Which track system suits you?
Two-rail or three-rail, ballasted or not, which system exactly — with real prices for the same example layout in every track system.
1. Why this is the second most important question
You have decided how much space you have, and from that your scale — H0, TT, N or Z. The next question follows immediately: which track do I buy?
It feels smaller than the scale decision. It isn’t. It determines:
- which radii you can run at all — and therefore which locomotives will work on your layout,
- which stations you can build — not every system offers curved turnouts or double slips,
- how much work the trackwork is — with or without ballasting,
- and what the whole thing costs you — not just at first purchase, but with every turnout you ever add.
Switching later means selling your track (at a loss) and buying new. There is no gentle migration. That is why an hour of thought here pays off more than anywhere else.
2. The chain of decisions
Four questions, in this order. Each one narrows the field.
- Two-rail or three-rail? (only relevant in H0) → §3
- With or without ballast? → §4
- Which system exactly? → §5
- Which rail height? (code 100 / 83 / 75) → §6
Turn the order around and start with “which brand is best?” and you land in the endless forum debates. Back to front there is no answer — front to back, usually only two or three systems are left.
3. Two-rail or three-rail?
What is this about? With almost all manufacturers the current comes through the two rails (two-rail, DC). In H0, Märklin uses a third conductor in the middle of the track — formerly a centre rail, today a row of small contact studs (Punktkontakte, “Pukos”).
Why does it matter? This is the only decision in the chain that affects not just the track but the rolling stock as well. A three-rail locomotive will not run on two-rail track and vice versa. You are not choosing a track here, you are choosing an ecosystem.
What does it cost you? Decided wrongly, this is the most expensive mistake in the whole hobby: it is not only the track that hangs on it, but your entire collection of locomotives and coaches.
How do you decide?
- Three-rail (Märklin) if you have inherited or second-hand Märklin stock, if reversing loops without wiring effort matter to you, or if you are in the Märklin world anyway.
- Two-rail if you want to buy freely from every manufacturer, if track appearance without contact studs matters to you, or if you are planning outside H0 — in N, TT and Z the question effectively does not arise.
In our libraries: 48 two-rail systems, 6 three-rail. The three-rail systems are all Märklin/Trix Express.
4. With or without ballast?
What is this about? Ballasted track brings the track bed along as a plastic moulding — you click it together and it looks finished. Track without ballast is just rail and sleeper base; you add the underlay (cork) and the ballast yourself.
Why does it matter? It determines how your layout building goes — and how it looks.
What does it cost you? Two items work against each other here:
- Ballasted track has the higher unit price.
- Track without ballast needs cork underlay and ballast as a separate item — plus the working time.
In our libraries 45 of 59 systems are unballasted, 14 ballasted. If you choose an unballasted system, budget for the ballast from the start instead of noticing once the track is down.
How do you decide?
- Ballasted if you build up and take down on the carpet or a table, if the layout has to be rearranged, if children play with it, or if you want to run trains quickly rather than spend a long time crafting. Ballasted track is sturdier underfoot and more reliable in operation.
- Unballasted if the layout is built permanently and the result should look as real as possible.
A detail that often surfaces too late: flex track — the metre length you bend freely — barely exists in ballasted systems. If you want flowing alignments, you are better served without ballast. In our data several ballasted systems (Kato Unitrack, Piko A with ballast, Bachmann E-Z Track, Roco geoLine) have no flex track; Rocoline with ballast is one of the exceptions.
5. Which system exactly?
What is this about? Usually only a few systems are left now. They differ in four points, which are worth checking in this order.
5.1 Is the smallest radius enough for your locomotives?
The smallest production radius is the hard limit of your layout. It differs markedly between systems — in H0 between 343 mm and 457 mm, in N between 150 mm and 550 mm.
This is not a nicety: Kato Unitrack HO starts at 430 mm, Roco Line and Rocoline at 358 mm. If you are short of space, Kato simply gives you fewer options. Conversely Rocoline goes up to 1962 mm — a radius in which long express coaches also look right.
The curve radii in your system determine which vehicles you can run. See also: minimum radius (check rule), radius (glossary).
5.2 Does it have the turnouts your station needs?
Any system will do for a simple loop. As soon as you plan a real station you need more: curved turnouts for tight track spacing, three-way turnouts and the double slip. That is the most common argument in the forums against an otherwise good system.
From the data (H0, current systems): a double slip is offered by Märklin C, Piko A, Piko A with ballast, Roco geoLine, Roco Line, Rocoline with ballast and Trix — essentially the continental European systems. No double slip in Kato Unitrack HO, Bachmann E-Z Track, Tillig Elite, Hornby and all Atlas and Peco systems.
5.3 How wide is the range?
More shapes means more freedom when planning. The spread is considerable: from 5 parts (Micro Engineering H0) to 68 parts (Kato Unitrack N).
Careful with the interpretation: a small range is not a knock-out criterion if the system has flex track — Peco Streamline carries no fixed curves at all, because everything there is bent from flex track. That is a different planning style, not a shortcoming.
5.4 What does a turnout really cost?
This is the item that blows up most calculations. In most systems the turnout motor is not included: the turnout comes in the box as a manual one, and the electric drive is a separate article you buy per turnout.
From our data across all 59 systems:
- 38 systems sell the drive separately — one extra article per electric turnout.
- 4 systems additionally offer their own built-in decoder for digital operation (Märklin C with drive 74491 + decoder 74462, Märklin K with 74490 + 74461, Roco geoLine with 61195 + 61196, LGB with 12010 + 55025).
- 3 systems have the drive integrated (Kato Unitrack).
Work this through on your own plan before you commit: a medium-sized station quickly has twelve turnouts — and therefore twelve drives, roughly €240 for Märklin C in drives alone. In some systems this block is larger than all the track put together.
Careful with the number of drives: “one turnout = one drive” is not always true. A three-way turnout needs two drives (Märklin C 24630, Piko A 55225 — both manufacturers say so explicitly), a double crossover up to four. Anyone calculating by turnout count undercounts exactly the most expensive components.
5.4a Digital is cheaper than most people calculate
This is where the reasoning error sits that makes digital operation look needlessly off-putting. People read “one decoder per turnout” and multiply. That is the more expensive of two valid routes:
| Type | Example | Drives per unit | 12 drives |
|---|---|---|---|
| Built-in decoder (sits inside the turnout) | Märklin 74462 · €38.27 | 1 | €459.24 |
| Decoder module (centrally mounted) | Märklin m83 (60832) · €62.99 | 4 | €188.97 |
| Decoder module | Viessmann k83 (5211) · €35.22 | 4 | €105.66 |
One four-channel module switches four drives. The requirement is therefore a division with
rounding up, not a multiplication: ceil(drives ÷ channels). At twelve drives, the gap between
the cheapest and the most expensive route is a factor of 4.3.
What you trade for it: the module sits centrally under the layout, and a pair of wires runs to it from every drive. The built-in decoder saves exactly that wiring — a real argument for a small layout, rarely one for a station.
The decoder has to match the digital format, not the track manufacturer: three-rail layouts (Märklin) run Motorola/mfx, two-rail layouts run DCC. A Viessmann k83 is a Motorola module — right for Märklin, wrong for a DCC layout.
And analogue is not free either. If you throw your turnouts electrically but without a digital command station, you need a control panel (Märklin 72720 from €21.59, Viessmann 5549 €27.84, Roco 10520 €33.21). We show the item but deliberately give no quantity: none of the three manufacturers states in its product data how many turnouts one panel switches.
The Trackplanner knows the drive model and the drive’s article number for every system. That is why drives appear as their own line in the parts list — you see the item before you order, not after. In the build plan the electrics appear as build stages (manual → electric → digital), each with its surcharge; the decoder requirement is calculated from the channel count, not from the turnout count.
5.5 Is the system still being made at all?
The point a price comparison hides most easily. A track system is not a one-off purchase: you buy more later. In two years you will be three curves short, a turnout will fail, the station will grow. A system that has been discontinued can then only be topped up second-hand or from remaining stock — at prices nobody can predict.
That is exactly why our tables carry the status next to every price:
| Status | What it means | What it means for you |
|---|---|---|
| current | in production and in the shops | can be topped up, even years from now |
| discontinued | production ended, remaining stock | often cheap — but the supply is finite |
| historic | no longer made | second-hand, swap meets, leftovers only |
Of the 59 systems recorded, 46 are current, 1 discontinued (Märklin K) and 12 historic (including Märklin M, Fleischmann Profi, Trix Express, Piko Hobby, Pilz TT).
Does that make discontinued systems off limits? No. If a discontinued system happens to be cheap right now, it can be a good choice — for a layout with a fixed final state, for an exhibition board, for getting started on a small budget. But two things you need to know:
- The low price is partly a clearance price, not a permanent property of the system. It says nothing about what the next component will cost you.
- Buy what the final state needs right away. With a current system you can top up piece by piece; with a discontinued one, today’s stock is your stock.
Wherever this knowledge base names a system that is no longer in production as the cheapest one, the caveat stands right next to it — together with the price of the cheapest system still in production, so you can see what future-proofing costs you.
6. Rail height: code 100, 83 or 75?
What is this about? “Code” is the rail height in thousandths of an inch. Code 100 = 2.5 mm, code 83 ≈ 2.1 mm, code 75 ≈ 1.9 mm.
Why does it matter? Lower profiles look markedly more realistic. Higher ones are sturdier and more forgiving — both when laying track and with older vehicles whose deep flanges can bottom out on code 75.
How do you decide?
- Code 100 if you run older or second-hand vehicles, if reliability comes before looks, or if the track gets handled often.
- Code 83, today’s standard for newly built layouts — a good compromise.
- Code 75 and finer if you photograph your layout and run modern vehicles exclusively.
Within one brand, profiles can be joined with transition tracks — so you do not have to commit for the whole layout, but you do for the visible areas.
7. Can you mix systems?
Short answer: yes, with limits. In two-rail DC the manufacturers are standardised; electrically it fits together. Mechanically you need transition rail joiners, and with different rail heights or at a ballasted ↔ unballasted transition you have to compensate for the height difference.
What does not work: two-rail with three-rail.
The usual and sensible case is not “mix everything”, but: one main system, supplemented by individual parts from another — a turnout shape your system does not offer, for instance.
In the Trackplanner you can load several libraries into the same plan and try the combination out before you buy anything.
8. The comparison tables
The complete generated tables per scale are in our generated comparison dataset (produced from the track libraries).
Extract H0, current systems, by range width:
| System | Parts | Current | Ballast | smallest R | largest R | Flex | Double slip | Drive |
|---|---|---|---|---|---|---|---|---|
| Bachmann E-Z Track | 50 | 2-rail | yes | 381 mm | 902 mm | no | no | — |
| Tillig H0 Elite (code 83) | 50 | 2-rail | no | 366 mm | 543 mm | yes | no | separate |
| Rocoline w. ballast | 49 | 2-rail | yes | 358 mm | 1962 mm | yes | yes | separate |
| Märklin C | 43 | 3-rail | yes | 360 mm | 1115 mm | no | yes | separate + decoder |
| Roco Line code 83 | 41 | 2-rail | no | 358 mm | 1962 mm | yes | yes | separate |
| Peco H0 code 100 Streamline | 30 | 2-rail | no | flex only | — | yes | no | separate |
| Piko A with ballast | 29 | 2-rail | yes | 360 mm | 908 mm | no | yes | separate |
| Kato Unitrack HO | 28 | 2-rail | yes | 430 mm | 790 mm | no | no | integrated |
| Piko A | 25 | 2-rail | yes | 360 mm | 908 mm | yes | yes | separate |
| Roco geoLine | 21 | 2-rail | yes | 358 mm | 511 mm | no | yes | separate + decoder |
Extract N, current systems:
| System | Parts | Ballast | smallest R | largest R | Flex | Double slip | Drive |
|---|---|---|---|---|---|---|---|
| Kato Unitrack N | 68 | yes | 150 mm | 718 mm | no | yes | integrated |
| Minitrix | 66 | no | 195 mm | 526 mm | yes | yes | separate |
| Atlas N code 55 | 51 | no | 254 mm | 1803 mm | yes | yes | — |
| Fleischmann Piccolo | 45 | no | 192 mm | 430 mm | yes | yes | separate |
| Roco N | 45 | no | 195 mm | 765 mm | yes | yes | separate |
9. The same layout in every system
Comparing systems on individual unit prices is misleading. So the generator builds the same small but complete layout in every system — and counts what it takes:
Oval (two half circles + two straight sides)
+ passing siding (2 turnouts)
+ stabling siding (1 turnout)
= 3 turnouts, so 3 drivesAll three are ordinary left/right turnouts, so “one turnout = one drive” really does hold here
(with a three-way turnout it would be two, §5.4). In digital operation one single four-channel
module is added — not three decoders: ceil(3 ÷ 4) = 1. For Märklin C that is €62.99 instead of
€114.81.
The layout is a real plan, not a calculation
This is where this comparison differs from most. Adding up a parts list from target dimensions is easy — and can contain parts that never join into a closed layout. An oval only works out if the curve angles add up to 360°; a passing siding only if there is a curve that brings the turnout’s diverging route back parallel, and if the straights fill exactly the gap that creates.
That is why the layout is actually built and checked in every system: every joint has to be coincident (gap 0), the loop has to close, the siding has to meet both turnouts. What does not close does not appear here as a number. These six systems carry the layout — and you can look at each one as a public plan:
| System | Parts | Size | Curve radius | Track spacing |
|---|---|---|---|---|
| Peco H0 code 100 Setrack | 31 | 2208 × 938 mm | 438 mm | 62 mm |
| Piko A | 35 | 2316 × 906 mm | 422 mm | 62 mm |
| Piko A with ballast | 35 | 2324 × 906 mm | 422 mm | 62 mm |
| Märklin C | 36 | 2255 × 953 mm | 438 mm | 78 mm |
| Roco geoLine | 39 | 2238 × 946 mm | 435 mm | 77 mm |
| Roco TT (TT gauge) | 41 | 1705 × 774 mm | 365 mm | 44 mm |
Every row is a real, publicly available plan — look at it, rotate it, rebuild it:
| System | View plan |
|---|---|
| Peco H0 Setrack | vergleichsanlage-peco-h0-code-100-setrack |
| Piko A | vergleichsanlage-piko-a |
| Piko A with ballast | vergleichsanlage-piko-a-mit-bettung |
| Märklin C | vergleichsanlage-maerklin-c |
| Roco geoLine | vergleichsanlage-roco-geoline |
| Roco TT | vergleichsanlage-roco-tt |
Tillig H0 Elite works out — but only just. With Elite turnouts (diverging at 15°, 12°, 9.4° or 6.34°) and only 30° curves, it is not obvious that the passing and stabling sidings can be closed without flex track. In the current dataset it succeeds with 37 catalogue parts. The character of the system is still a different one: Elite is a finescale system for scratch builders, where flex track is the normal case — if you want to plan from catalogue parts only, the geometric systems make it easier.
There is therefore no public plan for Tillig yet; the figure comes from the same calculation, but without a plan to click through.
Why the parts count varies so widely
The overview below still comes from the older, calculated method and therefore covers more systems. The verified plans above are what the cost statements rest on; this table shows the effect at issue:
| System | Parts for the same layout | of which curves |
|---|---|---|
| Hornby Standard | 24 | 8 |
| Peco code 100 Setrack | 28 | 8 |
| Bachmann E-Z Track | 30 | 12 |
| Kato Unitrack HO | 34 | 16 |
| Rocoline with ballast | 38 | 12 |
| Piko A · Atlas True-Track | 39 | 12 |
| Märklin C · Roco Line · Roco geoLine | 41 | 12 |
| Trix | 43 | 12 |
| Atlas H0 code 100 | 54 | 24 |
For exactly the same layout, Atlas code 100 needs more than twice as many parts as Hornby. That comes down to the grid: 15° curves instead of 30°, shorter straights. Comparing only the unit price of a curve means comparing the wrong number — what matters is how many pieces the layout needs.
On top of that, almost every system needs three turnout drives. In digital operation one four-channel decoder module is enough for this layout (§5.4a) — three built-in decoders would be almost twice as expensive for Märklin C (€114.81 against €62.99). With Kato Unitrack the drive block disappears entirely, because the drive is part of the turnout’s price.
Five H0 systems do not appear in the table: Peco Streamline, Peco code 75, Peco code 83, Shinohara and Micro Engineering carry no production curves — there the oval is formed from flex track, and the parts count depends on the radius you choose. Atlas code 83 is missing because our library carries no simple left/right turnout for it (see the data gap in §10).
The complete parts lists with article numbers are in our generated comparison dataset.
What the layout costs (DACH)
Price date: 11 August 2026 · Source: Modellbahn Union · Currency: euro Prices change. The date is in the generated dataset (column “price date”) and is carried along on every retrieval — the figure here is an extract from it, not a hand-maintained year.
What is calculated is the actual purchase: for every position the matching listing is chosen and rounded up to whole packs. At MU, track pieces are single articles, so no surplus arises here.
| System | Status | Parts | Cost of the example layout |
|---|---|---|---|
| Piko A | current | 35 | €226.36 |
| Peco H0 code 100 Setrack | current | 31 | €244.30 |
| Tillig H0 Elite (code 83) | current | 37 | €258.35 |
| Piko A with ballast | current | 35 | €276.18 |
| Roco geoLine | current | 39 | €283.93 |
| Märklin C | current | 37 | €437.56 |
These figures are comparable with one another — same scale, same layout, same region, same currency, same day.
Other gauges deliberately stay out of this table. Roco TT costs €171.77 for its version of the layout, but a TT layout is smaller (1705 × 774 mm instead of a good 2.2 × 0.9 m). Putting that number next to the H0 figures would make a gauge effect read as a price advantage of the system. Choosing between scales is a decision about space, not about the price of track.
All six are current systems — so here the cheapest is not the one being discontinued. If it were, the caveat from §5.5 would stand at this point.
Three things can be read from this that usually get lost in the forums:
Parts count and price do not run in parallel — they even run against each other. Peco Setrack needs the fewest parts at 31 and is still more expensive than Piko A with 35. Comparing systems by “how many parts do I need” gets it wrong; comparing them by the price of a single curve does too. Only the finished layout answers the question.
Ballast costs a premium — and we can now quantify it, but only halfway. Piko A and Piko A with ballast are the same layout at the same size; the ballasted variant costs €49.82 more (+22 %). That is the premium against bare track — with plain Piko A, underlay, ballast and adhesive still come on top. The underlay can be calculated (Noch cork, 3 m per pack, €16.63–20.79); for ballast (€11.70 / 500 g) and adhesive (€6.64–17.05) no manufacturer states a reach in metres of track. We therefore show these two items as “price known, quantity unknown” instead of inventing a number. The real gap between the two variants is thus smaller than €49.82 — but by how much exactly, the sources do not say.
Märklin C is by far the most expensive — and the reason is the turnouts, not the track. €437.56 against €226.36 for Piko A, so almost double for the same layout. It is not the metre of track: the three curved turnouts (24711/24712) cost €38.69 each, plus a drive per turnout (74491, €19.99) and a decoder (74462, €38.27). That block alone is around €290 — more than the entire Piko A layout. If you run C track, this is exactly where you pay for the operational reliability of the three-rail system.
This total calculates the built-in decoder per turnout, because Märklin offers one for C track itself. With a four-channel module instead of three individual decoders (§5.4a) the electrical block drops by €51.82, putting the layout at €385.74. That does not change the conclusion — Märklin C remains the most expensive of the six — but it shows how much hangs on this single decision.
A procurement note: with the standard turnouts (24611/24612, 24.3°) the layout would be somewhat cheaper, but not obtainable complete from our DACH dealer — the matching counter-curve 24224 is missing from their range. The plan therefore uses the curved turnouts, which are the operationally better choice anyway.
What the layout costs (UK)
Price date: 11 August 2026 · Source: Harburn Hobbies · Currency: pound
| System | Parts | Cost |
|---|---|---|
| Peco H0 code 100 Setrack | 31 | £195.30 |
The same layout is noticeably cheaper in the UK than in DACH — £195.30 against €244.30 for exactly the same 31 parts. That is not an exchange-rate effect but a home market: Peco is a British manufacturer, and British dealers cost their own range differently from an importer. If you are planning in the UK, Peco deserves a more serious look than the DACH table suggests.
For the remaining Peco ranges (Streamline, code 75, N, TT) individual articles are still missing at the dealer; they only get a total once every position is priced.
10. What is still missing
Costs are available for DACH and the UK (§9): Modellbahn Union (367 priced articles, seven systems complete) and Harburn Hobbies (161 articles, one system complete). Still open:
1. Gaps in our own library data. While pricing, it became clear that several positions carry no real manufacturer article numbers at all:
- Rocoline with ballast and Roco Line: parts with the internal identifiers
D4,D8,D12instead of a Roco article number — not orderable and not priceable that way. - Roco Line: combined numbers such as
40295/40296(drive left/right) and42616/617. Partly solved (2026-08-10): such numbers are now split into their parts when pricing, so the article is found under the individual number (42616/617→42616, found). That does not replace proper data maintenance, but it makes the pieces findable immediately. The split is deliberately cautious:3900 D1/1(Märklin) is left untouched, because the slash is part of the name there. - Atlas H0 code 83: no simple left/right turnout in the library (10 parts in total).
- Piko N: a placeholder instead of an article number.
This is not a finding about the manufacturers but about our data — and work on our side. As long as it is open, the two Roco systems of all things are missing from the cost comparison.
2. Märklin C — done. The missing decoder was the old number 74461; the successor 74462 has been retrieved (€38.27), and the layout now uses the curved turnouts whose counter-curve the dealer stocks. Märklin C is therefore complete in §9.
2a. USA without a total — and that stays that way for now. The US source (modeltrainstuff)
sells track largely in packs, and the piece count is only a convention in the product title.
Calculating from that would mean guessing (errors by a factor of 4 to 8), so the shop is set to
content_kind = unknown and deliberately returns no total. Solvable only with curated pack
contents or a second US source that sells singly.
2b. UK coverage. Peco Setrack is complete; for Streamline, code 75, N and TT, three to five
positions each are missing that Harburn does not stock. A second UK source (Rails of Sheffield)
would probably close the gaps — but their products.json sits behind bot protection.
3. Ballast as a cost item. For unballasted systems, underlay, ballast and adhesive are
still missing from the comparison — around 7 metres of track for the example layout. The
roadbed_products table and the quantity calculation are in place, and nine articles are now
maintained and priced. For the underlay the reach is documented (Noch cork, 3 m per pack); for
ballast and adhesive no manufacturer states a reach in metres of track, so they remain
without a quantity. Only then can the table in §9 be switched between “bare track” and
“including ballast”. Until then: the figures compare ballasted track with bare track, not
with fully ballasted track.
4. Turnouts with several drives — solved, the data maintenance is still running. A three-way turnout needs two drives, not one. Until 2026-08-10 the parts list counted turnouts instead of drives and therefore priced these parts too low; since 10 August 2026 it counts drives, and the decoder requirement is a division by channel count instead of a multiplication (§5.4a).
What remains open is the curation: across three-way turnouts, double slips and single slips our libraries hold 67 candidates; two are documented and entered so far (Märklin C 24630, Piko A 55225). All the others still count one drive. That is a decision, not an oversight: the research on Märklin C 24624 (double slip) produced contradictory information — several dealers write that the drive is already built in there, which would be a third situation not yet modelled. Setting a number you cannot document would be the same mistake as a guessed pack size.
For the tables in §9 this changes nothing: the example layout uses only ordinary left/right turnouts in every system. The correction takes effect on your own plans as soon as one contains a three-way turnout.
5. Digital operation for two-rail systems still has no total. The four-channel modules for Motorola/mfx are priced (m83, k83 — §5.4a). For DCC we are still missing a priced module: the documented candidate, ESU SwitchPilot V2.0 (51820), is not stocked by our DACH source. Until that is resolved, the build plan shows the digital stage for two-rail systems openly as “cannot be quantified” instead of estimating a price.
6. Control panel channels unknown. Prices for the common control panels are available (§5.4a), but none of the manufacturers states in its product data how many turnouts one panel switches. The panel therefore appears as a position without a quantity, and the electrical total of the analogue stage starts with “from” instead of “total”. Same rule as for ballast and adhesive (point 3): a named gap is better than an invented number.
7. The other regions. For the UK one complete total now exists (Peco Setrack, §9); the remaining systems stocked there are not yet priced throughout. For the USA the pack content is not curated (§9): the piece counts differ per listing and appear only in the product title, so no layout has a total there. Both are data maintenance within the existing model, no longer an open design question.