Shouldn’t the trend line pass through the 0,0 point?
Hey, good point, that would be nice, wouldn't it? Turns out I can force that condition, which produces a slightly different slope equivalent to ... uh ... 34.5% ... no, 33.3%! But the trendline doesn't fit as well, see R-squared value of 0.993.
Had to edit because my spreadsheet isn't the model of perfection I wish it was. Sorry again. Edit 2: Third time's the charm! Oof. I think I got it now. Sorry all around.
Shouldn’t the trend line pass through the 0,0 point?
If you want it to be less accurate...
It’s common for trend lines to have limited validity at the extremities of a graph.
I would imagine that if multiple data sets were included, the accuracy of the trend line would likely improve. That is, if several people tracked the same farm, and the cumulative data was used to generate a single trend line, that trend line would be more precise.
Also, I understand that it’s common for discrete trend lines generated from unconstrained data sets to be somewhat inaccurate. However, if there is a hard control point in the data (in this case, you cannot start farming with negative shards), shouldn’t that constraint should also constrain the trend line?
In any case, I’m an engineer, not a statistician. And it’s been a long time since I’ve done any deep dives into statistical analysis. So, I’ll defer to those that do this kind of thing more that I.
You should definitely separate out the signal data since that is not gear. It’s known to have WAY higher drop rates, and that will drag your average up.
You also need to separate out different types/colors of gear. Purples, blues, greens and golds all have wildly different drop rates. Some are higher than 20%. Some are lower. Supposedly.
Those two things + lots more drops will give you a more accurate picture.
Here's my gear data through today w/o the Blue/Green or the Signal in there.
Trending high still even with just Purple/Gold gears.
It's over. 4 months after starting this thread, my Shaak Ti farm is finally finished. Yay. It's a little fun doing these data captures, but I'm ready to be done with it and just play the game a while now.
TLDR:
Shaak Ti shards dropped at 33.5%.
B1 shards dropped at 32.8%.
Xanadu Blood shards dropped at 34.9%.
F5A Shaak Ti: 252/751 = 33.6% (7*)
Collected the 330th shard for Shaak Ti today. Started this farm with 56 shards already in the bank. Conclusion: no evidence of drop rate shenanigans that might spur crystal spending were found.
Results
755 attempts, 253 drops, 274+1 shards
33.5% drop rate = 253/755
5-pt moving average
Range: 12.0% - 56.0%
Special Event Range: 28.0% - 56.0%
Gear drops
Blue gear: Mk1 Bacta Gel (22.1%), Mk2 Holo (20.1%)
Purple gear: Mk10 Holo (18.8%)
Ship materials
Mk1: 2.9%
Mk2: 3.7%
Mk3: 4.8%
Reinforcement: 14.8%
Omega: 4.6%
Sim Tickets: 8.7%
Awesome shard drop figures
2 GAS events
Post announcement periods (yellow bars): batches 43-46, 124-131 (20/60=33.3% drop rate)
Event periods (orange bars): batches 47-50, 132-134 (13/35=37.1% drop rate)
2x drop event (green bars): batches 92-100 (22/45=48.9% drop rate)
Shard drops that occurred during 2x drops were only counted once per drop event when calculating the drop rate.
Drew a least-square trendline through the drop events, slope reflects the best-fit drop rate of 34.0%. The slope is 1.702 drops per batch of 5 attempts which translates into 0.3404 drops per attempt (1.702/5=0.3404). The R2 determination coefficient of 0.996 indicates this is a very good estimate of the drop rate.
Drop count histogram; I went 0/5 23 times, 5/5 only once.
0/5 events occurred at a 15.2% rate
1/5 events occurred at a 29.8% rate
2/5 events occurred at a 31.1% rate
3/5 events occurred at a 20.5% rate
4/5 events occurred at a 2.6% rate
5/5 events occurred at a 0.7% rate
Let's take a look at drop rates when approaching star thresholds ... rates look pretty normal to me.
Approaching 4* (80 shards):
Shard 72 collected on batch 9, shard 80 collected on batch 13
Batch 10-13: 8/20=40%
Approaching 5* (145 shards):
Shard 140 collected on batch 49, shard 145 collected on batch 53
Batch 50-53: 5/20=25%
Approaching 6* (230 shards):
During 2x drop event
Shard 219 collected on batch 94, shard 235 collected on batch 97
Batch 95-97: 8/15=53.3%
Approaching 7* (330 shards):
Shard 322 collected on batch 144, shard 330 collected on batch 151
Batch 145-151: 9/35=25.7%
Overall "approach rate" was 37.5% ... quite good really.
Threshold drop rates are within observed 5-pt moving average range.
0/5 events occurred 2/18 batches (2/18 = 11.1%), less often than the overall 15.2% rate for such events.
No evidence of drop rate roll-off at the star thresholds. In fact, drop rates were better than the nominal rate.
And let's not forget the bonus material! As if you're not bored enough already. See following posts for F5B (B1 Battle Droid) and L5B (Xanadu Blood).
Ok so I'm a little tired of writing about this stuff, so I'll let y'all digest this however you will.
F5B B1 Battle Droid: 172/525=32.8%
Pretty much the same conclusions as the previous discussion for F5A. Rates were a little depressed leading up to the second GAS event but nothing unusual.
Replies
Hey, good point, that would be nice, wouldn't it? Turns out I can force that condition, which produces a slightly different slope equivalent to ... uh ... 34.5% ... no, 33.3%! But the trendline doesn't fit as well, see R-squared value of 0.993.
Had to edit because my spreadsheet isn't the model of perfection I wish it was. Sorry again.
Edit 2: Third time's the charm! Oof. I think I got it now. Sorry all around.
I would imagine that if multiple data sets were included, the accuracy of the trend line would likely improve. That is, if several people tracked the same farm, and the cumulative data was used to generate a single trend line, that trend line would be more precise.
Also, I understand that it’s common for discrete trend lines generated from unconstrained data sets to be somewhat inaccurate. However, if there is a hard control point in the data (in this case, you cannot start farming with negative shards), shouldn’t that constraint should also constrain the trend line?
In any case, I’m an engineer, not a statistician. And it’s been a long time since I’ve done any deep dives into statistical analysis. So, I’ll defer to those that do this kind of thing more that I.
Here's my gear data through today w/o the Blue/Green or the Signal in there.
Trending high still even with just Purple/Gold gears.
TLDR:
F5A Shaak Ti: 252/751 = 33.6% (7*)
Collected the 330th shard for Shaak Ti today. Started this farm with 56 shards already in the bank. Conclusion: no evidence of drop rate shenanigans that might spur crystal spending were found.
Results
Awesome shard drop figures
- 2 GAS events
- Post announcement periods (yellow bars): batches 43-46, 124-131 (20/60=33.3% drop rate)
- Event periods (orange bars): batches 47-50, 132-134 (13/35=37.1% drop rate)
- 2x drop event (green bars): batches 92-100 (22/45=48.9% drop rate)
Shard drops that occurred during 2x drops were only counted once per drop event when calculating the drop rate.Drew a least-square trendline through the drop events, slope reflects the best-fit drop rate of 34.0%. The slope is 1.702 drops per batch of 5 attempts which translates into 0.3404 drops per attempt (1.702/5=0.3404). The R2 determination coefficient of 0.996 indicates this is a very good estimate of the drop rate.
Drop count histogram; I went 0/5 23 times, 5/5 only once.
Let's take a look at drop rates when approaching star thresholds ... rates look pretty normal to me.
- Approaching 4* (80 shards):
- Shard 72 collected on batch 9, shard 80 collected on batch 13
- Batch 10-13: 8/20=40%
- Approaching 5* (145 shards):
- Shard 140 collected on batch 49, shard 145 collected on batch 53
- Batch 50-53: 5/20=25%
- Approaching 6* (230 shards):
- During 2x drop event
- Shard 219 collected on batch 94, shard 235 collected on batch 97
- Batch 95-97: 8/15=53.3%
- Approaching 7* (330 shards):
- Shard 322 collected on batch 144, shard 330 collected on batch 151
- Batch 145-151: 9/35=25.7%
- Overall "approach rate" was 37.5% ... quite good really.
- Threshold drop rates are within observed 5-pt moving average range.
- 0/5 events occurred 2/18 batches (2/18 = 11.1%), less often than the overall 15.2% rate for such events.
No evidence of drop rate roll-off at the star thresholds. In fact, drop rates were better than the nominal rate.And let's not forget the bonus material! As if you're not bored enough already. See following posts for F5B (B1 Battle Droid) and L5B (Xanadu Blood).
F5B B1 Battle Droid: 172/525=32.8%
Pretty much the same conclusions as the previous discussion for F5A. Rates were a little depressed leading up to the second GAS event but nothing unusual.
Edit: populated the reserved post.
L8B Xanadu Blood: 178/510=34.9%
Gear drops were significantly better than fleet nodes, with blue gear dropping at ~30% and purple gear dropping at ~24%.
Edit: populated the reserved post.