inherit
2422
0
46
swordfish
135
February 2016
swordfish
|
Post by swordfish on Jan 23, 2017 23:40:43 GMT
Hello, I collected data from lucky flip. I have 100 datapoints of 4x draws, and 392 of single draws. Here is the link to the Google doc right here. Now when drawing, I found I'd get times where I draw all crap or a lot of good stuff. How could I analyze the data to see if I am more likely to draw good stuff if I have drawn some already? I'm a bit dusty with data mining.
|
|
#830909
Boxing Kangaroo
3476
0
Sept 11, 2018 22:49:30 GMT
372
BK✦Singing
Friendly and fun guild. Share experience & enjoy the game together.
1,097
Sept 10, 2016 20:42:20 GMT
September 2016
singingzibra
101
Boxing Kangaroo
SingingZibra
|
Post by BK✦Singing on Jan 24, 2017 0:41:52 GMT
This is a great log, thanks a lot. I love the way you use time to value the draws, and the fact that you compared 2 different drawing methods. My $0.02: - There seems to be a missing factor: the mat pool each day, that needs to be considered for more meaningful analysis. Let's use an extreme case, say in day 1 the pool is full of gold mats, and everything in day 2 is white. The value of draws apparently diverge, even the luck is exactly the same.
- Your data set is large enough so any inaccuracy in the cost calculation should be negligible. The conclusion of 2x+refrsh+2x being better than 4x should be very helpful for many of us. A rough mathmatical explanation could be that, the cost increases faster than the chance you gain when draw 1 more time. The former increases exponentially, while the latter only increases proportionally.
- Very very minor, if the x2, x3 or whatever gives you trouble in the cost calculation, simply take them out of your data. Your sample is large enough.
Since the pool is fully revealed before any draw, I think what can be useful is to come up with a calculator, taking remaining mats as input and giving out the expected return of next draws. Use this tool we'll know cost effictiveness and how far to go each day. Of course this can't stop IGG from cheating (no conspiracy, I don't mean they cheat, just for the sake of logic), but we at least have an idea when to take reasonable risk.
|
|
inherit
2422
0
46
swordfish
135
February 2016
swordfish
|
Post by swordfish on Jan 24, 2017 2:30:02 GMT
Thanks for all the input! It's encouraging that it benefited at least one person. As to your points, I decided to ignore the mat pool 1) because it would be way too much work to record and 2) because in the grand scheme of things, it doesn't matter unless you're going to adjust how much you will buy depending on the pool. I also think the 100 data points is not enough to give accurate estimations, but it does show a general tendency. I did answer my original question though about how to draw so I am satisfied. As to the calculator idea, it would be a really fun project, but I really don't have enough gems on my alt account to flip 8 times. Although, if you set you max spending at 6 flips, You'd only have to test it up to 6. Either way, the cost to my alts outweighs any benefits I may reap from it. But based on what I saw while running the experiment, if I was gonna hit high, it was most likely to be the second draw; you actually gain less from the third draw than the second. So if you wanted to guarantee higher rewards, I would guess that you'd see a significant change starting at the 5th draw. However, I personally can't afford to spend that much per day on my account. Thank you again for all your input, and if you have any experience with data mining, I'd love to evaluate the variable drop rates.
|
|
#830909
Boxing Kangaroo
3476
0
Sept 11, 2018 22:49:30 GMT
372
BK✦Singing
Friendly and fun guild. Share experience & enjoy the game together.
1,097
Sept 10, 2016 20:42:20 GMT
September 2016
singingzibra
101
Boxing Kangaroo
SingingZibra
|
Post by BK✦Singing on Jan 24, 2017 7:19:07 GMT
Hi swordfish , For the calculator suggestion, you don't have to experiment, just need to calculate. Let me explain my idea with a real example: Say the pool starts as: 2 white, 2 green, 3 blue, 1 purple & 1 gold. So the total is : 2x2 + 2*(2*4) + 3*(2*4*4) + 1*(2*4*4*4) = 756 SH (smelt hour) Of course the first draw is free, let's assume it gets a blue. Now the rest value is: 724 SH, and there are 8 mats left. The 2nd draw costs 5 gems, its amortized value is: (2/8)*2 + (2/8)*8 + (2/8)*32 + (1/8)*128 + (1/8)*512 = 90.5 SH. So on and so force. If our luck is like: blue, white, green, blue, purple, then the cost effectiveness of each draw is: - B: 1st draw, free, the expected value/cost of next draw is (32+90)/5 = 24 SH/gem.
- G: 2nd draw, tough luck, the actual income is less than expected. The next draw will cost 10 gems and the amortized value/cost is: 102/10 = 10 SH/gem, cumulative V/C = (32+8+102)/(10+5) = 9 SH/gem
- W: 3rd draw, still the IGG luck. The next draw will be 20 gems worth. The V/C = 119/20 = 6 SH/gem, cumulative V/C = 5 SH/gem
- B: 4th draw, average. The next draw costs 40 gems, resulting V/C = 3 SH/gem, and cumulative V/C = 3 SH/gem
- P: 5th draw. Ah ha, finally a purple. Want the gold? the next draw will have V/C = 2 SH/gem, and the cumulative V/C = 2 SH/gem
I might have some miscalculation, or wrong SH value for some mats, but you get the idea. In short, this can be calculated. Sorry I know it is easier said than done. I myself is dumb @ excel sheet, so maybe it is super simple for pros.
|
|
#ff3008
1791
0
1
749
Sinister22✪NFS
1,896
October 2015
sinister22
Lucy
|
Post by Sinister22✪NFS on Jan 24, 2017 8:19:04 GMT
If you guys could dumb it down, it will be great...
|
|
inherit
1003
0
Sept 26, 2019 9:51:15 GMT
51
moonshaker
239
Jul 31, 2015 15:55:50 GMT
July 2015
moonshaker
|
Post by moonshaker on Jan 24, 2017 12:32:45 GMT
If you guys could dumb it down, it will be great... Same here ha..
|
|
#ff0c0c
You are being watched...
3717
0
May 23, 2023 16:08:13 GMT
210
ruklo
586
Nov 24, 2016 18:00:23 GMT
November 2016
ruklo
ruklo
|
Post by ruklo on Jan 24, 2017 16:05:12 GMT
will try to explain what I undertand, hope it clarifies (and correct if wrong):
OBJECTIVE: which is better: 4 flips (better reward for gems spend) OR 2 flips , refresh , 2 flips
COST: A) 4 flips = 35 gems (free + 5 G + 10 G + 20 G) B) 2 flips / refresh / 2 flips = 20 Gems (free + 5 G + 10G (refresh) + free + 5 G
HOW WAS IT DONE:
1) for each material a value is given (white:2; green:8; blue:32; purple:128; orange: 512) this is because you need 4 white to 1 green, 4 green to 1 blue, etc) geometric progression
2) flips are made in 2 modes (4 flips ///// 2 flips , refresh , 2 flips) and an excel sheet is completed (log)
3) the flips (colors) are transformed into number (see point 1) to allow an AVERAGE calculation
4) this AVERAGE is then divided by the COST to give "THE AVERAGE GEM COST FOR EACH FLIP"
5) the conclusion: the returns for spending in a 4 Flip are LESS than doing 2 refresh 2
|
|
inherit
2422
0
46
swordfish
135
February 2016
swordfish
|
Post by swordfish on Jan 24, 2017 16:46:45 GMT
Sorry about my last post. It came out as a huge text block on my phone. Anyway, ruklo is explaining my original goal, and he did a very good job at it.
Zibra was trying to expand the data and use it to try and predict what your draw would be given your pool and what you've drawn already. The problem with your method Zibra is that each of the 9 cards displayed are not yet assigned a material. Instead, as you flip one, it assigns a material to that card with a bias towards lower level materials. So in your scenario, you don't have a 1/9 chance to draw the orange. At your first draw, it would be more of a 1% chance. I know, it doesn't make any sense since it appears to be an equivalent chance; lucky spin acts the same way.
So to predict your next draw, you'd have to test to determine what the bias is. To do this, you'd need plenty of data for different material pool configuration. You'd be talking about well over 1,000 data points as well as coding a prediction algorithm. Doing it on my own would be far too time consuming and difficult.
|
|
#830909
Boxing Kangaroo
3476
0
Sept 11, 2018 22:49:30 GMT
372
BK✦Singing
Friendly and fun guild. Share experience & enjoy the game together.
1,097
Sept 10, 2016 20:42:20 GMT
September 2016
singingzibra
101
Boxing Kangaroo
SingingZibra
|
Post by BK✦Singing on Jan 24, 2017 18:50:43 GMT
Each of the 9 cards displayed are not yet assigned a material. Instead, as you flip one, it assigns a material to that card with a bias towards lower level materials. So in your scenario, you don't have a 1/9 chance to draw the orange. At your first draw, it would be more of a 1% chance. I know, it doesn't make any sense since it appears to be an equivalent chance; lucky spin acts the same way. Now I see why we think differently. If what you described is true, at least the lucy spin part is, then it is not a draw, it is I-lure-you-to-think-it-is-a-fair-game-then-I-slap-you. In a word it is cheating. They already cheated on lucky spin, so I can't defend it on this account. Too bad. This conspiracy theory actually makes your study more meaningful. BK✦GamerGurl33 before also taught me the same luck flip strategy, and now your data materialized it.
|
|
inherit
2422
0
46
swordfish
135
February 2016
swordfish
|
Post by swordfish on Jan 24, 2017 19:02:05 GMT
I've been thinking of writing a dh guide for quite some time now, but it's the little things like lucky flip and spin being rigged that's just hard to teach. All of the features have their own little quirks like this that experienced players just know. It's hard to teach them to new people though because we assume they already have that basic foundation of how dh works.
|
|
#0d6ad6
2824
0
92
hetmasteen
133
May 2016
hetmasteen
|
Post by hetmasteen on Jan 25, 2017 6:58:38 GMT
First off, great job on the experiment, I really like it when people put down numbers and statistics. As for what Zibra said, I honestly thought that it was common knowledge that things like lucky flip, lucky spin, and any other similar things were all for show. Meaning the percentages for each prize are already hard coded in the game, and all the graphics that are presented to you don't mean anything. It is a similar thing like the arcade machines you see in places like Chuck E Cheese's, called Redemption Games, where sometimes a prize is guaranteed to never happen until a certain amount of players play it. This is why I believe stricter regulations concerning free to play mobile games are necessary as they begin blurring the line between normal games and gambling.
|
|
inherit
3009
0
Nov 10, 2017 13:23:50 GMT
9
hologram
88
June 2016
hologram
|
Post by hologram on Jan 25, 2017 7:32:25 GMT
It's the same as the tobaco industry. Deals with much money, and the countrys prefer to receive the income taxes than do anything about protecting the people. If tobaco it's bad for the health they could forbid the sales, like they do on other drugs (some of them far less worse that the tobaco). If tobaco gives cancer, they could at least ban the cancerous substances that it's in is composition (because the tobacco plant it self isn't cancerous). But they don't. Again, because of the income taxes. In the end, it will always by people fault if they want to spend money in a game or be addicted.
|
|
inherit
3827
0
Sept 1, 2017 13:45:38 GMT
4
klotto
13
Dec 29, 2016 15:26:04 GMT
December 2016
klotto
|
Post by klotto on Jan 29, 2017 18:01:16 GMT
You actually have fundamental error in your calculation for "single flip" draws. You using wrong cost. The thing is, you have to return the flip board to the initial state after you done flipping. I.e. you have to calculate cost of: flip (free) + flip (5) + reset (10) + flip (free) + flip (5) + reset (10). You didnt add another (last) reset into your calculation. This way someone using your system can not go on after he did first cycle.
|
|
inherit
2422
0
46
swordfish
135
February 2016
swordfish
|
Post by swordfish on Jan 29, 2017 18:51:45 GMT
You actually have fundamental error in your calculation for "single flip" draws. You using wrong cost. The thing is, you have to return the flip board to the initial state after you done flipping. I.e. you have to calculate cost of: flip (free) + flip (5) + reset (10) + flip (free) + flip (5) + reset (10). You didnt add another (last) reset into your calculation. This way someone using your system can not go on after he did first cycle. It isn't a fundamental error, but it is too narrow a focus to apply more generally. I was only thinking of resetting once so I wasn't considering the other reset at the end. I guess you have two values, one for the first two flips and then another for all flips after that. When I have the time, I'll add in a more general evaluation for you. If you want it now though, you have all the data you need; you just need to adjust the cost slightly.
|
|
inherit
3827
0
Sept 1, 2017 13:45:38 GMT
4
klotto
13
Dec 29, 2016 15:26:04 GMT
December 2016
klotto
|
Post by klotto on Jan 30, 2017 20:30:51 GMT
Actually, I just spent 2k gems doing "reset-flip-flip-flip" sequences. I averaged 2.9 hours per gem which is very good for me
|
|