DarwinBotsManual
Open the app EspañolES

Analyze

The screen for understanding what happened in a run: charts from a catalog of metrics, automatic findings, species, family tree, genetics, events and comparisons between runs.

Observe shows you the world cycle by cycle; Analyze shows you the whole movie. While a run advances, the app takes a sample every 100 cycles: how many bots there are of each species, how much energy they have, how long their DNA is, how many times they shot, who is whose child. Analyze draws all of that and helps you find what is interesting.

Open it from Analyze in the top bar. It has seven tabs: Dashboard, Species, Phylogeny, Genetics, Events, Compare and Reports. The first five look at a single run; the last two work with several.

Which run you are looking at

At the top right, the Run selector offers:

To have a saved run, use Save in Observe (see Observe). The app keeps the last 20 runs (see Your data).

Under the title, one line sums up the run: whether it is the current one or a saved one, the cycle it reached, its seed and how many cycles apart the samples were taken. If there are no samples yet, the app tells you at which cycle the first one arrives; if no run is in progress, it offers Go to Observe.

Next to it is the Cycle range: All, Last 10k or Last 1k. It trims the Dashboard, Species and Events charts all at once, so they stay aligned with each other.

The page address remembers the run and the tab you picked: if you reload or copy the link, you come back to the same place. With the focus on the tabs, the left and right arrow keys move from one to the other.

How the history is stored

A long run piles up a huge number of samples. To keep the history under about 5 MB, when it fills up the app merges the old points two at a time: the new point keeps the mean of the samples it combines, and also their minimum and maximum. That is why, in the old stretches of a long run, each point on a chart stands for several samples and carries a light band that goes from the minimum to the maximum. The latest sample is never merged.

Events (extinctions, seedings, live changes…) and the lineage are stored separately and are not merged.

Dashboard

The Dashboard shows four charts. By default they are living bots by species, births by species, mean DNA length by species and total energy. Each one has a Change button that opens the Chart catalog: about a hundred metrics split into six groups.

GroupSome metrics
Populationliving bots, vegetables, animals, corpses, living species, bots in multibots
Evolutionmean and highest generation, mutations, age, offspring per bot
GeneticsDNA length (mean, minimum, maximum), genes per bot
Behaviorshots of each type, reproductions, births, ties created, shell, slime, venom and poison made, deaths, kills
Energytotal energy, of vegetables and of animals, body, waste, chloroplasts
Environmentfield size, obstacles, teleporters, available light, sun position and range, day or night

Metrics that carry “· species” in the catalog are drawn as one curve per species; the others are a total for the whole world. Quantities that add up (bots, energy, shots) come out as stacked areas, one layer per species, and averages (generation, DNA length) as lines. The eight species that reached the most bots get their own curve; in the stacked areas, the rest are lumped into a gray layer, Others.

When you hover over a chart, you see the cycle and the value of each curve (with its minimum and maximum if the point is merged) and, in the areas, the total. Your choice of the four charts is remembered in this browser.

What each thing measures is in the simulation chapter: energy and body in Energy, body and waste, shots in Shots, defenses in Defenses and ties in Ties and multicellular bots.

Findings

To the right of the Dashboard, the Findings card reads the history with a set of detectors and writes one sentence for each notable thing it finds. These are fixed rules, not artificial intelligence: that is why the card says “rule-based text”.

TypeWhat it looks for
DominanceA species that holds more than half of the animals for at least 5000 cycles.
CollapseA sudden drop in the animal population: to half or less in about 2000 cycles, and it stays down.
ExtinctionA species that runs out of bots and doesn't come back. A vegetable that reappears through repopulation doesn't count.
DNA lengthA sustained growth or reduction (of 20% or more) in the DNA length of the animals.
ReplacementA species that displaces another as the most numerous.
OscillationA population that goes up and down with a regular period, at least five times in a row.

Vegetables count in none of them, except extinction. Each finding has a link with its cycle: when you click it, a dashed line marks that cycle on all four charts. If the detectors find nothing, the card says so. With the current run, the findings are recalculated every few seconds. They are the same sentences that open a report (see Reports).

Below there is a table with the most numerous species (now, their peak and their highest generation) and the All species link, which opens the next tab.

Species

A table with one row per species and these columns: Species, Now (living bots in the latest sample), Peak, Appeared, Extinct (“alive” if it still is), Max gen., Mutations, Mean DNA, Energy per bot, Mean age and Offspring per bot, plus a small Living bots curve. Click a header to sort by that column; another click reverses the order.

Clicking a species name opens its card, below the table:

The Show in Phylogeny and Dominant DNA vs founder buttons take you to the following tabs with that species already picked. The picked species is the same in Species, Phylogeny and Genetics. What a species is for the engine is in Species and lineage.

Phylogeny

At the top, the Species tree: one bar per species that goes from the cycle where it appeared to the last one where it had bots, with a × where it went extinct. If a species was born from another, it hangs from it, and the arrow next to the name collapses or expands its derived species.

In a simulation set up in the app the tree is usually flat: all species are founders, seeded by you or arrived through a teleporter. That is because the app doesn't turn on self-speciation, which is what makes new species appear (see Species and lineage). A tree with branches shows up when you load a .dbsim file that has it turned on.

Below, the Individuals of the picked species: a family tree ordered by generation, from left to right. The colored circles are the living bots; the dots, their ancestors. So that it doesn't grow without end, the app keeps only the living bots and their ancestors: branches that were cut off without descendants are discarded. If the chain goes further back than what fits in the drawing, it marks that with “…”.

When you click a bot, the Selected individual card shows its number, its Generation, the cycle it was born in, its Parent, its Mutations, its DNA length, its Offspring and whether it is alive or an ancestor. You can also move around the tree with the keyboard: left goes to the mother, right to the first child, up and down to the neighboring bot. The lineage follows the maternal line, as Species and lineage explains.

Genetics

The tab has two halves.

Histograms. With Histograms of you pick All species or just one, and with the buttons, what to measure: DNA length, Generation, Accumulated mutations, Age, Energy, Body, Genes, Offspring and Kills. The histogram is of the latest sample, with a dashed line at the median. For all species, Change over the run is added: a heat map with time on one axis and the value on the other, darker where there are more bots. There you can see, for example, how DNA length drifts over the run. For a single species there are four measures (DNA length, generation, mutations and offspring), which come from its living individuals.

Dominant DNA vs founder. Every 1000 cycles the app takes a snapshot of the dominant DNA of each species, the one that most bots carry, and saves it if it changed. Pick the Species and, if there are several, the Snapshot: the tab compares it with the first one, which is the founder's. At the top, four numbers: how many genes changed, the length before and now, the distance to the founder and what share of the species carries that DNA. Then, one cell per gene (same, modified, new or deleted) and, for each gene that changed, the words before and after with the differences highlighted. At the end, the full dominant DNA can be expanded.

This distance compares gene by gene; it is not the one the engine uses to decide whether two bots can cross (Species and lineage). How DNA changes when it mutates is in Mutations, and Your first evolution experiment suggests an experiment to see it in this tab.

Events

A chart of the total population with a pin for each event and, below it, the list of the run's events: the same ones Observe announces, stored without trimming. The buttons filter by type and count how many there are:

FilterWhat it gathers
New speciesspecies that appeared and ones that arrived by teleporter
Extinctionsspecies that ran out of bots
Recordspopulation and generation records
Live changesparameters changed while the run was going and changes to world objects
Seedingthe start of the run and every seeding
Save and loadwhen it was saved, resumed or loaded from a file

When you click an event, in the list or on its pin, its cycle is marked with a dashed line on this chart and on the Dashboard and Species ones. Clear the mark removes it. It is useful to see what happened to the metrics right after a live change or an extinction.

Compare

Compare has three views.

Two runs. Pick Run A and Run B (the current one or saved ones), a Group and a Metric: the two curves are drawn overlaid, each with its band. Below, the Settings differences: scenario, seed, settings base, each parameter that differs at the start, the seeded species, the world objects and the live changes and seedings of each one. It is the quick way to answer “why did these two runs end up so different?”.

Replicates. A single run can fool you: chance weighs a lot. Replicates repeat a run's scenario with other seeds to see which result is typical and which was luck.

  1. In Source run, pick the run. It must have come from a scenario; a simulation loaded from a file can't be replicated.
  2. Pick how many Replicates (1 to 64), how many Cycles and how many Workers at once (up to 8 by default, at most the cores of your computer minus one).
  3. Pick the metric you want to see and click Launch replicates.

Each replicate runs without drawing, at full speed, and repeats the source run's live changes at the same cycle. Replicate 1 uses the source run's seed. The result is a line with the mean of the replicates and a band from the 10th to the 90th percentile, plus a table with the final value of the main metrics (living bots, animals, vegetables, living species, highest generation, mutations, DNA length, energy and kills) with their mean, standard deviation, minimum and maximum. While replicates are still missing, you see what has already finished.

Sweep. The same, but changing one parameter: pick one with the Parameter search (by name, variable or key), its Values (From one value to another in equal steps or a List of values, between 2 and 32) and the Seeds per value. The app runs each value with each seed, up to 512 runs in total, and shows the metric at the end by parameter value, a table per value and the mean series of each value. If the source run had live changes of that same parameter, they are rewritten with each run's value. The parameters are in Experiment: advanced mode; by default the sweep proposes Solar share per cycle.

Replicates and sweeps are background jobs: they keep going while you use the rest of the app and resume if you close it. How to manage them, and how to make a report with their results, is in Reports.

Reports

The last tab builds reports of one run, of two runs or of a replicates job, and exports the raw data. It has its own page: Reports.