Info: Entropialoot Data Test #1 - High Eff Amp or No Amp?

[atomicstorm]

This month, I used my Entropialoot Tracker (not available for distro) and performed some data tests to prove or disprove the commonly held theory amongst high efficiency weapon enthusiasts to not amp to improve TT returns. While I cannot prove or disprove TT return theory, what I can look at is whether or not amp decay is returned and what the impact is, if any, on each kill for amping.

In the course of gathering information, I also attempted to detect efficiency impact within the metrics I gathered. I will include this as a byline with summary below but I will not include the data sets - more on that a bit later.

I will be giving my criteria and my results in full nerd mode. I will summarize to the best of my ability but if you want to make sense of it, you're going to have to read and put in the effort to understand it yourself. I will leave it to the community to determine if my conclusion is conclusive as I think it is.

So with that..... NERD MODE ACTIVATE!

**Criteria:**
3 hour test
LP-70 No Amp, LP-70 Imp-105, LP-70 Mayhem Alpha (L) Test
3 enhancers, not including any markup in any of the costs to eliminate variationssaaaaaaaaaaaaaaaaaaa
Do not calculate armor decay
Single maturity mob
DPP calculated with my buffs/rings

**Question**: Does amping with a high efficiency weapon create a significant impact on TT returns?
**Application of Question**: Avatar A does not amp because often only ammo is returned at the end of the data with most of the data being decay lost showing that amp decay is not returned in loot.

**Efficiency differential**: 1.3

**Data Results**

The first attempt at this dataset, results not included here, were more rolls. After about 300-400 "rolls", the data was apparent and most metrics did not change beyond several decimal places to the right; hence the reason for a limited number of rolls in each trial. For the test we are trying to conclude, we do not need a significant amount of rolls. If we were trying to prove TT returns, far more rolls would be necessary.

TotalRolls

DePS

APS

DecayRatio

TotalCost

Eff

ExDPP

CalcDPP

Average Loot

Low Range

Max Range

AvgLoss/Mob

AvgLootExcMulti

Ratio

AvgCost

AvgMulti

AvgExcludeMulti

Ratio

Cost/Eff

Cost/DPP

Damage

Shots

Hit Rate

Skill Lines

DPS

DPS Delta

NoAmp-Set3

632​

1.7056​

21.021​

8.114%​

22.7266​

92.2​

3.70365​

3.684​

4.028610866​

0.8674​

3.3395​

-2.306833074​

2.013559053​

0.465116611​

4.329149​

0.923618​

0.464954915​

0.107400999​

0.046954​

1.175122​

1023627​

12227​

89.90%​

2378​

113.982​

I105-Set3

711​

4.179​

22.451​

18.614%​

26.63​

90.9​

3.6688​

3.657​

3.979760864​

0.8578​

3.3853​

-2.305014613​

2.049552993​

0.469580193​

4.36465​

0.912884​

0.471121535​

0.107940289​

0.048016​

1.189667​

1160815​

11920​

90.14%​

2525​

132.312​

16.081%​

Alpha-Set3

697​

1.724​

30.001​

5.746%​

31.725​

87.3​

3.57737​

3.579​

4.201724113​

0.5172​

3.6257​

-2.390106181​

2.0765631​

0.465978458​

4.45635​

0.938548​

0.464409878​

0.104213062​

0.051046​

1.245706​

1133018​

9978​

90.15%​

2166​

153.697​

16.163%​

The below comparisons are directly between NoAmp and I105. The alpha test is to determine the impact of the L amp and to see if there is any impact of efficiency.

**Average loot stat**: Not relevant. Point in time statistic dependent upon multipliers. See multi-distribution at end.

**Low-Max Range**: Excludes multipliers over .8 -> The increase of the max range indicates inclusion of decay. The expectation here should be roughly ~50% of the differential, see multi-distribution later.

Average shots for LP-70 no amp: 19.04882 @ 22.7266pec/shot
Average shots for LP-70 I105: 16.40751 @ 26.63 pec/shot
(diff: -2.64131 shots @ 3.9034pec/shot = 10.310089 pec)

**AvgLoss/Mob**: This is the average loss per mob. **This is statistically negligent.**

Subnote: The increase average loss per mob in Alpha test is in line with the DPP change. See honorable mentions below. I believe the I105 being similar to noamp is due to luck with DPP. In a previous test with 2x more rolls, the average loss was fairly close to 1.7pec per mob - in line with what you see with the Alpha Set.​

**AvgLootExcMulti**: Excludes multipliers over .8 -> Difference here is 0.03599ped. This indicates an increase for decay.

**Ratio**: Ratio of base multi to cost: This shows the impact of decay. Small increase 0.004463. I am not sure this indicates any specific benefit.

**AvgCost**: Cost of killing a mob on average increased by 0.035501ped. This coincides with the drop in DPP.

**AvgMulti**: Point in time, dependent upon multipliers. Not weapon specific. Explanation later.

**AvgExcludeMulti**: Excludes multipliers over .8 -> Avg base multiplier. I think this change is statistically insignificant. In previous tests with other multipliers, there were not huge changes of this stat between 51 eff and 92.2 eff.

**Ratio**: Ratio of base multiplier to cost. Statistically insignificant.

Spoiler: Test Conclusion

: Amp decay returned in loot is confirmed.

The added cost of the decay and at roughly 50%, see multi-distribution in another post, makes this conclusive for me. Could not amping yield better TT returns? This is always possible since most amplifiers lower efficiency, but at the benefit of more DPS. This makes the idea of amping vs. not amping more situational instead of a must-do survival tactic.

Anecdote note: In my own tracking, my returns have not shifted between amping and not amping. The same low periods with not amping would have just yielded (maybe) worse returns when amping because of my increased cycle rate per hour - in this trial an increase of 16.2%. However, lack of amping might have also resulted in me losing out in the much needed multiplier to correct said downswing.

**Honorable Mentions**

(Not relevant for this, but wanted to include because I tried to dig a bit deeper on efficiency. These adjustments can be explained by change in DPP. The data set below was the first trial. I opted to redo the test to confirm my data.

TotalRolls

DePS

APS

DecayRatio

TotalCost

Eff

ExDPP

Average Loot

Low Range

Max Range

AvgLoss/Mob

AvgLootExcMulti

Ratio

AvgCost

AvgMulti

AvgExcludeMulti

Ratio

Cost/Eff

AddedCostPerEff

Cost/DPP

NoAmp-Set2

1078​

1.706​

21.021​

8.116%​

22.727​

92.2​

3.70365​

4.075329277​

0.9319​

3.2938​

-2.307634747​

2.025907816​

0.467744253​

4.33123​

0.933692​

0.467972374​

0.108046075​

0.046976​

1.169449​

i105-set2

1602​

4.179​

22.451​

18.614%​

26.63​

90.9​

3.66913​

4.372462692​

0.8261​

4.2577​

-2.324094945​

2.073430964​

0.47241988​

4.388958​

1.003143​

0.471609261​

0.107453585​

0.048283​

2.140%​

1.196185​

alpha-set2

1378​

1.7246​

30.001​

5.748%​

31.7256​

87.3​

3.57765​

4.178685376​

0.7031​

3.5973​

-2.387653039​

2.107988547​

0.46791833​

4.505035​

0.926902​

0.469702881​

0.104261756​

0.051604​

1.910%​

1.259216​

lp100modset2

430​

4.244​

23.66​

17.937%​

27.904​

83.9​

3.88349​

3.61913931​

2.51136​

6.13888​

-2.203429305​

1.910513897​

0.466092524​

4.099001​

0.88635​

0.463918879​

0.113178512​

0.048856​

-1.566%*​

1.055494​

lp65set2

575​

1.443​

26.55​

5.435%​

27.993​

69.4​

3.25053​

4.071995869​

3.07923​

7.27818​

-2.640554013​

2.283375541​

0.465310657​

4.907207​

0.827547​

0.464323168​

0.094620668​

0.070709​

3.085%​

1.509664​

vat100mentorset2

362​

4.811​

15.47​

31.099%​

20.281​

51.3​

3.02565​

5.702245109​

3.65058​

7.30116​

-2.801634892​

2.494770144​

0.473710325​

5.266447​

1.085213​

0.471387151​

0.089507629​

0.10266​

2.496%​

1.7406​

* The LP-100 modifier test breaks the Efficiency/DPP trend in all the trials. The fact that this stat is negative (negative to a trial not shown here, relative to the eff, which would be LP-70+Alpha) proves that the metrics changing is not because of Efficiency. DEBONKED!

Thank you to:

• Xanatos Rental service for letting me rent the VAT 100 Mentor for 51.3 efficiency test. Who knew it would be difficult to get a bad weapon to test?
• Dr. Zho Secret Asian Man 007-19 BBQ for initial math conversations, our "verbal concussion" matches, and sound boards.
• Bonnie - for the compromising pictures of John Black Knight and the space octopi. I will keep it "just in case".

What do you think?

---------------------------------------

For questions about some of the multi-distribution information mentioned above and the exclusion of .8 (as pointed out by Zho Secret Asian Man 007-19 BBQ), see my post below on appendix information.

Questions of Note

Q1: What about the small sample size? Doesn't this create issues with the conclusion here?

A1: It would change metrics in the data set that would not be relevant for this exercise - such as average loot. We are not making any conclusions from this information. Removing the multipliers from the equation eliminates this variation. Decay is either returned or it isn't. If it wasn't, then more rolls would be needed to determine if decay is later added in 1 big swirl much later in the equation. Instead you simply risk the extra cost against the 30-50% multiplier. It would only be relevant if we were trying to measure TT returns - and we are not. These tests were also done on different days - and some of these tests ranged from 99% return to 71% return. As you can see though, the multipliers did not change.

 

[atomicstorm]

Appendix, After-Thoughts -> Multi-Distributions of Trial Data Sets

While looking into efficiency impact of the data set in the previous post, under honorable mention, I came across a few findings:

• Multipliers are randomized floats, not whole numbers. This may or may not be a new mechanic of Loot 2.0 OR it was simply always this way and we always looked to 1000x multipliers as an easy conversation marker.
• Multipliers are not weapon dependent. The same distribution of multipliers can be found on 92.2 eff vs. 51 eff.
  • Hypothesis here: This may be influenced by looter levels. This test would need to be redone in 5 looter level intervals. Perhaps next mayhem.
• There is a multiplier black-out between 70% and 80% of cost. No multipliers occur during this window for any weapon - EVER
• Approximately 76-78% of all multipliers occur less than 60% of cost - for every weapon in the trial.
  • 18-19% of those are under 40%. - Also every weapon in the trial.

Multiplier Distribution of Trial Data Set #1 as a Percentage

Multiplier Distribution of Trial Data Set #2 as a Region

No conclusion here other than what I have mentioned above. These numbers will tighten with more "rolls". However, they will not change significantly from what we see here that will dissuade from the opinion one might reach from these graphs.

 

[jak]

Awesome! Thanks for the info! I’ve always appreciated your maths, you are Mozart with maths!

I can say though, even 1k rolls I don’t think is enough to get a good picture in my opinion. This game has some long rollercoaster rides, some down swings and upswings differing by weeks or months. Not too sure how some stats would change in the above if done at different times, might that change some stats?

 

[atomicstorm]

Awesome! Thanks for the info! I’ve always appreciated your maths, you are Mozart with maths!

I can say though, even 1k rolls I don’t think is enough to get a good picture in my opinion. This game has some long rollercoaster rides, some down swings and upswings differing by weeks or months. Not too sure how some stats would change in the above if done at different times, might that change some stats?

It would change metrics in the data set that would not be relevant for this exercise - such as average loot. We are not making any conclusions from this information. Removing the multipliers from the equation eliminates this variation. Decay is either returned or it isn't. If it wasn't, then more rolls would be needed to determine if decay is later added in 1 big swirl much later in the equation. Instead you simply risk the extra cost against the 30-50% multiplier. It would only be relevant if we were trying to measure TT returns - and we are not. These tests were also done on different days - and some of these tests ranged from 99% return to 71% return. As you can see though, the multipliers did not change. I will add this note to the first post.

Tl;dr: a 3 hour test is sufficient as we are not trying to track luck - multipliers, swirls e.g. TT return. We just want to know if decay is returned in this test.

 

[jak]

Ah, got cha!
Thanks!!!

 

[eyeshark]

I didn’t even know some people thought amp decay wasn’t returned in loot regularly... Excellent exercise.

 

[buckaroobanzai]

+rep. Removing high multi’s from the analysis is key to understand the loot mechanics. If you search back in this forum (2006-2008) you will find plenty of info by Falkao and me applicable to loot 1.0. I still log my runs in the same way and nothing in this respect really changed with the shift to loot 2.0.
Take care, BBB

 

[jak]

There is a multiplier black-out between 70% and 80% of cost. No multipliers occur during this window for any weapon - EVER

M8, I thought I understood this well but then some friends questioned me on it and I felt lacking.
Can you elaborate on this a little with examples?
Also, might this be a bug? Guess only MindArk would know but seams like it should be reported as a bug...

 

[atomicstorm]

M8, I thought I understood this well but then some friends questioned me on it and I felt lacking.
Can you elaborate on this a little with examples?
Also, might this be a bug? Guess only MindArk would know but seams like it should be reported as a bug...

It is what it says. There are no multipliers between 70% and 80% of cost. The example is simply the graph. This was true for all 9 data sets.

I doubt it is a bug. It gives it a bit of the variance and may very well be a by product of what ever formula they use for the final result of each loot.

 

[VandalHand]

 

[eyeshark]

I doubt it is a bug. It gives it a bit of the variance and may very well be a by product of what ever formula they use for the final result of each loot.

i.e. TT returns were too high, so they cut a gap from the range to hit their desired target.
/s

 

[Starkiller]

Bumping this up for relevance.

Thanks for the data gathering and publishing.

 

[Leshrac Brevin]

Very cool data set, thank you. And bump that again.

Let me try if I understood the gap correctly, as at 10% and 70% of costs. So like I throw in 3000 PED of costs in a run, then there would be a multiplier gap at 300 ped and 2100 ped of this and the multiplier could return after I reach like over 2400 ped, so at beyond 80% of costs?

I am not sure if this is what you all are talking about and what you meant, please correct me if I am wrong, @ Crimson Eyeshark Tide.

 

[arielos]

• Multipliers are randomized floats, not whole numbers. This may or may not
• There is a multiplier black-out between 70% and 80% of cost. No multipliers occur during this window for any weapon - EVER
• Approximately 76-78% of all multipliers occur less than 60% of cost - for every weapon in the trial.
  • 18-19% of those are under 40%. - Also every weapon in the trial.

wow! The same multipliers for different efficiency weapons at the same Looter lvl, very interesting discovery! Thx for sharing...

But I'm afraid I don't understand what is the "% of cost"... in the above quote. Would you please further explain it?

 

[Stefan 008 Bond]

Nice work! This is the first mention of (one of) the loot gaps in 2.0 on the forums. It has been there since the beginning of 2.0. There are others, smaller less noticeable and possibly looter affected fyi...

M8, I thought I understood this well but then some friends questioned me on it and I felt lacking.
Can you elaborate on this a little with examples?
Also, might this be a bug? Guess only MindArk would know but seams like it should be reported as a bug...

Dont think its a bug, it is part of the rebalance of 2.0 imo.

 

[Stefan 008 Bond]

Very cool data set, thank you. And bump that again.

Let me try if I understood the gap correctly, as at 10% and 70% of costs. So like I throw in 3000 PED of costs in a run, then there would be a multiplier gap at 300 ped and 2100 ped of this and the multiplier could return after I reach like over 2400 ped, so at beyond 80% of costs?

I am not sure if this is what you all are talking about and what you meant, please correct me if I am wrong, @ Crimson Eyeshark Tide.

Based on a single kill.. (then repeat)

 

[Ferial]

Nicely conducted test! Well done.

And...