It is a Neopagan thing, created sometime in the last century.
It has 8 festivals spread around the year roughly (give or take a week or so) evenly spread.
Each of those 8 festivals (or at least 6 or 7 of them) were celebrated at or about that time of year by people variously described as Celts, or Saxons, or Danes or, well, lots of different people associated with England, and the bits of France that were at various times England, and lets not forget Wales nor Ireland. (We can largely ignore the Scots, as the English did for centuries.)
We have various bits of textual and archaeological evidence for most of these 8 festivals, at various times and places over a couple of thousand years, and at various places and times they have been celebrated on dates calculated by various calendars.
The easy ones are the Solar festivals, the Equinoxes and Solstices, or they would be easy if the Earth's orbit were completely stable. They drift, by a day or two each year from their mean point, and over time they drift around the Zodiac too. So whatever method we use to decide this (points vaguely at late September) is the (Southern) Spring Equinox, there will be perfectly good documentary evidence that this choice is wrong, in someone's opinion. The Solstices (Summer and Winter) are the longest and shortest day and night, and vice versa each year. You'll notice there's one in December that is disturbingly close to the major festival of a Roman era cult or three.
Calendars often simply declare these orbital mileposts to be the 21st or 22nd or 23rd of March, June, September and December. Ephemerides and other observation and mathematical predictions give us an hour or ever minute based on some agreed definition. That's the bottom line, agreed definition.
Is the Equinox the moment of passage of the sun through some arbitrary point on the Ecliptic - entry to some particular house of the Zodiac (or achieving the midpoint of it's time in that house). Is it that day which is equal to the night either side of it where you live, or at some other place? Is it a day chosen by a calendar publisher? By the calculations of a religious authority, or even Friday, Saturday or Sunday closest to the nearest full or new moon? We'll avoid the question of moon phases for now, they have similar complexities.
At every turn we find different authorities for placing these days over a spread of a few days and valid reasons in the minds of some to celebrate them perhaps a week or two either side.
Earth's orbit is not circular, we mark off slightly different arcs of the sky each day, each month, and this means the four Solar mileposts are not quite evenly spaced, and due to the Moon and the various planets, their wobble is not synchronous, so the span between them increases and decreases a day or two year by year. Add to that the complexity of Leap Years and you can see where this leads.
If we lived in the Neolithic (New Stone Age, when farming appeared and we became slaves to our crops), this is all pretty simple. We'd have a fair idea, as farmers, that the Solar year was 365 days give or take, and we'd be able to observe the length of day and night, and the progress of sunrise and sunset across our horizon with the seasons. Counting days, coupled with observing which day the sun rose past a certain notch on the cliffs or even a large rock stood on the ridgeline, and we could reset the count easily enough. Local climatic knowledge would tell us things like don't plant until after the Equinox, or plant no more than 78 days after the Winter Solstice depending on where we lived.
We'd celebrate harvests and plantings, but we'd do those things according to lifetimes of accumulated wisdom on local seasons, when it rains, when it snows, when it frosts. We'd do so according to Sun and Moon (because tides influence the air as well as the sea) and sometimes, we'd get it wrong or there'd be unseasonal rain or frost and we'd starve. We'd live by feast and famine, as is the way of the subsistence farmer of the Neolithic.
We'd almost certainly celebrate a special day for the dead, variously (remember I'm talking Europe in the Northern Hemisphere) in early November halfway between the Autumn Equinox and Winter Solstice, or like the Romans, 6 months later (or earlier, remember the year is like a wheel) in early May. We'd celebrate one or two harvests depending on where we lived and what crops we grew. We'd celebrate somewhere in the depths of winter when we used up stores that would not make it through to spring.
Lets look at the Day of the Dead. Samhain it is styled by the celtic revivalists who "borrowed" it, mostly from the Irish. As Halloween, it is locked into the (Roman) Gregorian Calendar as October the 31st, which, if you count the days betwixt Northern Autumnal Equinox and Midwinter, is about a week early. We know from astronomic alignments in stone aged northern European burial sites that there was some significance to a date in early November going back at least a couple of thousand years. Unfortunately we know next to nothing about what anyone in northern Europe did to celebrate or mark such annual observances until only the last 1,000 to 2,000 years, and that often only from what was condemned or copied by Romans and Christians and folk revivalists.
We live in the Century of the Fruit Bat, er, Twenty-first Century. So we have several thousand years of progress in calendrical and astronometric calculations to draw on. We also have several thousand years of assorted and disparate traditions. By the Early to mid Twentieth Century several groups had come out of the woodwork with more or less well thought out and/or historical revivals of "Pre-Christian" European beliefs. These groups fueled an explosion of interest in the Occult, Witchcraft, Paganism, and along the way revived, exposed, composed, created, synthesized and faked a new folkway, which included the now almost universal "Wheel of the Year". Sure, at least one of the festivals was pretty much made up to complete the symmetry, and a lot of stuff picked and polished from 10,000 years and a continent and a half of "Celtic" archeology and a couple of centuries of theorizing and romanticizing cobbled together to make the others. The urban roots of a lot of this revival meant a lot of mistakes were made that could have been corrected by looking out the window, Latitude and local climate can make a huge difference to the perceived "season" at any given day even in places only a few hundred kilometres apart. Adding "late September" events from southern France and from Highland Scotland lead to hybridization of Early Spring and Early Summer imagery. Traditions drawing on one or the other saw the other as wrong. Books more assembled than authored muddied the waters. New traditions made up to fill markets with poor scholarship and even poorer editorship (Ancient Celtic Potato Goddess chips anyone) left ever more genericized and blended authorities for the ill informed to bring to the debate.
