Do you feel like doing 32 test runs on Grinder + Flower to complete my set of 100? Otherwise I might just exclude them from the data to keep it to round numbers - I doubt I'll complete these any time soon as I don't want any more Steam or Astro! (plus the long fail times suck)
Unless I'm misunderstanding, someone else doing 32 attempts in a different account shouldn't really be added to yours to make 100 attempts because it isn't consistent. I would think that you'd have to do all 100 attempts in the same account for it to be more accurate. Wouldn't it? Anyway, I'm super busy this week, but I will try and fit in 100 attempts with this combo for you.
Unless the % is set individually for each player, which is very improbable, it won't matter. And if it is set individually than all this testing is for naught.
Wapa: Thanks for that, no worries how long it takes - just let me know when you have completed your 32 breeding attempts, and I'll collate them with mine and update the OP.
Maidy: No it shouldn't matter at all. After all, that's exactly what this thread is doing, collating everyone's data. The only reason I suggested batches of 100 is so it is nice and neat. There is another complication which I'll explain in my next post (or might amend my first reply).
Pixie: Batches of 60 are fine. Ideally 50 or 100 for ease of collation however. There is one catch though, which is the same as what I was referring to in my reply to Maidy above. I'll post this seperately.
Just one thing I wanted to clarify with regards to this testing. It is important that you test a set number of breeding attempts, regardless of whether you get a master bot or not.
So say you are going for Nuclear, and you get it on your 123rd attempt, and then you stop there. These stats (122 vanilla, 1 nuclear) cannot be included in the results.
The reason for this is that this is actually a different probabiilty distribution and including a sample from this distribution in the overall results will skew the data.
So it doesn't really matter what sample size you select, but as long as you see that sample through (regardless of whether you get a master bot on your first attempt, or not at all). Of course, you can choose not to complete the set, but then it can't be included in the data.
I've suggested 50 or 100 breeding attempts for ease of collation.
Just one thing I wanted to clarify with regards to this testing. It is important that you test a set number of breeding attempts, regardless of whether you get a master bot or not.
So say you are going for Nuclear, and you get it on your 123rd attempt, and then you stop there. These stats (122 vanilla, 1 nuclear) cannot be included in the results.
The reason for this is that this is actually a different probabiilty distribution and including a sample from this distribution in the overall results will skew the data.
So it doesn't really matter what sample size you select, but as long as you see that sample through (regardless of whether you get a master bot on your first attempt, or not at all). Of course, you can choose not to complete the set, but then it can't be included in the data.
I've suggested 50 or 100 breeding attempts for ease of collation.
To say this in a different way, it needs to be a random selection of breeding attempts. If you end your data at the point you achieve your goal, it is not a random selection.
Just one thing I wanted to clarify with regards to this testing. It is important that you test a set number of breeding attempts, regardless of whether you get a master bot or not.
So say you are going for Nuclear, and you get it on your 123rd attempt, and then you stop there. These stats (122 vanilla, 1 nuclear) cannot be included in the results.
The reason for this is that this is actually a different probabiilty distribution and including a sample from this distribution in the overall results will skew the data.
So it doesn't really matter what sample size you select, but as long as you see that sample through (regardless of whether you get a master bot on your first attempt, or not at all). Of course, you can choose not to complete the set, but then it can't be included in the data.
I've suggested 50 or 100 breeding attempts for ease of collation.
To say this in a different way, it needs to be a random selection of breeding attempts. If you end your data at the point you achieve your goal, it is not a random selection.
So, in breeding for Double Rainbow quitting at attempt 355 when finally successful wouldn't have been a random selection? Should have continued to what number? 400?
To say this in a different way, it needs to be a random selection of breeding attempts. If you end your data at the point you achieve your goal, it is not a random selection.
So, in breeding for Double Rainbow quitting at attempt 355 when finally successful wouldn't have been a random selection? Should have continued to what number? 400?
Well, erring on the long side is not as bad as on the other. I got my first Doctor on attempt number 6, which, if I stopped there, would indicate it has a 16.6% breeding success rate. Yours indicates a 0.28% success rate. If the actual success rate is 1%, than you were out by 72%, whereas mine is out by 1660%
So, in breeding for Double Rainbow quitting at attempt 355 when finally successful wouldn't have been a random selection? Should have continued to what number? 400?
Well, erring on the long side is not as bad as on the other. I got my first Doctor on attempt number 6, which, if I stopped there, would indicate it has a 16.6% breeding success rate. Yours indicates a 0.28% success rate. If the actual success rate is 1%, than you were out by 72%, whereas mine is out by 1660%
Is the assumed 1% based on it being the same percent as plain Rainbow?
Well, erring on the long side is not as bad as on the other. I got my first Doctor on attempt number 6, which, if I stopped there, would indicate it has a 16.6% breeding success rate. Yours indicates a 0.28% success rate. If the actual success rate is 1%, than you were out by 72%, whereas mine is out by 1660%
Is the assumed 1% based on it being the same percent as plain Rainbow?
That's just a number I choose somewhat at random. If it's actually 0.5% than you are only out by 44% and I'm out by 3320%
So, in breeding for Double Rainbow quitting at attempt 355 when finally successful wouldn't have been a random selection? Should have continued to what number? 400?
No, not at all - it is definitely a random selection and a useful statistic at that. It's just that they are just different probability distributions, so when collecting samples (to test the unknown probability of the distribution), all samples need to be collected from the same distribution.
Quick maths lesson for anyone interested.
Binomial distribution
Taking samples of say 100 breeding attempts and recording the number of epic bots can be tested using a binomial distribution, where the probability of success (a master bot) is p%. The overall statistics can be used to estimate this (i.e. total number of thermals from total number of breeding results that COULD result in a thermal, regardless if other master bots were also possible). We can then also test if the breeding percentages are cumulative by:
a) First establishing what the likely value of p is for each 'lesser' master bot (should be the same - if they are similar, we could reasonably assume this) b) Testing if the breeding success rate of ANY master bot is indeed the sum of the individual probabilities. (e.g. Say Astro = 1%, and thermal=steam=2% - Then we can look at the samples that can produce all three of these (e.g. Grinder + Botanist ) and see if they produce ANY of these bots ~5% of the time)
Geometric Distribution
Recording the time it takes until your first success can be tested using a geometric distribution (this distribution calculates the probability that it takes x number of attempts until you get your first success).
This is another way we can predict what the probability of success is, but as mentioned earlier, it is a distinct probability distribution and you can't take samples from both to test a hypothesis.
There is another issue with this in that the first 'results' we'll get will inevitably be from people who get the success earlier. It is just simpler to analyse fixed samples (via the binomial distribution) as it eliminates any potential reporting bias.
Anything else you want tested? Otherwise I'll just randomly scroll through my list and choose 2 bots.
I love my vine dragon avatar created by ElektronX. Now, if only BFS would add it to the game...
DV friend ID: Likylind#5293 Baby park: DORULIND#1400
'... overwhelming goodwill, and witty humour, profound grace and compassion blended with the spice of relentless taunts.'
Note: This is an independent forum and is not affiliated with Deca Games, the developer of DragonVale.
It was created by and is managed by DragonVale fanatics.
All who observe the following brief guidelines are welcome to this great community: no personal attacks & no spam.
