I don't think it needs it, but BUMP!
Awsome idea, finally getting some clarification on the drop rates of the boxes. I'll start noting what I get from mine.
I don't think it needs it, but BUMP!
Awsome idea, finally getting some clarification on the drop rates of the boxes. I'll start noting what I get from mine.
EGO rating 3044. My head goes *WUB, WUB, WUB, WUB, WUB*
X-BOX
ive opened a few tier 2 dont remember what i got but also got 12 tier 4 lockboxes. with all blue. no green, no purple, no roange. just blue
This is the part of your argument that I disagree with. A T4 give you exactly 4 trials at 8% each trial. A T3 gives you exactly 3 trials at 6% per trial. A T2 gives you exactly 2 trials at 4% per trial.Originally Posted by Stickasylum
72 Keycodes will therefore yield 4 trials at T4, 9 trials at T3, and 18 trials at T2.
We have all the exact numbers we need to calculate exact cumulative probability in each scenario per 72 available keys to spend, and that tells a very clear and accurate story: you have the best chance to score at least one orange by buying only T2 lockboxes, but you'll see very few purples and greens in the process. You have a lower chance to score oranges with either T3 or T4, but you'll see more purples and blues in the process.
IMO, the "cost per" measure does not tell nearly as clear of a story.
Moat people see plenty of good purples show up on the vendor Sale slots or as guaranteed rep items. Same for blues, which also show up frequently as drops. However, nobody sees oranges on the vendors or as drops. Therefore, most people buy boxes specifically to get oranges.
BTW, -IF- the drop rate for oranges from a T2 is only 2%, then the cumulative probability for 72 Keycodes is 30.5%. In which case T3 would in fact be the best probability even if the true drop rate for T3 is also only 5% (which drops the *** prob for T3 to 36.5%).
but so far the collected data indicates 4, 6, and 8 percent drop rates, making T2 the best bet for oranges.
Wow! great idea. I just bought 2 Tier 3 boxes and with drastically different results.
#1: 2 Green - 1 Blue - 0 Purple - 0 Orange
#2: 0 Green - 3 Blue - 0 Purple - 0 Orange
Haha. I've submitted in the form too. Good job for making this statistics study. Keep it up!
I think this metric should be renamed "Expected Items per 100 keys" and the values multiplied by 100. That should solve any confusion around it. It's not a percentage (they don't add to 100% anyway), it's just a benchmark metric.
Also, can you add a column with number of observations? Just sum up total items for that tier (G + B + P + O) divided by number of items per box. That might help put the statistical significance into perspective.
My numbers aren't supposed to add up to 100%. Cumulative probability doesn't work that way.
May I suggest that you look at the link in my sig block for a more comprehensive explanation of what my numbers mean and what types of questions the notion of cumulative probability answers?
Wasn't talking about your numbers, rather the presentation in the spreadsheet. I've edited the quoted text to be more accurate.
I'm pretty sure he was referring to the "Drop rate / cost per item" column in the table. He seems to be addressing the OP. (Edit: beaten on that one!)
But where does the 72 keycode value come from? This analysis would be ideal if we had 72 keycodes to spend and that was it, ever, but that's not how keycodes work. The player can always accumulate more - left-over keycodes get spent later! Ignoring them can lead to your analysis giving different answers on which tier to buy depending the number of keycodes to spend (see my previous example). Should we tell a player with 8 keycodes to spend that they should buy a T2 lockbox because they have a 0% probability of an orange from T3 and T4 lockboxes (because they can't buy any)? T2 may be the correct choice, but we want to be able to say that buying a T2 is better than saving for a T3 or T4.
To get a consistent answer on which lockboxes to buy to maximize the long-term probability of at least one orange, we thus need to use equal-keycode-value ratios of lockboxes to avoid leftover. The simplest such ratio is 3:8:24. This can be simplified slightly, because there is no need to use an integral number of lockboxes, we just need an integral number of items. Hence, to compare the efficiency of the lockboxes at producing at least on orange, compare 3 T4 items, 8 T3 items, and 24 T4 items.
With the current values in the table, you'll actually expect to get 6.7 Blues/Purples per 100 keycodes spent on T2 boxes, 6.8 per 100 keycodes spent on T3s, and only 4 per 100 keycodes spent on T4s. T2 is better than T4 even for blue/purple accumulation (though slightly worse than T3), you just get a metric ton of greens to go with them!
Another statistic that might be useful to add to the table is "Expected keycodes spent to first orange drop". This would give similar information to the probability of at least one orange (though not equivalent), but be easy to compute and guaranteed to be consistent with the "Drop rate / cost per item" statistic. Computation is just (cost per item) / (orange drop rate as decimal).
For example, the current T4 orange drop rate is 7.68%, so on average we expect to spend 16 / 0.0768 = 203.3 keycodes on T4 boxes before our first orange. The T3 drop rate is currently 4.43%, so we would expect to spend 180.6 keycodes before our first orange. With the T2 drop rate at 2.72%, we expect to spend 142.1 keycodes before our first orange.
he probably chose 72 because 75 is the maximum you can hold at anytime but 75 does not divide into the number of keys it takes to open a box
This is true, but the point still stands - leaving leftover keycodes in the analysis will lead to a biased analysis (so long as those leftovers can be used later). Suppose T4 really did have the best chance at producing oranges efficiently. With 72 keycodes you could get a better probability of at least one orange by buying one T4 and one T2, spending all 72 keycodes. But this is a bad long-term recommendation! Instead you'd want to buy one T4 and save the leftovers for the next T4. This is the same problem with comparing probabilities obtained while leaving leftovers - you're ignoring the fact that those leftovers can be used later! It might give the same answer as a valid analysis, but that doesn't make the analysis valid (except in the case where you only have 72 keycodes with no possibility of obtaining more).
Posting Permissions
Reply With Quote