American Flyer S gauge: the more accurate train set, and the one that lost

In 1946 A.C. Gilbert threw away the third rail, rebuilt every tool in the New Haven plant, and put American Flyer on track that was seven times closer to the real thing than Lionel's. He was right about the engineering and wrong about the business, and twenty years later Lionel bought the wreckage for the price of a house.

What S gauge actually is

S gauge is model railroading at three-sixteenths of an inch to the foot, which works out to a ratio of 1:64, running on two rails set seven-eighths of an inch apart. It sits between HO and O in size, and among the American toy-train gauges it is the one almost nobody outside the hobby can name.

The letter itself was an argument settled by committee. The proportion had been kicking around since the 1890s, and by 1937 the National Model Railroad Association had accepted it under the unlovely name C-D gauge, for Cleveland-Designed. At the 1943 NMRA convention it was renamed. The National Association of S Gaugers records the reasoning with some charm: the letter s appears so many times in three-sixteenths and sixty-fourths that S beat out the rival proposal, Half-1.

None of which would matter to anyone but standards committees, except that three years later the largest toy company in Connecticut bet its entire train division on it.

How Gilbert came to bet on S

Alfred Carlton Gilbert was not a train man. He was the Erector Set man — a physician by training, an Olympic pole vault champion, and a manufacturer of the sort of improving toys that promised to turn a boy into an engineer. In 1938 he bought the American Flyer name from W.O. Coleman, whose family had run it out of Chicago, and moved production to New Haven. (A few accounts put the deal in late 1937; 1938 is the date most Gilbert histories use.) The Train Collectors Association's western division records that Coleman took a profit-sharing arrangement rather than a lump sum, and died in 1939 before it paid him anything.

Gilbert's first move was conservative. He kept the trains on Lionel-compatible three-rail O gauge track, then in 1939 introduced a new line built at three-sixteenths scale — 1:64 bodies — still running on that same O gauge track. The 1939 catalog, a 48-page book titled New Challengers of the High Iron, is where the strategy goes on record. Smaller trains on standard track meant they would take a 20-inch radius curve, which meant a layout that fit a real American living room.

Then the war stopped toy production, and Gilbert spent it thinking.

The features came fast after that, and they are how collectors date the boxes today: choo-choo sound from a motorized piston in the tender, red glowing smoke in 1946, knuckle couplers in 1952, a diesel-roar unit in 1955. Gilbert was selling realism, item by item, year by year.

Why S gauge was the more accurate system

This is usually argued as a matter of taste. It is not. It is a division problem, and it takes about thirty seconds.

American standard gauge track — the real thing, under real trains — measures 4 feet 8½ inches between the rails. That is 56.5 inches. To model it correctly at any scale, you divide.

	Gilbert's S	Lionel's O
Nominal scale	1:64	1:48
Correct track gauge at that scale	0.883 in	1.177 in
Track gauge actually used	0.875 in	1.25 in
Error, in scale inches	About ½ in too narrow	About 3½ in too wide
Rails	Two	Three

Gilbert's seven-eighths of an inch is off by half a scale inch. Lionel's inch and a quarter is off by three and a half. That is roughly a sevenfold difference in error, and it is before you account for the rail running down the center of Lionel's roadbed, which has no prototype anywhere in American railroading and which every O gauge modeler has spent eighty years learning not to see.

The bodies tell the same story. O gauge and O scale share that 1.25-inch track, but traditional Lionel rolling stock was never built to a consistent 1:48 — it was compressed to clear tight curves and toughened for children, with oversized couplers and deep wheel flanges, producing what the hobby politely calls semi-scale. Gilbert's Flyers were closer to a single honest ratio throughout.

One footnote for the pedants, because it is a good one: Gilbert's own figure was not perfect either, and the hobby eventually said so. In 1981 the National Association of S Gaugers revised the S track standard from 0.875 to 0.883 inches, which is exactly 56.5 divided by 64. The NMRA adopted the corrected standards in 2004. It took the hobby thirty-five years to fix a discrepancy of eight thousandths of an inch. Nobody ever bothered to fix Lionel's, because fixing it would have meant admitting what it was.

Why the better system lost anyway

Gilbert won the argument and lost the decade, and the reasons are not mysterious once you stop thinking like a modeler and start thinking like a nine-year-old in 1955.

Lionel owned the word. By 1953 the Lionel Corporation described itself as the largest toy manufacturer in the world, with some two thousand employees. Children did not ask for a train set. They asked for a Lionel, the way they asked for a Kleenex. American Flyer was a good answer to a question nobody was asking.

Two rails made switching harder and accessories fussier. The third rail Gilbert threw away is electrically convenient: it lets you wire a reversing loop without insulating anything, and it makes track-triggered accessories trivial. Gilbert solved all of this, but he solved it with more parts, and more parts is a worse Christmas morning.

He was fighting on a shrinking field. Both companies were selling into a market that was quietly ending. Lionel posted its first annual loss since the Depression in 1958 — $470,000 on sales of $14.5 million — and its response was to diversify into stereo cameras and fishing reels, which went about as well as that sounds. Television, and then the space program, took the imaginative territory that trains had held since the 1920s. The Wish Book pages tell the story better than any balance sheet: the train spread shrinks year over year while the plastic and the rockets expand.

And the cheap seats were already taken. Below both of them sat Marx, selling lithographed tinplate at a fraction of the price. We looked at that squeeze in detail in the guide to the 1952 Sears Christmas Book, which is the year the bottom of the market went plastic.

A.C. Gilbert died in 1961, at 76. The Wrather Corporation took a controlling block of stock the following year. Nineteen sixty-six was the last year American Flyer trains were made, and in February 1967 the A.C. Gilbert Company went bankrupt — the National Association of S Gaugers puts the 1966 loss at $13 million. Lionel bought the American Flyer name and its tooling out of the estate. The figure the NASG records is $157,000. General Mills acquired Lionel in 1969, unused Flyer tooling and all.

There is a familiar shape to this. The better-engineered thing loses to the better-distributed thing, and then the winner buys the corpse and sells the name. It is the same arc we traced through the calculator watch, where the most sophisticated example on the market was also the one that died first.

How to tell a real Gilbert-era set

Gilbert-era American Flyer means 1946 through 1966, made in New Haven. Four checks will place almost any piece, and none of them require a reference book.

• Measure the track. Two rails, seven-eighths of an inch between the inside faces. If there are three rails, it is not S gauge, whatever the listing says. If the rails are an inch and a quarter apart, it is O.
• Look at the coupler. A link coupler — a flat hook and a dangling link rather than a hinged jaw — means 1946 to 1954. A knuckle coupler, the hinged jaw that looks like a real one, means 1952 or later. Link couplers were dropped after 1954, so a link on the car dates it firmly to the first nine years.
• Count the digits in the number. Three-digit catalog numbers — 300, 310, 312, 322, 466, 484 — belong to the earlier run. Five-digit numbers beginning 21 or 24 came in with the renumbering in 1957 and run to the end.
• Read the lettering, then the box. Rolling stock is lettered AMERICAN FLYER LINES on the sides. The A.C. Gilbert Company attribution and the New Haven, Connecticut address live on the boxes, instruction sheets and catalogs rather than on every car, which is one more reason the paper is worth as much argument as the train.
• Check the wheels if you are mixing eras. Original Gilbert wheels are wide, with deep flanges — code 220 in the hobby's shorthand. Modern hi-rail wheels are code 172: narrower, shallower, visibly finer. They run on the same track, which is exactly why mixed lots get described as all original.
• Note the track style. The familiar black-tie T-rail sectional track ran from 1946 into 1960. Pike Master track, introduced in 1961, uses a different rail profile and plastic ties and is not what people picture when they picture Flyer track.

The look-alikes that get sold as Flyer

Five things get listed as postwar American Flyer S gauge. One of them is, and the most persistent impostor is Gilbert's own work.

What it is	How it fools you	The tell
Gilbert 3/16 scale on O track, 1939–1942	Identical 1:64 bodies. Same company, same designers, genuinely the same trains to look at.	Three-rail wheels on inch-and-a-quarter track. 400- and 500-series numbers. It is prewar O gauge wearing S gauge clothes.
Lionel-made American Flyer, 1979 onward	It says American Flyer, because Lionel owns the name and has since 1967.	Modern box art, modern catalog numbers in the 6-4xxxx range, code 133 track. New tooling in an old name.
American Models, and the old S-Helper Service / MTH line	True S gauge, runs on Flyer track, sits in the same box lots.	Finer detail than Gilbert ever managed, code 172 wheels, code 148 track. These are modelers' trains, not toys.
Lionel O gauge	Sellers who inherited a basement full of trains use the two brand names interchangeably.	Three rails, inch and a quarter apart. Not close. Measure it.
Pre-1938 Chicago American Flyer	Same brand name, and prewar Standard Gauge pieces carry big numbers in price articles.	Different company, different city, different gauge entirely. Those four-figure Standard Gauge results have nothing to do with what a postwar S gauge set is worth.

That last row is responsible for more disappointed sellers than any other single fact in this category. Price lists headlined most valuable American Flyer trains routinely mix 1928 Standard Gauge President's Specials in with postwar S gauge freight sets. They are not the same market and never were.

What American Flyer sells for now

Here the honest answer is uncomfortable, and worth stating plainly before the table: common postwar American Flyer is cheap, and getting cheaper.

The figures below are realized hammer prices from Stout Auctions, which specializes in toy trains and publishes its results. They are what the material actually brought, not what anyone hoped for. Buyer's premium sits on top.

Lot	Sale	Hammer
Boxed set No. 20365 The Challenger, engine, 8 cars, accessories	Sept 11, 2026	$1,000
Boxed set No. 20558 Truscott, complete and unused, with station	Sept 11, 2026	$600
Three No. 790 Trainorama sets, original boxes	Sept 11, 2026	$400
Freight set No. 20740 Defender, engine 234 and 5 cars	Sept 11, 2026	$300
Mint No. 20800 Game Train, complete and unused	Sept 11, 2026	$190
Klondike set, engine 21207 and 5 cars, original box	Sept 18, 2025	$150
Boxed set 5106WT, engine 325AC, 4 cars, original boxes	Sept 18, 2025	$130
Yard King set K5440T, engine 343, 4 cars	Sept 18, 2025	$100
Set with 360/361 Alco AB and 4 cars, with controls	Sept 18, 2025	$90
Set with 312 Pennsylvania Pacific	Apr 16, 2026	$70
No. 593 Signal Tower, original box	Sept 18, 2025	$70
No. 785 Coal Loader, original box	Sept 18, 2025	$60
Set with 290 engine, 3 cars, boxes and accessories	Sept 18, 2025	$50
Set with 300 engine, 3 cars, transformer	Sept 18, 2025	$50
310 engine with eight freight cars	Apr 16, 2026	$40

Read the bottom half of that table twice. A complete set — locomotive, three to five cars, often the original boxes, often a transformer and a couple of accessories — clears at fifty to a hundred and fifty dollars, repeatedly, at the house that sells more of this material than anyone. An engine with eight cars made forty.

What dealers ask, which is a different number

These are asking prices on live listings, not sales. The gap is the whole point.

Item	Dealer ask
Boxed Yard King set, 1956 — the same set that hammered at $100	$400
Boxed Thunderbolt set, 1958	$280
New Haven Clipper set 5670TRH, 1957	$875
484/485/486 Santa Fe Alco ABA, graded C-6	$336
342 Northern Pacific switcher	$225
322 Hudson with smoke, damaged box	$184
21573 New Haven EP-5 electric	$150
499 New Haven EP-5, graded C-6	$148
350 Royal Blue Pacific	$110
312AC Pennsylvania Pacific	$85
Common freight cars, each	$10–60

Where the market is going

Down for common material, and flat to strong at the top. The reason is demographic and nobody in the hobby pretends otherwise: the people who collect postwar American Flyer are overwhelmingly the people who got one as a child, and that cohort is now in its seventies and eighties. As those collections come to market, supply rises exactly as demand falls. WorthPoint's own dictionary entry says it flatly — as the generations die off, the secondary market for common and above-average trains diminishes.

What holds value is what always holds value in a thinning market: complete, boxed, unused, unusual. A Challenger in its box at $1,000 and an engine with eight cars at $40 are the same hobby on the same day.

The track that was right

There is a particular kind of object that is worth collecting precisely because it lost, and this is one of them. The American Flyer S gauge line is a complete, coherent, fully realized argument about what a toy train ought to be, executed at enormous expense by a man who was right, and then comprehensively beaten by a competitor who was wrong in a way that turned out not to matter.

You can still see the argument on a table. Put a Gilbert Pacific on its two rails beside a Lionel steamer on its three, and the thing Gilbert spent his company's money on is immediately, obviously visible — the roadbed reads as railroad rather than as toy. He was not imagining it. The numbers say he was right by a factor of seven.

It is also, at forty to a hundred and fifty dollars a set, one of the last genuinely complete vintage systems a person can still assemble on an ordinary budget. The locomotives are diecast, the track is steel, the catalogs are beautiful, and almost nobody is bidding against you. That will not be true forever, but it is emphatically true right now, and there is something fitting about the more accurate model finally being the affordable one.