it to fire again and that a widow mine can be triggered by anything as long as that unit isn't killed before the mine is triggered, means that it's deadweight supply after the initial firing if the opponent is attentive enough to pick it off as it recharges.
In terms of the OP, well, I think a lot more stats are needed. I could care less about NA GM level lol. Take a look at GSL and PL games, look at # of mines made, compile that and let's see what they say. Either way, as mentioned, I would prefer if there were some OP units in the game like BW. Nothing like seeing mass tanks sieging up at once, storm drops, reaver drops, plagues going down on MnM etc... and watch the opposing player deal with it well 
|
On April 08 2013 10:25 BronzeKnee wrote:Show nested quote +On April 08 2013 10:13 Entirety wrote:On April 08 2013 10:09 denzelz wrote:On April 08 2013 10:06 BaaL` wrote:
Pretty big selection bias? Games where T is winning already are much more likely to have 10+ widow mines, as the pushes would not get shut down and so the mines never get mopped up.
A 10% difference is nothing really if you consider any statistical effects. Statistical effects doesn't play into this since the data was not sampled from a larger set. You could say that the GM level games does not have enough data points to justify a conclusion, but it's a 10% difference across all levels. Do you understand statistics? He is not talking about the sample size, he is mentioning that games in which the Terran has 10+ Widow Mines usually means the Terran is already winning. That's the essence of sampling bias. What if I went to Wings of Liberty, found all the games where the Zerg has 30 Brood Lords and calculated the win rate? It would probably be 95% win rate for the Zerg. This tells us nothing because a Zerg with 30 Brood Lords has already won. Unless the players played "no rush 30 minutes", no Zerg player is able to produce 30 Brood Lords without a significant advantage. Okay first, this isn't about a moment when the Terran has 10+ Widow as BaaL' suggests, it is when Terran produces 10 or more Widow Mines in a game. So the Terran pushes might be "getting shut down" left and right and the Terran may have no more than 1 Widow Mine at a time, but that is irrelevant, it only focuses on production. In other words, the focus is correctly on whether or not the Terran is committing to Widow Mines. And so there is no "selection bias" as he suggests. Second, comparing Widow Mines to Brood Lords is silly. 10 Widow Mines cost exactly as much as 10 Roaches, which 750-250. It isn't a lot over a the span of a game, and they can be Reactored. So it isn't like "well he has 10 Widow Mines, thus the game is already won." So what we see is that people who make a lot of Widow Mines (10+) win 10% more games. That is pretty impressive.
There is still selection bias.
For example... someone making 10 Widow Mines in a 1 hour game... not as much as someone making 10 Widow Mines in a 20 minutes game. Having high win rate with 10 mines+ could suggest that TvZ is favoured in Terran's direction in long games as opposed to shorter ones (just saying it's a possibility, not that it is actual).
|
I think there are two problems with the widow mine. The first being the fact that it is a no skill, very strong unit, much like the infestor in WOL. The widow mine is guarenteed damage. There is absolutely no reason not to make them in t v z.Terrans can place them in random spots on the map, forget about them and and the they will do damage to a pack of lings that euns by them They completely shut down ling runbys and do so extremly cost efficiently, for 75 mins and 25 gas, your third is protected form both lings and mutas. They also shut down all forms of zerg aggresion in the early game. Zerg has to macro, which makes gamplay stale, when everygame is a no rush 20.
For the viewer I find the game extremely boring to watch when widow mines come into play. FYI I play zerg, and I used to love the art of t vs z specifically muta ling bane vs marine tank. The marine splits, the ling flanks, mutas focusing tanks, tanks focusing blings. Then infestors happened and we lost that. And now its widow mines, where you can make insane amount of them, and be safe and they decimate everything. It just seems to have lost the greatness that it once was.
|
On April 15 2013 08:27 SC2Frozen wrote:Show nested quote +On April 14 2013 08:45 shaldengeki wrote:On April 12 2013 07:17 SC2Frozen wrote:On April 09 2013 23:17 mau5mat wrote:
Terrible statistical testing OK, this thread has had enough of the "this is not significant" BS. Let us pull back the foreskin of ignorance and apply the steel wool of enlightenment. For those of you needing a little refresher on their stats (or would like to learn something new): http://en.wikipedia.org/wiki/Binomial_testSo, we have, at masters level, 345 games with more than 10 widow mines made @ 10 minutes, with a win rate of 60% (where we would expect 50%). Fires up R, dum-de-dum:
> binom.test(0.6*345,345,0.5,alternative="greater")
Exact binomial test
data: 0.6 * 345 and 345
number of successes = 207, number of trials = 345, p-value = 0.0001203
alternative hypothesis: true probability of success is greater than 0.5
95 percent confidence interval:
0.5546467 1.0000000
sample estimates:
probability of success
0.6
So, given the sample, the probability that there is actually a 50% winrate if more than 10 widow mines have been produced @ 10 minutes is 0.000123 (translation: really fuckin' small). There is a 95% probability that the winrate is between 0.5546 and 1.000. Note, this does not say that the widow mine production caused the win, but it pretty definitely shows that in games where the WMs were produced in these volumes, the winrate was very likely higher than 55%. You sound pretty confident here, so it's really bizarre to me that you seem to have forgotten that a simple binomial test assumes random sampling - which is violated if there's significant selection bias at work. It also doesn't rule out the possibility of confounding factors, both of which are the primary objections to the analysis performed in this thread. I think you'd be well-served by qualifying your statements more fully in the future. It's tempting to throw yourself out there and pretend to be super confident in your conclusions -- and I totally understand that sometimes you may "have" to do so to be taken seriously -- but you do yourself and the entire field of statistics a disservice by understating your potential errors. Thanks for the reasoned insight. Given the size of the sample and the way ggtracker works (primarily auto-uploading), I would think the selection bias would be low, however that could be wrong. Further data analysis could test this hypothesis, and that was obviously not done. Confounding factors (in my understanding) are only relevant if you are trying to show cause, which I'm explicitly not doing. These winrates could be produced because every one of these games had banshees (or speedvacs, or lagged zergs), it would not change the facts as stated above ("in games where the WMs were produced in these volumes, the winrate was very likely higher than 55%.").
With regards to selection bias, I do think that:
- the population of masters players who have gone to the trouble of setting up auto-uploading to GGTracker is unlikely to be representative of masters players as a whole (though to what extent and in what direction this effect might lie I don't pretend to know),
- for players who haven't uploaded all of their replays, the games that were important or interesting enough to merit the trouble of uploading to GGTracker are unlikely to be representative of your typical masters game. Knowing the proportion of replays that fall into this category would be useful.
So I take issue with the assertion that the numbers presented show that "in [presumably all masters] games where the WMs were produced in these volumes, the winrate was very likely higher than 55%". If you instead want to restrict the claim to just this set of games, obviously that's true.
I brought up confounding factors because the OP presents the relationship between the two as specifically meaningful in a significant way, although it does state that it doesn't prove anything. And I think we ought to be interested primarily in "meaningful" relationships; that is, if you're implying that there being a spurious relationship between the two doesn't detract from the importance of the analysis performed here, I think you're probably (nearly) alone in that sentiment. So I do believe that the (in my opinion likely) possibility that there are confounding variables at work here is a problem if you want to say that the analysis performed is particularly meaningful. But that's just my opinion of what "meaningful" is!
|
You're spending all your time trying to justify the validity of your sample set and ignoring all posts that give you examples of top Korean players dealing with widow mines with finesse.
|
|
|
|
|
|
EPT
