Despite the myriad ways you can now collect your favorite Pokémon, catching wild Pokémon still remains a core component of Pokémon GO. Today, we’re going to take a close look at a less-studied feature: how day and night cycles affect the rarities of wild Pokémon.
Whether you’re a shiny hunter, a PvP guru, or a Pokédex completionist, understanding the when and where of Pokémon spawn patterns can aid you in your hunt for wild Pokémon. Silph Researchers have amassed a data set of over 290,000 spawns since mid-2018 that we used to examine day and night cycles in detail. We found that a surprisingly large number of Pokémon are affected by these cycles. Let’s take a look!
- We found 79 Pokemon species that were either more common at night (nocturnal) or more common during the day (diurnal).
- Some nocturnal and diurnal species have been either removed, added, nerfed, or boosted since the release of the Seasons feature.
- Pokémon species within the same evolutionary line are often (but not always) boosted during the same time of day.
Summary of Data Collection and Analysis Methods
For a more detailed methods discussion, see the appendix at the bottom of this article.
Researchers recorded the species, date, time, weather, and location of each Pokémon they observed. We used our previously determined formula to calculate the in-game sunrise and sunset times for the location and date of each observed Pokémon. Because nocturnal and diurnal Pokémon appeared to change their rarity around 2 hours after sunrise or sunset, we removed spawns that occurred during the transition period and classified the remaining data points as day or night.
In general, event-highlighted species did not appear to be affected by day and night cycles, so Pokémon observed during events were removed.
For each species, we then counted the number of times that it was seen during the day and night. The data set was then shuffled within each unique combination of location, weather, and period of continuous spawn tables to determine how the Pokémon would have been distributed if day and night cycles had no influence. Controlling for these parameters accounted for any differences that may have emerged from biome variability or the tendency for some researchers to record more data during the day (or night) than others. We re-sampled the data without replacement 10,000 times and compared the observed day/night distribution of each species to the 10,000 randomized data sets. A true nocturnal or diurnal species was expected to show a more skewed distribution in the observed data than in the randomized data. We used the q-value to estimate a cutoff that would limit the number of false positives to 1% or 5% of detected species.
Nocturnal and Diurnal Species
The following heat map shows all species that were found to be significantly more common during the day or night before the release of the Seasons feature or in any of the four completed Seasons.¹ The color of each square indicates how much more common, on average, each species is during the day or night.² The darker the blue, the more common the species was at night, and the darker the orange, the more common the species was during the day. Note that the actual increase in spawn rate for each species in a particular weather condition and biome may vary from the overall differences we observed across this highly diverse data set.
We identified 30 nocturnal species and 37 diurnal species before the release of the Seasons feature. Interestingly, species could be boosted by different amounts. For example, some nocturnal species like Lunatone and Shuppet were almost exclusively found at night, while Murkrow was only moderately more common at night. Similarly, some diurnal species like Swablu and Solrock were almost exclusively found during daytime, while species like Exeggcute and Machop were only moderately more common.
The analysis validates many of the observations made by experienced travelers on the Road, such as Ghost and Dark-type Pokemon being found more commonly at night. Overall, Pokémon with a primary Ghost typing were almost two-fold more common at night, and Pokémon with a primary Dark typing were more than three-fold more common at night in this data set (χ2 test of independence, χ2 = 2079 and 2047, df=1, both p-values<10-3). In addition to the tendency for Ghost and Dark-type Pokémon to appear more commonly at night, many Grass-Type Pokémon appear more commonly during the day (25% increase overall, χ2 = 346, df=1, p-value<10-3). Do you see any other species that reference Pokémon lore?
The results also reveal a large number of previously unknown nocturnal and diurnal Pokémon species. These include PvE-relevant Pokémon like Ralts, Machop, and Roselia, and several PvP-relevant species like Slakoth, Meditite, and Gastly.
Changes to nocturnal and diurnal patterns across Seasons
While we didn’t observe any major changes to the list of nocturnal and diurnal species before the start of Seasons, each of the first four Seasons has offered its own updates to the effect of day and night cycles on wild Pokémon.
The inaugural Season of Celebration saw the removal of many Pokémon species from the wild, including almost three dozen nocturnal and diurnal species. Several new nocturnal and diurnal species also emerged during this Season. The diurnal Pokémon Fletchling was released in the wild, and Clefairy and Misdreavus became significantly more common at night. All three of these species retained these characteristics for the following Seasons when they were spawning.
The Season of Legends saw the most dramatic changes to nocturnal and diurnal species. Several species gained new nocturnal or diurnal tendencies, while others were removed or modified. Most strikingly, Illumise and Volbeat changed from strongly diurnal to strongly nocturnal. Given their resemblance to fireflies, this seems a more appropriate choice. The Season of Legends also saw Venonat and Wooper added to the pool of nocturnal species and Growlithe added to the diurnal species roster. Interestingly, Venonat was a bit of a one-season-wonder — when it re-appeared during the Season of Mischief, it’s nocturnal tendencies had been severely diminished. Other species likely lost their day or night boost, including diurnal species like Bidoof, Drilbur, and Teddiursa. Finally, the classically nocturnal species Gastly became even more common at night.
The Season of Discovery retained many of the changes from the previous two Seasons, but brought its own tweaks and fine-tuning. Purrloin made its return to the wild after a two-Season absence on the nocturnal species list, and Zubat and Misdreavus were even more skewed towards nightime. Additionally, Ledyba may have been removed from the nocturnal species, or at the very least was much less boosted at night.
The most recently completed Season, the Season of Mischief, saw further shifts and removals. Spoink, Cubone, Shuppet, Drifloon, Sableye, and Stunky all lost their strongly nocturnal tendencies. These removals were counter-balanced by the addition of Spinarak to the nocturnal species list and Spearow and Litleo to the diurnal list.
It’s unclear why specific species are added or removed from the nocturnal or diurnal species list. Both hemisphere-exclusive species and those found globally can be nocturnal or diurnal. Species associated with a particular biome can also be diurnal, as evidenced by Hoppip’s daytime enrichment during the Season of Legends. Uncovering the pattern might help travelers to predict nocturnal and diurnal species at the outset of each Season.
Correlation with Evolutionary Lines
Observant travelers will note that there are several nocturnal and diurnal species in the same evolutionary line. Can this observation be extended to all species?
Below we show the percent difference between the day spawn rate and night spawn rate for each base species (e.g. Hoppip) against its first-stage evolution (e.g. Skiploom) before the release of the Seasons feature.³ Diurnal Pokémon species have larger positive values (right for base species or top for first-stage evolutions), nocturnal Pokémon species are more negative (left for base species or bottom for first-stage evolutions), and Pokémon that do not change rarity tend to be around 0 (center). Each species is represented by a sprite of its base form (Hoppip for Hoppip/Skiploom).
We found a high correlation between the percent differences in spawn rate for species in the same evolutionary line (Pearson’s correlation coefficient = 0.80, p-value<10-3). This means that an evolution tends to spawn during the same time of day as its base form. Second-stage evolutions were too rare to perform this same analysis; however, Crobat and Jumpluff followed the trends of Zubat/Golbat and Hoppip/Skiploom, respectively, suggesting this pattern applies to the entire evolutionary line.
While this trend was highly significant, we note that the relationship was not perfect. For example, Ariados was nocturnal, while Spinarak was not affected by day and night cycles.
We hope that you’ve enjoyed exploring day and night cycles with us, travelers. If a species you’re hunting is nocturnal or diurnal, consider exploring during those hours to maximize your chance of finding them. Regardless, next time you’re out for your morning or evening walk, be sure to take note of the differences in the Pokémon you see based on the time of day!
Until next time, travelers.
Article author: Scientist Titleist
Analysis: Scientist Titleist
Project Leaders: Scientists Cham1nade, Gustavobc, and PhoenixCrystal7
Graphics: Scientists WoodWoseWulf and Titleist
Editing: Scientist Cham1nade and Senior Researcher JinianD
Data Collection: Over 600 amazing researchers collected data for this article. These 26 put in the extra miles:
- Brush are Best
- Cheery Dog
¹ Species were included in the graphic if they were significantly enriched in either the day or night at a 1% false discovery rate cutoff in any of the five time periods examined. Solosis was manually added despite only reaching a 5% false discovery rate cutoff as Niantic previously announced that they would spawn more commonly during the daytime.
² Enrichment is calculated as the percent increase in spawn rate for each species relative to its non-boosted time of day. For nocturnal species, we took the difference between the average day spawn rate and night spawn rate and divided by the day spawn rate. This ensures all enrichment values for nocturnal species are negative for plotting purposes. For diurnal species, we took the difference between the average day spawn rate and night spawn rate and divided by the night spawn rate. If the absolute value of the enrichment was greater than 16, the value was set to 16 (or -16 for nocturnal species) for the heat map. Species that were seen less than 5 times in a particular time period are plotted in white regardless of the calculated enrichment.
³ Here, the percent difference is defined as the day spawn rate minus the night spawn rate divided by the arithmetic average of the two.
Appendix: Detailed Methods
Researchers collected data from identified “spawn points”. A spawn point is a single place on the overworld map where a Pokémon will appear once per hour. Researchers then recorded each Pokémon that appeared at their tracked spawn points, along with the weather, date, and time. If the researcher did not observe the weather when the Pokémon first appeared, the weather was recorded as “Not Observed”. A small number of data points were collected using a modified data collection procedure where a researcher walked a route many times and recorded the Pokémon seen. In this case, data was only recorded when the weather had not changed from the previous hour to account for any leftover spawns.
Researchers recorded over a million data points before any filters were applied. After removing data collected during in-game events affecting spawns, they recorded 266,205 spawns before the release of the Seasons feature, 5,634 during the Season of Celebration, 7,846 during the Season of Legends, 6,865 during the Season of Discovery, and 5,887 during the Season of Mischief.
Some spawn points generate their spawns 15 or 30 minutes before the Pokémon becomes visible on the overworld map. If the spawn is generated in the prior hour, it will use the weather condition at that time to determine which types to boost. We determined which spawn points were informed by the weather in the previous hour by counting the number of times the Pokémon’s type matched either the previous or current weather boost. Spawn points whose spawns appeared in the latter half of the hour were used as the null model. Remaining spawn points were classified as using the previous hours’ weather if they more skewed towards the previous hour’s weather than any spawn point in the null model. Weather data from these spawn points were re-assigned to the previous hour’s weather when it was recorded.
We used our previously deduced formula to determine the in-game sunrise and sunset times for the location and date of each observed Pokémon. Previous analyses on the Road suggested that nocturnal and diurnal Pokémon change their rarity around 2 hours after sunrise or sunset, which we confirmed in our data set by filtering for a list of highly skewed nocturnal and diurnal Pokémon at spawn points with a large amount of data. Spawns that appeared between 1.5 and 3.5 hours after sunrise or sunset were classified as “ambiguous”. If the spawn appeared between 3.5 hours after sunrise and 1.5 hours after sunset, it was classified as a “day” spawn. Similarly, spawns that appeared between 3.5 hours after sunset and 1.5 hours after sunrise were classified as “night” spawns. This buffer also accounts for any noise that was introduced from inaccurate spawn point coordinates.
All spawns that appeared during an event with a significant change in spawns were removed. In general, event-boosted species are not affected by day and night cycles. One exception to this rule is the Psychic Spectacular of 2020, where Solosis and Gothita appeared more commonly during the day and night, respectively. If an event boosted one or two species (e.g. Summer Hat Pikachu), the boosted species was removed, but the rest of the data from the event was left intact. Spawn points within nests were omitted from this analysis.
We compiled a list of dates where global baseline spawn tables changed, such as when a new species was added to the game or a new Season started. These dates were used to break the data set into blocks with consistent spawn tables, so the data could be randomized within these blocks for comparison with the recorded data. For each species, we counted the number of times that it was seen during the day and during the night. The data set was then re-sampled within each spawn point, weather, and period of continuous spawn tables to determine how the Pokémon would have been distributed if day and night cycles had no influence.
Controlling for these parameters accounts for any differences that might have emerged from biome variability or the tendency for some researchers to record more data during the day or night than others. Not accounting for these parameters could have led to the detection of more potentially nocturnal or diurnal species, but would likely have increased the false positive rate. For this analysis, we chose the more conservative strategy. The data re-sampling was performed 10,000 times, and we compared the observed distribution of each species during the night and day datasets to its distribution in the 10,000 randomized datasets.
To control the false positive rate, we used the q-value. Species were split into those that appeared more commonly during the day and those that appeared more commonly at night compared to the median shuffle (no matter the effect size). For each species, we counted the number of times that the re-shuffled data was more skewed towards day or night than the observed data. This value was then divided by the total number of shuffles in the data set and multiplied by 2 (two-tailed test) to get a “p-value”. The qvalue package in R was then used to estimate a p-value cutoff for a false discovery rate of 1% or 5%. Species were removed if they had been observed fewer than 10 times in the data set. The two species removed by this filter were Quagsire and Sigilyph, who were both nocturnal candidates.
After this filtering step, the ten nocturnal and diurnal species with the highest p-values were examined in closer detail. First, the data for each species was re-shuffled after removing the five spawn points where it was seen most frequently to ensure that its enrichment was not driven by a small number of biomes. Next, data points from one weather condition were removed from the data set and the shuffles repeated for each of the ten species. This process was repeated for all seven weather conditions to confirm that the observed differences were generalizable across weather conditions.