Don't worry about the statistics trolls, they'll say anything to show that deflect activates more than it does.
Deflect has a chance of activating 43.75% on each hit regardless if the last 20 hits in a row were deflected.
Fallen Sword Mythbusters: Part 1 - Deflect
Started by RJEM, Jan 13 2011 03:40
101 replies to this topic
#61
Posted 19 January 2011 - 12:46
Deflect has a chance of activating 43.75% on each hit regardless if the last 20 hits in a row were deflected.
#62
Posted 19 January 2011 - 12:54
43.75% is relevant and all that matters. Anyone who understand odds will understand that sometimes it'll be 10 hits on a 10 hit bounty and other times it'll be 20.
#63
Posted 19 January 2011 - 13:05
43.75% is relevant and all that matters. Anyone who understand odds will understand that sometimes it'll be 10 hits on a 10 hit bounty and other times it'll be 20.
Whose to say you wont get a bounty where youll need 30 attacks to clear it? What if you get 4 of those in a row? There is a probability of that happening. Op is stating probability as fact, when in fact its all just a probability.
#64
fs_nthnclls
fs_nthnclls
-
- Guests
Posted 19 January 2011 - 13:10
43.75% is relevant and all that matters. Anyone who understand odds will understand that sometimes it'll be 10 hits on a 10 hit bounty and other times it'll be 20.
And other times you'll get 20 goddammed deflects on one bounty...(not that that's ever happened or anything).
I think we all get the point of RJEM's first post. The average over time will probably be close to 43.75%.
That being said, truckdrift, you're wrong because the RNG cannot be completely random, if I understand this correctly. If this were real life and we were flipping coins to find out who was deflected, you would be right, but this involves computer generated randomness.
#65
fs_coyotik
fs_coyotik
-
- Guests
Posted 19 January 2011 - 13:36
but you CAN be confident that over MANY more bounties you will have no worse luck than the average player.
As long as you believe that the RNG is getting seeded by something truly random, i.e. definitely NOT something like player_id + current_time
.And I won't believe that until Hoof swears that he grepped the whole codebase for RNG seeding.
#66
Posted 19 January 2011 - 14:00
The OP is stating what his results were based on "A" sample. The sample is representative of what SHOULD happen. The larger the sample the better the representation. Deflect works as intended, annoying or saving whichever perspective you have. Personally I like the skill.Op is stating probability as fact, when in fact its all just a probability.
[Signature removed]
“When you tear out a man's tongue, you are not proving him a liar, you're only telling the world that you fear what he might say.” -GRRM
#67
Posted 19 January 2011 - 14:12
truckdrift - I really don't get your point. Everyone knows the chance of deflect activating on any given attack is 43.75%. Everyone knows you can flip 25 heads in a row, but it isn't very likely.
I am stating 'as fact' that the statistical likelihood of the deflect rate in this game, as it was programmed when I did my sample, being over 50% as being EXTREMELY low - so low that it would be valid to rule this out scientifically. Nothing more, nothing less.
A run of bad luck does not guarantee a run of good.
I'm interested that you agree with me that over time the average will be close to the stated rate and yet continue to argue about what I'm saying. The myth I was 'busting' was that deflect activated more than 50% of the time - it doesn't. That fact that a few horrible bounties happen now and again is neither here nor there.
Sure, you can get worked up over 4 30 hit bounties in a row - I would be frustrated. There is a small chance it would happen and no, you can't say anything about the next bounty because of that. However, like I demonstrated, any subsequent bounty at the normal activation rate will cause the average over time to tend towards 43.75%. That is a fact, not a probability.
By the way, on the 'time and effort' front, it doesn't matter where the deflects fall in the 'whole sample' - the time taken is the same overall. It has to be, because you did 10000 attacks overall. If you break it down into lines of 10 successful hits then you'll have several of 25+, but some of only 10 or 11 to go along with it. They're just as important and you're neglecting that point.
I am stating 'as fact' that the statistical likelihood of the deflect rate in this game, as it was programmed when I did my sample, being over 50% as being EXTREMELY low - so low that it would be valid to rule this out scientifically. Nothing more, nothing less.
A run of bad luck does not guarantee a run of good.
I'm interested that you agree with me that over time the average will be close to the stated rate and yet continue to argue about what I'm saying. The myth I was 'busting' was that deflect activated more than 50% of the time - it doesn't. That fact that a few horrible bounties happen now and again is neither here nor there.
Sure, you can get worked up over 4 30 hit bounties in a row - I would be frustrated. There is a small chance it would happen and no, you can't say anything about the next bounty because of that. However, like I demonstrated, any subsequent bounty at the normal activation rate will cause the average over time to tend towards 43.75%. That is a fact, not a probability.
By the way, on the 'time and effort' front, it doesn't matter where the deflects fall in the 'whole sample' - the time taken is the same overall. It has to be, because you did 10000 attacks overall. If you break it down into lines of 10 successful hits then you'll have several of 25+, but some of only 10 or 11 to go along with it. They're just as important and you're neglecting that point.
#68
Posted 19 January 2011 - 14:22
RJEM, the point of this thread is obvious to the knowledgeable player. Some, however, don't understand the basics of statistics. Your point was well made and factual. Deflect isn't as bad as players make it out to be but rather annoying, depending on your perspective.
[Signature removed]
“When you tear out a man's tongue, you are not proving him a liar, you're only telling the world that you fear what he might say.” -GRRM
#69
Posted 19 January 2011 - 14:31
RJEM, the point of this thread is obvious to the knowledgeable player. Some, however, don't understand the basics of statistics. Your point was well made and factual. Deflect isn't as bad as players make it out to be but rather annoying, depending on your perspective.
Thanks Luis - just trying to convince a couple of naysayers that there isn't a problem with what I've done.
As an interesting aside for those people:
Assumer I did 100 hits and got deflected on 60 of them
Let's ask what the probablility the next 100 DON'T lower the average towards the expected rate:
P(Successes <= 40 | 100 trials) = 0.08% with 43.73% deflect rate as the defined Deflect rate.
No problems there for me - you could be hugely, hugely unlucky and end up with 120 or more deflects from the combined 200 hits, but it's unlikely, regardless of the outcome of the first 100.
#70
Posted 19 January 2011 - 14:39
Or vice versa. These likelihoods are rare exceptions, not the norm. Getting deflected 10-15 times in a row happens. Most likely is that some one else is NOT getting deflected. Sounds like a delevel party to me. So who cares as long as xp remaining = 0.No problems there for me - you could be hugely, hugely unlucky and end up with 120 or more deflects from the combined 200 hits, but it's unlikely, regardless of the outcome of the first 100.
[Signature removed]
“When you tear out a man's tongue, you are not proving him a liar, you're only telling the world that you fear what he might say.” -GRRM
#71
fs_kashen
fs_kashen
-
- Guests
Posted 19 January 2011 - 17:23
So you proved what we already know.
0.25*175=43.75
The math is inarguable, my guff with the "debunking" is the fact is, no matter the proof of theoretical evidence the 43.75% is NOT relevant where it really matters. You and I are arguing 2 different things I think, but where your argument uses the WHOLE (which is not realistic) im trying to break it down to a incident by incident basis (which is the way PvP works), which you do not bring to a full circle anywhere in the original post. I may hit a bad run on a 200 ticket bounty, but I may hit a good run on a 1 ticket bounty, 2 entirely different situations. Or vise versa, making the 200 ticket bounty pretty much a waste of time. Your only accounting for stats, not time and effort, thats where the gamblers logic im talking about comes in. Your numbers mean absolutely nothing to me, at any time. Im not going to look at 43.75% out of 10000 attacks because the 43.75% does not matter in such a broad spectrum. Im looking at a hit by hit rate, because i put time and effort into it (i honestly dont mean to disrespect your research in that aspect). Youre looking at a bounty clear as one in a line of numbers, but when only 10 kills at a time are involved, 43.75 doesnt mean much. Its sets of 10 you need to look at, but also with that if you line up 100 sets of 10 43.75% WILL show up, as it should. BUT the 10 kills and 20 deflects will happen, and thats what im talkin about. No matter if the odds look like i shouldnt get deflected, it might or might not. 43.75% is absolute through a infinite number, but 43.75% in 10 hits means very very little. Each bounty will have a different % of deflected attacks, and you could never try to calculate those numbers (its just impossible).
Your comparison of runs is exactly the point im trying to get across, your first run could be 60%, and the second could also be 60%. if it happened in one bounty whats to say it wont happen again ( or 4 times in a row) there is a probability of it happening. Now if those 4 were counted into 100 bounty clears, your numbers would still come up on top, but those few bounties i would be straight up screwed. You see what im getting at? this + this doesnt necessarily mean that.
You seem to enjoy writing long passages of text explaining things that most people in this thread, and definately OP, already know and find to be pretty trivial. However, your first sentence is wrong and it is also the reason for the test. There are plenty of players that think that the activation rate of Deflect is something different than 43.75%, that there is an error in the game code.
#72
Posted 19 January 2011 - 19:59
Congratulations OP!! You proved HCS's numbers are correct -statistically speaking- while at the same time showing the Myth to be true. .....
Sorry Michael - this post is totally wrong.
I got 31 deflects on 81 attacks (50 of which were successful), not 31 on 50. The myth is false.
i stand by my post.
#73
Posted 19 January 2011 - 20:53
The myth is true that deflect could activate more than 43.75% on any bounty at any time.
It could also activate less.
As I said, I do not care about an infinite amount of hits only the 10 that will make me some gold/FSP.
You dont care about the 10 hits, only the infinite number that will prove you right.
In this respect we are both right, deflect as a whole works the way it should, but during a bounty it can activate more than the average (or less).
I feel the average of 175 is fine, it may be a lil high at 43.75%, but personal bias makes me feel that way. I absolutely DO NOT want to see deflect 200.
It could also activate less.
As I said, I do not care about an infinite amount of hits only the 10 that will make me some gold/FSP.
You dont care about the 10 hits, only the infinite number that will prove you right.
In this respect we are both right, deflect as a whole works the way it should, but during a bounty it can activate more than the average (or less).
I feel the average of 175 is fine, it may be a lil high at 43.75%, but personal bias makes me feel that way. I absolutely DO NOT want to see deflect 200.
#74
Posted 19 January 2011 - 21:29
About the 43%:
Most people would assume 43/100 means 21.5/50, 21.5 + 50 = 71.5 OR expecting 71 or 72 attempts on average in GvG.
What a statistician would do is:
43% = .43 = d/(50+d) where d is number of deflects, getting d = 37.7 meaning 37.7 + 50 = 87.7
So, layperson expects 72, statistician expects 87.7 attempts to complete on average.
Like i said i stand by my post.
About the losers/winners:
The "roll" may be evenly distributed but people and there experiences are not. So, assessing a normal curve over all players in FS we would expect losers and winners. The idea that one person always happens to invent at "wrong" time would be expected, and applies to winners.
About red then black:
If ten red, then yes next play black has no better chance. Interestingly enough, red has same chance as black.
About RNG:
I actually like individual RNG for things like inventing, but not as Coyotik puts it at putting a player at a disadvantage. i.e. i agree with coyotik - coyotik may not agree with me.
About feel:
People get comfortable about outcomes ("rolls"), if the cows do anything drastic to RNG it may cost FS players.
About losers/winners:
The cows can put in parameters to end lucky and or unlucky streaks.
Most people would assume 43/100 means 21.5/50, 21.5 + 50 = 71.5 OR expecting 71 or 72 attempts on average in GvG.
What a statistician would do is:
43% = .43 = d/(50+d) where d is number of deflects, getting d = 37.7 meaning 37.7 + 50 = 87.7
So, layperson expects 72, statistician expects 87.7 attempts to complete on average.
Like i said i stand by my post.
About the losers/winners:
The "roll" may be evenly distributed but people and there experiences are not. So, assessing a normal curve over all players in FS we would expect losers and winners. The idea that one person always happens to invent at "wrong" time would be expected, and applies to winners.
About red then black:
If ten red, then yes next play black has no better chance. Interestingly enough, red has same chance as black.
About RNG:
I actually like individual RNG for things like inventing, but not as Coyotik puts it at putting a player at a disadvantage. i.e. i agree with coyotik - coyotik may not agree with me.
About feel:
People get comfortable about outcomes ("rolls"), if the cows do anything drastic to RNG it may cost FS players.
About losers/winners:
The cows can put in parameters to end lucky and or unlucky streaks.
#75
Posted 19 January 2011 - 21:38
The myth is true that deflect could activate more than 43.75% on any bounty at any time.
It could also activate less.
As I said, I do not care about an infinite amount of hits only the 10 that will make me some gold/FSP.
You dont care about the 10 hits, only the infinite number that will prove you right.
In this respect we are both right, deflect as a whole works the way it should, but during a bounty it can activate more than the average (or less).
I feel the average of 175 is fine, it may be a lil high at 43.75%, but personal bias makes me feel that way. I absolutely DO NOT want to see deflect 200.
What the hell are you even going on about?
The myth RJEM showed not to be true is that deflect's activation rate is higher than what the game suggests. 10 hit, 100 hit or 1000 hit, the chance of deflect activating on any single hit is what is displayed in the game. This isn't a discussion on deflect of its potions at all, so stop trolling.
#76
Posted 19 January 2011 - 22:29
Congratulations OP!! You proved HCS's numbers are correct -statistically speaking- while at the same time showing the Myth to be true. .....
Sorry Michael - this post is totally wrong.
I got 31 deflects on 81 attacks (50 of which were successful), not 31 on 50. The myth is false.
i stand by my post.
Michael, my exact words in the OP are:
"50 successful attacks required just 81 attempts"
I don't care if you're a layperson or a maths professor - if English is your first language it should be pretty clear that I did 81 attacks, not 50, and that 31 were deflected. If you can't see that then the rest of the post is bound not to make sense.
#77
Posted 19 January 2011 - 22:41
The myth is true that deflect could activate more than 43.75% on any bounty at any time.
It could also activate less.
As I said, I do not care about an infinite amount of hits only the 10 that will make me some gold/FSP.
You dont care about the 10 hits, only the infinite number that will prove you right.
In this respect we are both right, deflect as a whole works the way it should, but during a bounty it can activate more than the average (or less).
I feel the average of 175 is fine, it may be a lil high at 43.75%, but personal bias makes me feel that way. I absolutely DO NOT want to see deflect 200.
So when it boils down to it my original post is absolutely fine, doesn't fall into the gambler's fallacy, and states exactly what you just stated.
FYI, X successful hits in a row with no deflect should happen more often than X deflects by a factor of 1.3^X - so if you're looking for which is more likely, a deflect free bounty or one in which your first ten are deflected then it's the former (and by a factor of 12 too).
#78
Posted 20 January 2011 - 00:24
Next up, I'll be taking a look at whether FI 800 really is better than FI 1000 when scavenging. I just need to find some (lots of) gold before I start
I was going after the leaf dragon with FI1k and spent 10 fsp on gold yet never got a thing... So I started using 25k for scavenging instead of the 15k recommended by HCS and it still didn't work so i lost about 10 more fsp.. Either FI doesn't work in the caves or I'm just really unlucky...
#79
Posted 20 January 2011 - 00:31
Next up, I'll be taking a look at whether FI 800 really is better than FI 1000 when scavenging. I just need to find some (lots of) gold before I start
I was going after the leaf dragon with FI1k and spent 10 fsp on gold yet never got a thing... So I started using 25k for scavenging instead of the 15k recommended by HCS and it still didn't work so i lost about 10 more fsp.. Either FI doesn't work in the caves or I'm just really unlucky...
Sounds like a case of bad luck to me - it often takes a lot of investment in the caves to get worthwile returns, but sometimes it just isn't your day
#80
Posted 20 January 2011 - 00:35
Next up, I'll be taking a look at whether FI 800 really is better than FI 1000 when scavenging. I just need to find some (lots of) gold before I start
I was going after the leaf dragon with FI1k and spent 10 fsp on gold yet never got a thing... So I started using 25k for scavenging instead of the 15k recommended by HCS and it still didn't work so i lost about 10 more fsp.. Either FI doesn't work in the caves or I'm just really unlucky...
Sounds like a case of bad luck to me - it often takes a lot of investment in the caves to get worthwile returns, but sometimes it just isn't your day
It was my first time in the caves so i just assumed FI didn't work in them until i read this and before anyone says anything, auto discard was off lol
0 user(s) are reading this topic
0 members, 0 guests, 0 anonymous users