Anecdotic evidence is a problem when researching on how the western Eurasian wild horses looked like. Since many authors simply copy off from other authors instead of looking for the primary references, therefore the information of the original descriptions sometimes get quite altered and might also be influenced by personal preferences of the respective authors. I therefore spent some time searching for references as close to the base as possible and also found primary sources, and hereby I want to give an overview over what I found. I present them here chronologically. A “P” at the end of the paragraph means that it is from a primary source, written by people who actually saw the animals, “S” indicates a secondary source which is written by someone who based himself directly on a primary source, and “T” is a tertiary source which is based on a secondary source. For example, my reference for Kozmian’s report is a passage in a book by the Polish Academy of Sciences, therefore a secondary reference. Note that many of the primary sources are written in Latin, so that I am basing myself inevitably on translations, except in the case of C. H. Smith.
Herodot, 5th century before Christ: lightly coloured (leukos) wild horses in the Ukraine. (S) 
St. Isidore, 600: Iberian wild horses: colour like a donkey, ash-coloured. (P) 
Albertus Magnus, 13th century: greyish-coloured wild horses with a dorsal stripe in Germany (S). 
F. Chr. Dahlmann, 1840: Large numbers of wild horses that were hunted lived in Denmark of the 12th century (T). 
H. Röslin, 1593: Wild horses were still living at Vogesen, Elsass-Lothringen. They were faster and wilder than deer and difficult to catch. Once caught, they got tame after some time (S). 
Balthasar Haquet, 18th century: Wild horses at Zamosc: small, blackish brown, large and thick heads. Mane and tail comparably short, stallions had a beard. Were used in fight shows with predators and showed great bravery (S). 
Eugeniusz Rozdzynski, 1721: Wild horses at Zamosc: tan and isabelline in colour (T).
Rytchkof, 1762: colour dun or bluish, other shades exceptional (S). 
Samuel Gottlieb Gmelin, 1768: Wild horses at Woronesh, Russia. The largest of those wild horses barely reached the size of the smallest Russian domestic horses. The ears were very pinned and the size of domestic horses or sometimes longer and hanging down. The eyes were fierily. The mane was short and frizzy, the tail shorter than in domestic horses. The were typically mouse-coloured, but white or grey horses were also reported. The belly was ashy at the base, the legs were black from the knees downwards. The hair was long and dense during the winter and felt more like fur than horse hair. (P, S). 
Berenger, 1771: Ural. middling size, roundish and short, big heads, ewe-necked and of a bluish grey colour (S). 
Peter Pallas, 1771: Free-roaming horses at eastern Russia and western Siberia. Considered them feral domestic horses. Resembled Russian farm horses, but they had thicker heads, pinned ears, short and frizzy manes and shorter tails. They were of a greyish brown colour and had lighter coloured legs, brown and greys would appear. The colour of the head was white/light on the snout and black towards mouth. Black horses were rare, and there were no piebald ones. They lived in herds of 20 individuals. (S) [1,2]
Kajetan Kozmian, 1783: Wild horses at Zamosc: small, strong limbs, enormous strength and uniform dark mouse colour (S). 
C. H. Smith, 1841: (Probable) Wild horses in the Russian steppe: “coupled with different proportions and position of the ears, an arched or plane forhead, a straight or curved nose, a difference of colour in the eyes, of the skin, of the hoofs, the constancy of their liveries, of their marks, in a streak along the back and bars on the limbs, of dappled croups and shoulders, or of dark uniform colours, dense or thin manes and tails, although traits now mixed,” […] “All seem to refer to a sturdy form of mountain-forest ponies, still found in the province of Cordova, in the Pyrenees, the Vogesian range, the Camargue, the Ardennes, Great Britain, and in the Scandinavian highlands: all remarkable for an intelligent but malicious character, broad forheads, strong lower jaws, heavy manes, great forelocks, long bushy tails, robust bodies, and strong limbs; with a livery in general pale dun, yellowish brown and a streak along the spine and cross bars on the limbs, or the limbs entirely black, as well as all the long hair and mostly having a tendency to ashy and gray, often dappled on the quarter and shoulders”. […]“These horses are evidently again referred to by Andr. Schneebergius, who states, that “there were wild horses in the preserves of the prince of Prussia, resembling the domestic, but mouse-coloured, with a dark streak on the spine, and the mane and tail dark;” […] “Real Tarpans are not larger than ordinary mules, their colour variably tan, Isabella or mouse, being all shades of the same livery, and only varying in depth by the growth or decrease of a whitish surcoat, longer than the hair, increasing from midsummer and shedding in May: during the cold season it is long, heavy, and soft, lying so close as to feel like a bear’s fur, and then is entirely grizzled; in a summer much falls away, leaving only a certain quantity on the back and loins: the head is small, the forehead greatly arched, the ears far back, either long or short, the eyes small and malignant, the chin and muzzle beset with bristles, the neck rather thin, crested with a thick rugged mane, which, like the tail, is black, also the pasterns, which are long: the hoofs are narrow, high and rather pointed; the tail, descending only to the hocks, is furnished with coarse and rather curly or wavy hairs close up to the crupper; the croup as high as the withers: the voice of the Tarpan is loud, and shriller than that of a domestic horse; and their action, standing, and general appearance, resembles somewhat that of vicious mules.” (P) 
C. R. Darwin, 1868: “It seems that not very long ago a wild breed of dun coloured horses with a spinal stripe was preserved in the royal parks in Prussia. I hear from Hungary that the inhabitants of that country look at the duns with a spinal stripe as the original stock, and so it is in Norway”. (P) 
Heptner, 1989: Last living Tarpan (Dubrowka Tarpan), died 1918. It was 140-145 cm tall, had a large head, small ears, short neck, mouse-coloured coat, broad dorsal stripe, faint shoulder stripes, black mane, tail and lower legs, semi-erect mane, broad and arched front head and a straight head profile. (S) 
What is going to follow now is my personal interpretation of these contemporaneous descriptions of the wild Equus ferus ferus. If anyone knows further sources, I’d be grateful if it was pointed out to me, I want this list to be as complete as possible.
It seems that some features are consistent over all the reports, such as the large and thick head, strong and sturdy body, dun coat colour, short manes and tails, and pinned ears. Except the description of the last captive Tarpan, all the texts describe the wild horses as small compared to domestic horses. The exceptional size of the Dubrowka Tarpan can be interpreted as sign of intermixing with domestic horses, yet all the other features are consistent with the animals of other reports. The ears are either described being smaller or larger than in domestic horses, I don’t know which is more likely to be correct. I think that the notion of the ears hanging down refers to the ears being clapped towards the rear as an agonistic gesture. I suppose that “the difference of colour in the eyes” written by Smith is a comparison with the less-pigmented irises of some domestic horses.
There is some controversy in the literature whether western wild horses had an upright, standing mane like the Przewalski and other wild equines or a falling mane like domestic horses. As far as my interpretation goes, there is no clear reason to assume the former is the case. “Short mane” does not necessarily imply a standing mane as there are horses having a short but falling mane, and “short but frizzy” definitely implies a mane like in, f.e., Exmoor ponies to me. Sometimes the manes of these horses are short enough to be semi-erect. The comparison with mules by Smith and the description by Heptner implies such a mane as well. The wild horses seemingly had a shorter tail, and interestingly some of the semi-feral Koniks in the Oostvaardersplassen have comparably short tail hair as well. The notion of a beard for the stallions in the Zamosc population fits the fact that Przewalski’s horses and other horses with a long and dense winter coat have a “beard” too (longer hair on the underside of the lower jaw), but in both sexes. I think it was present in both sexes of the European wild populations too, but perhaps the author didn’t recognize it or was misinformed.
Charles Hamilton Smith reports of free-ranging mountain-forest ponies with heavy manes, forelocks and tails that seemingly had been around in western Europe and Scandinavia back this time. I don’t know what to make out of that, because the widely held opinion is that wild horses died out centuries earlier before they ceased in Eastern Europe during the 19th century, yet these ponies apparently were contemporaneous with the last (and mixed) Tarpan populations. I would suggest they were feral animals, but their wild-type colours seem to suggest otherwise.
The colour is almost consistently reported as dun, which is indicated by the more or less light colour, dark legs and the broad dorsal stripe. The seemingly rare and locally limited presence of other colours like grey or white legs suggest domestic influence in some herds. The colour of the horses Pallas had seen makes me think that he did in fact see feral horses (or hybrids) like he suspected, although the rest of their morphology resembles wild horses. Terms like “donky-coloured”, “ashy” or “greyish” and probably also “blue” definitely indicate grulla/black dun to me, as much as dark mouse-colour. As you might know, there are various shades of grulla, and I think the horses linked below might come pretty close to the colour of the black dun horses witnessed by those authors (unless stated otherwise, all of those photos show Koniks):
|Vyatka horse, looks pretty wild horse-like to me|
“Blackish brown”, used by Haquet for the colour of the horses of Zamosc, could be interpreted as the sooty brown colour of ponies like the Exmoor, but since the mealy mouth and white belly are not mentioned (a pretty apparent detail that probably would not be ignored) and the reports of other authors implicate shades of grulla in this population, I think Haquet saw grulla horses that looked like the darker ones above. Bay duns were seemingly – and not surprisingly – present in the wild populations as well. “yellowish brown”, used by Smith for some of those western forest ponies, definitely indicates bay dun to me. Tan probably also refers to bay dun, but could be grulla as well. Isabella would be an indicator of domestic horse introgression if the authors indeed referred to its modern meaning in horse colour terminology. But it could mean bay dun as well (cite from that post): “The tan or isabelline colour described by Rozdzynski does not necessarily imply what modern terminology identifies as “Isabella”, it could have been bay dun or even black dun as well, historic horse colour terminology was not as strict as it is today (there are some confusions even today, f.e. the British often call dun “cream”, which is a different gene).” The only author possibly describing a white muzzle is Gmelin. The white muzzle is part of the pangare condition, where the ventral side of the body is of a light colour. This feature can be present in both bay dun and black dun horses, but the white hair around the mouth in the latter is shed after some months so that the snout gets dark in adult grulla horses (photo of a black dun Sorraia foal with mealy mouth, photo by Lynne Gerard), therefore Gmelin’s notion implicates that some of the horses he saw were bay dun. Some grulla horses, Koniks for example, do show a lightly coloured belly too, what Gmelin obviously referred to (“ashy at the base”). Because of the lack of notion on the mealy mouth in all other sources I suppose that this feature was not common among bay dun wild horses (and absent in the grulla ones anyway). I think this Dartmoor pony I photographed at the Dartmoor National Park might be a good model for the colour of those European wild horses that where bay dun, also because of its tone:
Although the alleles for both bay and black have both been identified in European wild horses (the dun factor, which acts upon these alleles, was not evaluated yet) there is an interesting discrepancy between historic and genetic data: the bay allele predominant in the predomestic samples of the genetic studies [5,6], and considering that the black allele is recessive, grulla horses must have been quite rare compared to their bay dun conspecifics – yet it seems that the majority of the wild horses described in historic sources are probably of a grulla colour. I think there are two possible explanations for that: 1) the black allele is a mutation that turned up in the early Copper age  and since that constantly spread all over Europe [5,6], implying that it had a selective advantage over the bay allele. Perhaps the selection process was that far in the second millennium after Christ that the grulla horses started to outnumber the bay dun ones. 2) There might have been at least one genetic bottleneck which coincidently increased the frequency of the black allele. I admit that this is speculative.
The “surcoat” Smith describes obviously is the winter coat of the horses. The winter fur is, as far as I know, always lighter than the summer coat, and Smith’s notion that the Tarpani have white hair during winter is in accordance with Przewalski horses, whose colour gets very light in the cold season as well.
 Tadeusz Jezierski, Zbigniew Jaworski: Das Polnische Konik. 2008. Polish Academy of Sciences
 Charles Hamilton Smith, 1841: The Natural History of horses, with Memoir of Gesner
 Hardy Oelke: Wild horses then and now. Kierdorf-Verlag.
 Ludwig et al. 2009: Coat color variation at the beginning of horse domestication
 Pruvost et al. 2011: Genotypes of predomestic horses match phenotypes painted in paleolithic works of cave art