**TL;DR - Greens, Blues, Purples and Legendaries give about 4, 21, 53 and 85 (Lesser) Essence on average when salvaged. We can use this info to accurately compare running Trials with running Morgenheim and other maps for Essence Farming**

This is an attempt at kicking our data collection in a higher gear.

Coda has made a great effort with his spreadsheet to record and compare Crafting Resource Data.

One point where we in our discussions thought of a different approach is the qualifications of a level itself.

I believe data collection will be more accurate if we treat Essence farming as 2 different steps:

- Play a level to collect items
- Salvage said items for Essences

Both of these steps have their own random distribution associated with them, making data collection much more complex.

If we would somehow “score” what the expected Essence output is from each item category, then we only need to focus on step 1 for each map: determining the spread of items on each map.

As an example, if we discover Morgenheim drops (warning: made up numbers!) e.g. on average 5 greens, 7 blues, 0.5 purples and 0.2 legendaries, then we could easily compare this to Trial farming for 3 legendaries per run if we could “score” 3 legendaries to be equal to (again, made up numbers!) e.g. 15 greens and 10 blues in terms of average Essence yield.

At that point the only other variables become: Gold, Gemstones, Mark of Titan drops and the time it takes to run each map. And perhaps the chance of the occasional Spirit dropping.

I’ve set out to do exactly that. Provide a baseline for Essence yield on salvaging lvl70 items.

__Phase 1: Data collection__For this exercise I’ve gathered a lot of items and salvaged them.

I deeply apologise to the mathematicians and statisticians amongst us, because I’ve taken a couple lazy shortcuts which cause the calculations afterwards to be taken with a grain of salt (since almost nothing in my dataset has identical sample sizes except for the Legendaries) but sadly there’s little other choice because we have no access to infinite storage space, infinite spare time, nor consistent and replicable item drops.

The result is data on the salvaging of:

- 500 Greens (salvaged in 25 batches of 18-24 items)
- 575 Blues (salvaged in 25 batches of 21-26 items)
- 290 Purples (salvaged in 25 batches of 9-15 items)
- 140 Legendaries (salvaged in 7 batches of 10, and another 70 items one by one)
- 24 Purple Unique items (salvaged one by one)
- 21 Purple Set items (salvaged one by one)
- 13 Legendary Unique items (salvaged one by one)
- 48 Legendary Set items (salvaged one by one)

Total: 1611 items salvaged and recorded

*If you want access to the data set for your own analysis, PM me ;-).*

__Phase 2: Comparing regulars and Unique/Set items__The first thing I wanted to check is if a regular item has the same Essence salvage value as a Unique or Set item, with just the extra Mark of Titan and special materials bonus.

For this, we assume Unique and Set items can be merged together into one dataset. Comparing the mean values of my data does point in that direction, but no explicit analysis was done due to the small data samples.

A second assumption is that we should independently count Lesser Essence, Essence and Greater Essence values for greater accuracy: Greater Essences count for 100 Lesser but are such infrequent drops that some bad luck in the sample data would immediately skew the outcome to levels I don’t like.

**2.1: Purples:**

Our 25 samples of (in total 290 regular purples) shows the following confidence intervals (0.05 T-test interval with sampled Standard Deviation) on the “per item” yield in each salvage batch.

*** Disclaimer: this is actually a violation of proper statistical procedure. Sampling batches should always have the same batch size. Sadly this game doesn’t just drop equal numbers of purples in your lap every few runs, so I’ve had to work with batches of 9-15 and calculate the average of each batch. Expect the margin of error to be significantly larger than what it should be because of that ***

Mean | 95% Confidence Interval | |

Lesser Essence | 32.902 | ±0.509 |

Essence | 1.511 | ±0.141 |

Greater Essence | 0.044 | ±0.031 |

The 45 Purple Unique and Set items have the following Salvage yields:

Mean | 95% Confidence Interval | |

Lesser Essence | 33.222 | ±1.366 |

Essence | 1.556 | ±0.437 |

Greater Essence | 0.067 | ±0.076 |

Doing a 2-tailed T-test with unequal variances on the two datasets brings us the following statistical probabilities that the means of both populations samples are actually the same:

Lesser Essence | 0.658 |

Essence | 0.845 |

Greater Essence | 0.571 |

Not overly convincing, but let’s be honest, with numbers this close the odds the developers in fact indeed intend these numbers to be equal is pretty damn large.

**2.2: Legendaries:**

Our 14 samples of (in total 140 regular legendaries, batched nicely per 10) shows the following confidence intervals (0.05 T-test interval with sampled Standard Deviation) on the “per item” yield in each salvage batch.

Mean | 95% Confidence Interval | |

Lesser Essence | 49.793 | ±0.792 |

Essence | 2.693 | ±0.287 |

Greater Essence | 0.071 | ±0.048 |

The 61 Legendary Unique and Set items have the following Salvage yields

Mean | 95% Confidence Interval | |

Lesser Essence | 49.754 | ±1.060 |

Essence | 3.033 | ±0.385 |

Greater Essence | 0.066 | ±0.064 |

Doing a 2-tailed T-test with unequal variances on the two datasets brings us the following statistical probabilities that the means of both populations samples are actually the same:

Lesser Essence | 0.952 |

Essence | 0.151 |

Greater Essence | 0.881 |

As you can see, the difference on the Lesser essences is exceptionally low, and the T-test actually reaches 95% confidence the means are the same.

We also clearly see in our data set that the salvage values for every legendary tend to situate around 50, being the expected sum of 4 uniformly distributed Lesser Essence yields of 7-18 per type.

**Conclusion:**I’m quite confident a Unique item or Set item salvages the same Essences as its normal quality counterpart, with just extra Mark of Titan and Special Materials on top as a bonus.

__Phase 3: Scoring the quality of each item__Now that we have shown in Phase 2 that Unique and Set items give the same yield, we can merge the datasets for a more accurate overall mean calculation.

I’ll report 2 means: the mean of our data samples (so the mean of the averages per salvaging batch, which is sadly not 100% accurate but does allow some

**sort of kinda indicative**confidence intervals to be shown, at least for an expected yield in the case of big numbers.

Next to that I’ll also calculate the overall mean of the full data set, simply summing all Essence yields together and dividing by the number of items salvaged.

These two numbers are very close together, but reported both for completeness. In the case of Purple and Legendary data sets, due to the mixed individual and batched data the Batch mean and Confidence Interval will only refer to the 290 and 140 regular items, the overall mean includes Unique and Set items.

To calculate total Essence yield, we’ll take the yield of each subtype, and multiply this by the Lesser Essence Equivalent (so *10 for regular Essence and *100 for Greater Essence).

*** We now make abstraction of the gold cost needed to combine/shatter essences, let’s consider it out of scope ***

Batch mean | 95% Confidence Interval | Overall mean | |

Green | 3.801 | 0.168 | 3.806 |

Blue | 21.251 | 1.580 | 21.217 |

Purple | 52.373 | 3.535 | 52.848 |

Legendary | 83.864 | 5.996 | 84.706 |

__Phase 4: Scoring maps__This Overall Mean calculation now finally gives us a tool to score the efficiency of maps in terms of Essence farming.

- A TL80+ level drops 3 legendaries, guaranteeing us (besides the potential of valuable set/unique crafting materials) on average 254 Lesser Essences worth of yield.
- A run of a story map that takes the same amount of time should drop at least (just a made up split of course) e.g. 1 purple, 8 blues and 9 greens to beat the TL80+, being on average 257 Lesser Essences worth.

__Conclusion__:Since for Trials we know the score is 254 for end-game farming, the only thing that remains for us as a community is to determine the average expected item drops for the story maps, something that is very fast to note down just by tallying greens, blues, purples and legendaries in the inventory every couple of runs.

Then, for example, if Trial 80 takes you 2 minutes per run, as long as you find Story mode maps yielding >127/minute in Lesser Essence Score, they’ll be faster for you for Essence farming.

## Comment