How many Bildschirmspiel 01 consoles were built is not documented; at least, we know of no production record. Eyewitnesses involved in the production consistently report around 1,000 units. That is the current state of knowledge. The serial numbers (Einlagenummern) in the archive raise questions, though: they have at most three digits, and some occur more than once (see The serial number). This page sets out the explanations we consider possible and checks, for each one, what the data says about it and what it does not. None of them is proven.
The observation
230 validated devices have a fully readable serial number; 22 of these numbers occur more than once. In November 2026 there were 230 devices and 22 repeated numbers: 20 occurred twice, 2 three times. That makes 26 pairs of devices sharing a number.
The well-known birthday problem shows how many repeats to expect: among just 23 people, there is about a 50 percent chance that two share a birthday, although the year has 365 days. Likewise, two devices carry the same number with a probability of 1 in 999 if every number from 1 to 999 is equally likely. 230 devices form 26,335 possible pairs, so about 26 pairs with the same number are expected on average. 26 were observed. The two numbers occurring three times also match the expectation (2 on average).
The crucial point: this expectation holds for two very different scenarios, A and B. The number of repeats alone therefore cannot tell them apart.
A: Around 1,000 units, numbers in no fixed order
Assumption: About 1,000 units were built, and the numbers were not assigned in sequence but without regard to those already used – for example from a stock of pre-printed stickers or pre-made inserts fitted in any order. Then any number can occur several times and some not at all.
What fits: The number of repeated and triple numbers matches exactly what such an assignment predicts. The numbers are spread across the whole range from 1 to 999, and their last digits are evenly distributed. And the scenario agrees with the eyewitness accounts.
What remains open: A does not explain why a manufacturer would apply three-digit numbers but not use them as unique serial numbers; how the numbers were actually assigned is not documented. With random assignment to 1,000 units, about a third of the numbers would appear on no device at all. With around 230 recorded devices this cannot be checked.
B: Considerably more units, numbers sequential and run through several times
Assumption: The devices were numbered in sequence, 1, 2, 3 … 999, and then counting started again, for example because the numbering stamp only had three digits. Within one run every number is unique; equal numbers only occur between devices from different runs.
What follows: With around 1,000 units there would be practically no repeats. The more units were built, the more often two surviving devices share a number, but never more often than with random assignment (A); as the total grows, the expectation approaches that of A. We simulated, for various totals, how often at least 26 pairs occur among 230 randomly surviving devices (state of November 2026):
| Units built | Expected pairs (mean) | Probability of 26 or more pairs |
|---|---|---|
| around 1,000 | 0 | practically zero |
| 2,000 | 13 | below 0.1% |
| 3,000 | 18 | about 3% |
| 4,000 | 20 | about 10% |
| 10,000 | 24 | about 34% |
| 50,000 | 26 | about 51% |
If B is true, hardly fewer than about 3,000 to 4,000 units were built. No upper limit can be derived: above that, the expected number of repeats changes so little that the data cannot distinguish 10,000 from 100,000 units. The table only answers which totals are compatible with the observation under assumption B, not whether B is true.
An additional check: If the last run stopped before 999, low numbers would occur more often than high ones. With 2,500 units, for example, the numbers up to 502 would have been used three times and the rest twice, and about 60 percent of the devices would have a number below 500. In the archive it is 111 of 230 (48 percent). That speaks against a few runs with a last one stopped early, but says nothing about many runs or a last run ending close to 999.
What fits: Sequential numbering is the usual method for serial numbers, and a three-digit counter readily explains why no number above 999 is known. What speaks against it: B contradicts the consistent eyewitness accounts and requires a production several times larger that no source known to us mentions. The archive would then hold at most about 6 percent of all units built, under A about a quarter. How many devices survived and get reported is unknown, however, so this argument does not rule out either explanation.
C: Special events in production
Assumption: The numbers were meant to be unique, but something broke that. Conceivable are, for example, a faulty or misadjusted numbering stamp that repeated numbers; renumbered returns or replacement inserts; or covering up theft and shrinkage, such as replacing missing units with ones carrying numbers already used, or building units off the books with copied numbers. We know of no evidence for any of these.
What the data says: Isolated errors are not enough. For C alone to explain the observation, it would have to have produced almost exactly as many repeats as purely random assignment; with around 1,000 units, a large share of the devices (almost two thirds with a random-like spread) would then carry a number that at least one other device also carries. An error affecting that many devices would amount to scenario A. Errors limited in time would also mainly affect devices from the same production phase. If the three families correspond to different phases – which is not established – devices sharing a number would then belong to the same family more often than random pairs. In November 2026, 10 of the 26 pairs belonged to the same family (same sticker and chassis colour), about 11 were expected by chance; 3 pairs matched in all four properties, about 5 were expected. Devices sharing a number are therefore no more alike than any two devices. That does not support C, but does not rule it out either, because it is unknown whether the families were built at different times at all. As an addition to A or B that explains some of the repeats, C remains possible.
D: Other explanations
- Several number ranges: Different production lines, plants, batches or years could each have started at 1. This is a variant of B. If the ranges covered the whole span up to 999, the same estimate as for B applies (two full ranges, i.e. around 2,000 units, do not explain the repeats). If they ended earlier, low numbers would have to be clearly more frequent; that is not observed.
- The number does not identify the device: It might, for example, identify the insert itself, perhaps made by a supplier in its own series, or be an inspection, batch or workstation number. Then neither the repeats nor the limit of 999 say anything direct about how many units were built.
- Recording errors: Misread digits (such as 6 and 9 or 1 and 7) or a device reported twice could fake repeats or hide real ones. All counted devices were checked against photos, the last digits are evenly distributed, and devices sharing a number look no more alike than random pairs, which would be different with many double reports. As the main cause this is therefore unlikely; single cases cannot be ruled out.
- Mixed forms: Several mechanisms may have acted together, for example mostly sequential numbers with stickers occasionally used out of order. The data cannot separate such combinations.
- Biased sample: All calculations assume that the recorded devices are a random selection of the surviving ones, and these a random selection of those built. Neither is certain. If, say, most surviving devices came from a single run, B would produce fewer repeats than calculated, and the estimate would shift.
Comparison
| Explanation | Number of repeats | Eyewitnesses (around 1,000) | Additional assumptions needed |
|---|---|---|---|
| A: around 1,000, unordered assignment | fits exactly | fits | unusual numbering method, undocumented |
| B: sequential, run through several times | fits from about 3,000–4,000 units | contradicts | production several times larger, mentioned by no known source |
| C: errors, shrinkage, rework | alone only if most devices were affected | fits | so far without any evidence |
| D: other explanations | depends on the variant | depends on the variant | depends on the variant |
What could settle the question
- Documents from production or trade: plans, delivery notes, sales figures.
- Specific questions to eyewitnesses: How were the serial numbers assigned – with a counting stamp, from a roll of stickers, by hand?
- Manufacturing dates on components, where present, such as date codes on chips or circuit boards: under B, devices sharing a number come from different runs and should carry clearly different dates. Under A, their dates would be no further apart than those of any two devices.
- More recorded devices make all estimates more precise.
Do you know something about the production, or can you read date codes in your device? Then write to kontakt@bss01.de or add your device.
Conclusion
According to the current state of knowledge, around 1,000 units were built; that is what the eyewitnesses report. The repeated serial numbers are fully compatible with this if the numbers were assigned in no particular order (A). But they are just as compatible with a considerably larger production and sequential numbers run through several times (B, then at least about 3,000 to 4,000 units). Isolated production errors (C) can hardly explain the number of repeats on their own, and recording errors are unlikely as the main cause. Which explanation is correct cannot be decided from the numbers alone.
Back to The serial number or the overview.