The eldest game that can (with some reliability) be classified as a chess game is the so-called Baoying-Xiangqi. This is the name given to the game described in the Xuanguai lu (An alternative title for the book is Youguai lu, the meaning being the same as Xuanguai lu.) (often rendered as Book of Marvels, but possibly more accurately Notes on the Strange and Memorable, a book written by a Minister of State of the Tang dynasty, Niu Sengru (779--847). In the story Cen Shun we are told about an impoverished scholar named Cen Shun who, in a dream, is witness to a battle. Later it turns out that he had been living in a house that had been erected above an old grave that amongst other things contained a fully set up chess-board.(The name Baoying-Xiangqi comes from the date, Baoying 1 (= 762 AD), at which these events are supposed to have taken place.) From the text one might deduce that once there was a type of Xiangqi that actually used three-dimensional pieces. The pieces the story tells about were made of gold. Although the eldest extant version of the text is from the latter half of the 10th century, (the text is included into the collection Taiping guangji, 'Gleanings from the era Taiping', compiled 977--978.) there is another text that corroborates the existence of a chess game at the beginning of the 9th century. It is a poem (this poem belongs to the cycle He chun shen ershi shou, 'Twenty poems on the quiet in spring'.) written by Niu's contemporary and friend Bo Juyi (772-846), that mentions Xiangqi and other board games. But alas, it doesn't give particulars on the pieces. So there is a slight possibiliy that there might be three-dimensional Xiangqi pieces dating from the 9th or 10th century somewhere. Needless to say that a properly documented find of a set of such pieces would be quite a sensation. From that time on there is no further mention of the possibility of three-dimensional pieces. (With one exception: in the early 12th century text Pingzhou ketan, 'Leisurely talks from Pingzhou', chess-playing 'foreigners', most likely Persian or Arab merchants, are mentioned. Their pieces were described as 'unlike Horse and Chariot ( the Xiangqi pieces, P.B.); all pieces are made from ivory, rhinoceros horn, and aloeswood'. To my knowledge no such pieces have ever been found.) The oldest Xiangqi game of which we have a complete description of is the so-called Qiguo Xiangqi ('Seven Realms-Xiangqi') which is played on a board with 19 by 19 lines, and uses 120 pieces: one General, one Commander, one Colonel, one Diplomat, one Catapult, one unit of Archers, one unit of Crossbowmen, two units of Swordsmen, four units of Broadswordsmen, and four units of Cavalry each. It was created (or propagated) by Sima Guang (1019--1086). We do not know of a single set of pieces for this game, nor do we know whether it was played at all. Most likely the pieces for this game were shaped like these for ordinary modern Xiangqi, but there would have to be pieces that aren't found in ordinary Xiangqi.
All we know about from the texts are flat, disc-shaped pieces that are inscribed with Chinese characters and/or -- for some period of time at least -- are engraved with pictures corresponding to the Chinese names of the respective pieces.
The earliest set we know of was found in 1984 near Anxi in Jiangxi province. By sheer coincidence a wooden box was found that contained 32 copper pieces: two Generals, four Ministers, four Chariots, four Horses, four Elephants, four Cannons, and ten Soldiers. They were divided into two sides through different colours (red/black); the discs had characters on one side and pictures on the other. The two Ministers of each side were differentiated through two slightly different characters; we don't presently know for what purpose, if any. The discs were about 3.8 cm wide and 7 mm thick. From coins that were found alongside they could be dated to the era Chongning (1102-1106) during the Northern Song Dynasty (960--1126).(The find is described in Zhang Ru'an, Zhongguo xiangqishi ('A History of Chinese Chess'), Beijing 1991, p105. I could not find another description nor pictures of these pieces.)
A similar full set of copper pieces, dated to the same period, was unearthed in Kaifeng. They, too, were inscribed with Chinese characters on one side, and bore pictures on the other side. The pieces were smaller (2.0--3.0 cm), and the ministers were not differentiated.(Pictures of these pieces in Li Songfu, Xiangqi shihua ('Talks on the History of Chinese Chess'), Beijing 1981, p60-61.
A third similar set (copper pieces) of unknown provenance, but tentatively dated to about the same period of time, a full set as well, measures 2.6--3.1 cm, the thickness of the pieces is given as 1--2 mm.(Photographs and 'expertise' in Arbeitspapiere zum Privatissimum''Seidenstrasse'', Förderkreis Schach-Geschichtsforschung (ed.), Kelkheim/Ts., p100--101.
A fourth (full) set of similar pieces was found in Inner Mongolia in 1954. These pieces are made from bronze; the are inscribed on one side and engraved on the other side as well. From an inscribed piece of wood that was found in the same context they were dated to about 1270.(A description and pictures in Nei Menggu chutu wenwu xuanji ('A selection of cultural relics unearthed in Inner Mongolia', Wenwu chubanshe (ed.), Beijing 1963, p123, fig. 165.) It is well later than the other sets but its form is very similar, so I included it here.
All these sets are cast (not minted or really engraved), it seems they have been finished to remove grates, and the two sides are usually differentiated by colours (red and black or green). From a poem (a poem written by Emperor Huizong (r. 1101-1126), included in the collection Xuanhe yuzhi gongci, 'Poems written by the Emperor himself during the era Xuanhe (1119-1126)'.) from the early 12th century we know that there were ivory pieces as well, presumably with engraved characters that were later filled with a lacquer containing powdered gold. We find a similar descrition in a poem by Cai Shen (1088-1156). (This poem is part of the cycle Linjiangxian ('The Immortal of Linjiang'), which is part of the collection You gujushi ci ('Poems of two Scholars of Old'), that has been included into the collection Song liushi mingjia ('Six famous Song poets').)
In Song times we also know of pieces made of porcelain. These have been found at the Liangcheng site near Fengtai in Anhui province; the pieces were black and white. One of the white pieces was engraved with the character 'General', the others were plain; perhaps to be later inscribed with ink. (Description and pictures in Ge Zhihong, Anhui Fengtai 'Liancheng' zhinei faxian Tang--Yuan shidai wenwu ('A number of cultural relics from Tang to Yuan times dicovered at the Liancheng site near Fengtai in Anhui'), in Wenwu ('Cultural Relics') 10/ 1965, p46--56, esp. p49 (description), and fig. 39.)
In the years 1973 to 1974 an excavation of a sea-going vessel was carried out in the Quanzhou bay in Fujian province. During this excavation 20 wooden Xiangqi pieces were found. They must have belonged to two sets, as one piece, a Horse, was marked through an engraved character that was filled up with red paint; the others were inscribed with red resp. black China ink. The ink-marked set had the differentiated Ministers as well. The wreck was dated to Southern Song times (1127-1279). It may be interesting to note that the simpler pieces were found near the crew quarters, so it may be acceptable to presume that Xiangqi began to gain popularity at about this time (though it is of course possible ( maybe even probable) that it became more widespread earlier).
From the inventary of the Imperial treasury for the year 1565 (that is, during the Ming dynasty ( 1368--1644 )), the Tianshui bingshan lu ('Record of Heavenly Waters and Frozen Mountains'), we know that Xiangqi pieces made from jade and ivory were kept there.
In conclusion we may state that
Since it seems safe to conclude that wooden pieces may have been used from the time on that Xiangqi earned popularity (for the simple reason that they were easy to produce and to replace, and most likely quite cheap). Wooden pieces might therefore be relatively frequent, even if their state of preservation might be not too good. Bronze and copper pieces might be somewhat rarer, but they should be better preserved than wooden pieces. Jade pieces should be rarer still, but as they are easily recognized as valuable, they might even be in better state than metal pieces -- jade simply doesn't corrode. One of the problems with the sets of metal pieces describes in short in the first part of this article is, of course, that all of these sets are full. They may seem used, and the paint may be worn off, but nevertheless they all consist of 32 pieces. In my humble opinion that is something to think about: how can it be that we don't know a single set of corroded, imperfect, incomplete metal Xiangqi pieces? It may well be that none thinks it worth while to report, photograph, or describe an imperfect set, but it would remove the nagging doubts about the sets that haven't been found in a properly documented excavation.