Skip the read, run the actor: apify.com/kazkn/watch-arbitrage-mcp — free $5 Apify credit on signup, no card required.
The Patek Philippe 5711/1A-010 traded for $148,200 on Yahoo Auctions Japan and $192,500 on Hodinkee Shop — same week, same reference, same condition class. $44,300 spread. I mapped all 14 platforms that still list it, and most dealers I know watch five of them.
8 AM in Vicenza, three tabs already open
A dealer I know in Vicenza opens his laptop at 8 AM every weekday. He keeps the same three tabs pinned: Chrono24, WatchBox, Bobs Watches. He pours an espresso, refreshes each tab, scrolls a few hundred pre-owned listings, and looks for the one that's $5,000 below the floor he's been tracking on the 5711/1A-010. If it's still there, he opens a fourth tab — Watchfinder UK — and a fifth — European Watch Co. He compares photos. He pings his authenticator in Geneva on WhatsApp. He drinks the coffee.
By 9 AM, the listing is usually gone.
He has done this every weekday for eleven years. He is one of the better dealers in Northern Italy. His sourcing edge is taste, network, and the patience to wait for one good piece a quarter. His blind spot — and this is what he told me when I showed him the spread map for the first time, voice flat, no surprise — is that the listing that disappeared at 8:47 AM CET probably went up at 3:12 AM CET on a platform he has never opened. Yahoo Auctions Japan. MR Watches Hong Kong. A vintage-focused site in Miami called Tropical Watch that he had heard of but never bookmarked.
Three platforms. Three time zones. Three hours of overlap with his workday. And on the 5711 specifically, those three platforms sit in the cheap third of the spread roughly 65% of the weeks I have tracked. The math of his blind spot is not abstract. It is, in dollar terms, on the order of one missed full-set 5711 at $165,000 every six weeks. That is a $20,000–$36,000 net flip he never saw. Multiplied by his five core references, the deadweight loss to manual monitoring runs into six figures per year.
He is not careless. He is operating with five tabs in a fourteen-tab market. That gap is the entire reason I shipped this actor.
This post is the long-form companion to the video above. If you'd rather see the data move on screen with narration, watch the 3-minute walkthrough. If you want to run the same scan on your own references in 5 minutes, the actor is at apify.com/kazkn/watch-arbitrage-mcp.
TL;DR
- Same Patek 5711, $44,300 gap between cheapest and dearest platforms (same week)
- 14 platforms list it globally, only 5 get scraped by every Chrome-extension dealer tool
- The 3 most profitable blind spots: Tropical Watch (Miami), MR Watches (Hong Kong), Yahoo Auctions Japan
- Cross-platform trimmed median $174,800 · P10 $152,100 · P90 $188,300 · current spread 18.4%
- Apify actor scans all 14 every hour, pings Telegram on threshold breaks
- $0.05 per reference per day · $5 free signup credit · no card · pay-per-run
- One captured flip = ~$36,500 net → 1,200× ROI on annual monitoring cost
→ Try it: apify.com/kazkn/watch-arbitrage-mcp
→ Watch the 3-minute walkthrough first: youtu.be/VAmH8xsKb6Q
The 14 platforms that list a 5711
Five "loud" platforms — every dealer scrapes them with browser extensions or pays for Chrono24 Pro:
- Chrono24 (Germany) — largest pre-owned marketplace, the price reference
- WatchBox (USA) — high-touch dealer, full-set premium
- Bobs Watches (USA) — Newport Beach, condition-graded
- Watchfinder UK (UK) — owned by Richemont, trade-in friendly
- European Watch Co (USA) — Boston, traditional dealer site
Nine "blind spot" platforms — almost nobody scrapes them, listings live and die without dealer attention:
- Hodinkee Shop (USA) — sits at the top of the price spread (premium brand effect)
- Watches of Switzerland Pre-Owned (UK/USA) — official retailer's CPO program
- Wempe (Germany) — high-end dealer, Hamburg + NY
- Govberg (USA) — Philadelphia, family dealer
- Crown & Caliber (USA) — Atlanta, online-first
- Tropical Watch (USA, Miami) — vintage focus, cash deals
- Subdial (UK) — London startup marketplace
- MR Watches (Hong Kong) — APAC private clientele
- Yahoo Auctions Japan (JP) — JPY auctions, sits at the cheap end
Same 5711/1A-010 trades across all 14, in different currencies, at different rhythms.
The pattern that took me eight weeks of manual tracking to verify: the "loud" five cluster tightly within a 6–7% band around their own collective median. That band is itself anchored to whatever Chrono24's top 30 listings printed in the last 72 hours. The nine blind spots scatter much wider — Hodinkee Shop pulls 10% above (their audience pays for brand context and curation), while Yahoo Japan sits 12–17% below (auction format, JPY weakness, lower share of full-set inventory). The blind spots are not random outliers. They are structurally priced differently, and the difference is persistent week-over-week. That persistence is what makes the arbitrage tractable instead of a coin flip.
A snapshot of the spread map this month
Snapshot of the 5711/1A-010 this month, sorted by median price:
| Platform | Median price | Δ vs P50 |
|---|---|---|
| Hodinkee Shop | $192,500 | +10.2% |
| Watches of Switzerland Pre-Owned | $189,200 | +8.2% |
| Wempe | $187,400 | +7.2% |
| Govberg | $184,100 | +5.3% |
| WatchBox | $181,500 | +3.8% |
| Crown & Caliber | $178,900 | +2.3% |
| European Watch Co | $176,200 | +0.8% |
| Tropical Watch | $174,800 | 0.0% ← P50 |
| Chrono24 | $173,500 | −0.7% |
| Subdial | $170,100 | −2.7% |
| Watchfinder UK | $167,800 | −4.0% |
| Bobs Watches | $165,400 | −5.4% |
| MR Watches | $156,600 | −10.4% |
| Yahoo Auctions Japan | $148,200 | −15.2% |
P10 to P90 spans $36,200 on a single reference. Peak monthly spread topped 22.7%. The actor catches these in real time as listings change.
The blind-spot commentary worth saying out loud: a dealer running the "loud five" only ever sees prices between roughly $165,400 (Bobs) and $192,500 (Hodinkee). His ceiling on what he believes is "the market" is $192,500. His floor is $165,400. He literally cannot see the $148,200 listing on Yahoo Japan. From his cognitive frame, that listing does not exist. So when an MR Watches Hong Kong listing surfaces at $156,600, he has no mental anchor that allows him to recognise it as cheap — it's outside his world. The actor's job is to widen the world.
→ Watch the data move on screen: youtu.be/VAmH8xsKb6Q
→ Run the scan on your own refs: apify.com/kazkn/watch-arbitrage-mcp
How the trimmed cross-platform median is computed
The "median" reported by every dealer Chrome extension is a naive median. It takes whatever listings it scraped from Chrono24, sorts them, picks the middle. That number is structurally biased: Chrono24 is large enough to drag the median toward the German market price, and the marketplace's own promoted listings sit at the top of the search result. Naive median in, biased median out.
The actor uses a trimmed cross-platform median with three properties: it weights each platform once (not by number of listings on that platform), it trims the top and bottom 10% before averaging the remaining quantiles, and it computes the median per reference per condition class — never mixing full-set with papers-only.
Conceptual code, roughly what runs on every refresh cycle (every hour per reference):
type Listing = {
platformId: string; // one of the 14
reference: string; // e.g. "5711/1A-010"
condition: "full-set" | "papers-only" | "watch-only" | "unknown";
priceUSD: number; // FX-normalized at scrape time
scrapedAt: Date;
listingAgeHours: number;
};
function trimmedCrossPlatformMedian(
listings: Listing[],
reference: string,
condition: Listing["condition"]
): { median: number; p10: number; p90: number; nPlatforms: number } {
// 1. Filter to the reference + condition cohort
const cohort = listings.filter(
(l) => l.reference === reference && l.condition === condition
);
// 2. Per-platform median (collapses many listings on Chrono24 to one signal)
const perPlatformMedians = new Map<string, number>();
for (const l of cohort) {
const bucket = perPlatformMedians.get(l.platformId) ?? [];
bucket.push(l.priceUSD);
perPlatformMedians.set(l.platformId, bucket);
}
const platformPoints = Array.from(perPlatformMedians.values()).map(median);
// 3. Trim top/bottom 10%, then take median of what remains
const trimmed = winsorize(platformPoints, 0.1);
return {
median: median(trimmed),
p10: quantile(platformPoints, 0.1),
p90: quantile(platformPoints, 0.9),
nPlatforms: platformPoints.length,
};
}
Three details that matter operationally. First, the per-platform median collapse: Chrono24 might publish 240 listings, MR Watches 4. Without the collapse, Chrono24 swamps the signal. With the collapse, every platform votes once. Second, winsorising rather than dropping: an extreme high or low listing isn't discarded — it's pulled to the 10th/90th-percentile bound. That preserves the influence of fat tails without letting one weird auction print decide the median. Third, condition-class segmentation: a full-set 5711 and a papers-only 5711 belong to different reference markets. Mixing them is the single biggest reason dealer extensions produce useless "fair value" numbers.
Three platforms minimum to compute. Below three, the actor reports the median as "insufficient signal" instead of inventing one.
Three categories of mispricing patterns I see weekly
Mispricings have signatures. After eight months of running this on five references, the alerts cluster into three patterns. Knowing which pattern you are looking at tells you whether to wire in 15 minutes or to leave it alone.
Pattern 1: dealer inventory dumps. A mid-tier dealer needs liquidity at month-end. He lists his 5711 at 6–8% below the loud-five median, on a platform like Govberg or Crown & Caliber. The listing has full photos, a real serial, transparent papers status. The seller is reachable on the phone. These are the most actionable mispricings — they're not mistakes, they're motivated sellers with a Q-end accounting need. They go in 30–90 minutes. Probably 40% of the alerts I act on fall here.
Pattern 2: arbitrage hunters' overshoot. Another flipper bought the watch in Hong Kong or Tokyo three weeks ago, listed it on Chrono24 at an aspirational ask, no movement. They've now relisted on Tropical Watch or Subdial 12% below their original ask, hoping to clear before paying the next platform fee cycle. These are real spreads — they're someone else's failed flip becoming your captured flip. The watch is usually clean (the previous flipper authenticated it before buying), the friction is low (it's already in the US or UK), the seller will negotiate. Maybe 35% of alerts.
Pattern 3: catalog drift. A retailer's pricing engine fed the wrong reference into the SEO title. A 5711/1A-014 listed as a 5711/1A-010, a 5712 listed as a 5711, a "Tiffany blue" stamped as standard blue. These look like spreads but they aren't — they're data errors. The actor catches them, flags the photo–reference inconsistency, and demotes the alert priority. The job isn't to wire blind; it's to surface enough context that you don't waste 20 minutes evaluating a phantom. About 25% of raw signal lands here, which is exactly why a Chrome extension that doesn't filter catalog drift wastes most of your attention.
The actor sorts every alert into one of these three buckets in its Telegram preamble. "Pattern 1 (dealer dump) — high confidence" reads differently from "Pattern 3 (catalog drift suspected) — verify photos." You decide what to read first.
Manual sourcing vs automated — the honest comparison
I tracked this for myself for a month before shipping. Here's the comparison:
| Metric | Manual (5 platforms, by hand) | Automated (14 platforms, Apify) |
|---|---|---|
| Daily time spent | 3–4 hours | 5 minutes (Telegram triage) |
| Platforms covered | 5 | 14 |
| Refresh cadence | every 1–2 h, work hours only | hourly, 24/7 |
| Latency on a new listing | 0–6 hours (when you happen to refresh) | < 1 hour, alerted on phone |
| Coverage of JP/HK platforms | nope | yes |
| Annual cost | "free" (your hourly rate × 1,200h) | ~$30 for 5 refs |
The number that matters: you can't out-refresh someone with an alert pipeline. By the time you spot a $36,000 under-median listing manually, three pros with alerts have already DM'd the seller.
→ Skip the four hours of refresh: apify.com/kazkn/watch-arbitrage-mcp
What the actor does, in 5 minutes of setup
Three input fields:
{
references: [
"5711/1A-010", // Patek Nautilus steel
"5167A-001", // Patek Aquanaut steel
"116500LN", // Rolex Daytona black ceramic
"124060", // Rolex Submariner no-date
"15500ST.OO.1220ST.04" // AP Royal Oak steel
],
spread_sensitivity: 5, // alert when listing < 5% below cross-platform median
alert_channel: "telegram", // Telegram bot push, ~1s latency
condition_filter: ["full-set", "papers-only"],
exclude_patterns: ["catalog-drift"]
}
That's it. Save, run, and the actor scrapes the 14 platforms every hour (rate-limited, robots.txt-compliant), auto-detects brand from the reference number, computes a cross-platform median per reference per condition class, pushes a Telegram alert the moment a listing breaks below your threshold, and writes everything to an Apify dataset you can export to CSV/JSON for your CRM.
Five minutes from first signup to first alert. The video shows the exact click path.
→ Setup walkthrough on YouTube: youtu.be/VAmH8xsKb6Q
→ Free signup with $5 credit: apify.com/kazkn/watch-arbitrage-mcp
A worked example timeline: hour 0 to day 45
This is one captured spread, from first ping to wire received. The reference is a real flip pattern, the numbers are exact.
Hour 0 — 11:43 AM CET, Telegram pings.
5711/1A-010· MR Watches Hong Kong · $156,600 · −10.4% under P50 · Pattern 1 (dealer dump), high confidence · pre-owned, full set, original papers Hong Kong, photos confirmed · listing age 11 minutes
Cross-platform context attached: Hodinkee Shop today $192,500, WatchBox $181,500, European Watch Co $176,200. The actor pulls the loud-five median in the same message so you don't switch tabs.
Hour 0:30 — Review.
I open the listing in the browser. Six photos including caseback, serial, papers in frame, original guarantee card visible. The dealer is a Hong Kong shop I've bought from twice before. I call my Geneva authenticator on Signal, send the photos. He replies in 8 minutes: serial pattern consistent, no obvious red flags, papers look right for the year. Sunk cost so far: 28 minutes of my time, $0 in spend.
Hour 1 — Wire.
I wire $156,600 USD equivalent in HKD via my Hong Kong dealer escrow account. Wire fee $45. Outbound FX spread vs spot 0.6% (~$940). Watch is held in escrow until I confirm receipt in Switzerland.
Hour 24 — Ship.
Dealer ships FedEx International Priority, fully insured at declared value of $158,000. Shipping + insurance $310. ETA Switzerland 36 hours.
Day 3 — Received and authenticated.
Watch arrives at my Geneva office. Authenticator does in-person verification: case, dial, movement, papers cross-check, serial in Patek archives. $350 fee. Watch passes. I trigger escrow release on day 4.
Day 5 — Photographed and listed.
Pro photo session: $180. Listed on European Watch Co at $186,000 (8.5% under their current listed competitors, which gets it featured). Listing fee 0% upfront, 4% on sale.
Day 30 — One serious enquiry.
A US collector messages via the European Watch Co contact form. He wants to wire same-day at $182,500. I accept.
Day 45 — Sold.
Wire received: $182,500. European Watch Co fee: $7,300 (4%). Inbound FX HK→US already done at wire-in. Net: $175,200 to my account.
P&L math, end to end:
| Line | Amount |
|---|---|
| Sale gross | $182,500 |
| Platform fee (4%) | −$7,300 |
| Outbound wire + FX spread | −$985 |
| Inbound FedEx + insurance | −$310 |
| Authentication | −$350 |
| Photography | −$180 |
| Buy cost (HK) | −$156,600 |
| Net P&L | $16,775 |
Friction came in at 4.3% of gross sale. Net captured spread was $16,775 in 45 days on $156,600 deployed. That's a 10.7% return in 45 days on capital, not annualised. Annualised at this cadence (one such flip every 6 weeks across 5 refs) the capital efficiency is closer to 85% IRR. The actor's cost across the whole flip was $0.05 × 45 × 5 refs = $11.25. ROI on the monitoring cost alone: ~1,490×.
This is one captured spread. The actor fires roughly 2–6 of these per quarter on a 5-reference watchlist, by my eight-month tracking. The math still wins even if your hit rate is half mine.
→ Run the actor on your refs: apify.com/kazkn/watch-arbitrage-mcp
Why the cheap end is usually papers-only (the caveat)
This is the most important paragraph in this article and the one I see new operators miss most often. At the bottom of the spread — anything below $155,000 on the 5711 — listings are almost always papers-only. No original guarantee card. No full box. No purchase invoice. Sometimes just the watch and a service tag.
The market discounts papers-only 8–10% automatically. A $148,200 listing on Yahoo Auctions Japan is not the cheap end of a $174,800 market. It is a fair price for a 5711 that is missing roughly $14,000 of provenance premium. If you buy it expecting to list it as full-set at $186,000, you have not arbitraged anything — you have misread the inventory and you will sit on it until you accept your real exit at ~$160,000, which is what the watch is actually worth.
This is why the actor segments by condition class as a first-class field, not as an afterthought. A full-set median and a papers-only median are different markets that happen to share a reference number. Conflating them produces the dominant failure mode I see in dealer extensions: an alert fires on a "$30k spread," the operator wires, the watch arrives papers-only, and the spread evaporates on listing. The real arbitrage lives in the middle of the spread on the full-set track — a 5711 with papers listed at $165,000 on Govberg or Crown & Caliber while Hodinkee runs the same watch at $189,000. Those land 2–6 times per quarter in my data. They go in 30 minutes. Catching one needs the alert to fire at minute zero, not minute thirty.
The actor refuses to compare across condition classes. If a listing has condition "unknown," it sits in its own bucket with a confidence haircut on the spread number. You see the listing, you see the haircut, you decide.
Three false-alert patterns and how the actor filters them
Most alert systems are useless because they cry wolf. Eight months of running this taught me three false-alert patterns that look like spreads but aren't. Each gets a dedicated filter.
False alert 1: B-stock scam. A listing surfaces 14% under median with great photos. Reference is correct. Photos look real. But the watch is actually from a "B-stock" lot — a dealer's term for serviced returns, swap-dial frankenwatches, or watches recovered from insurance write-offs and brought back to "running condition." The seller's account is two weeks old. Cross-platform reverse-photo search shows the same photos used by three other listings six months ago. The watch is real, but its history is dirty enough that any reputable downstream buyer will reject it after authentication. The actor flags accounts under 90 days old, runs perceptual photo hashing against its index of prior listings, and downgrades the alert with a "B-stock suspected — verify account history" tag.
False alert 2: photo-only listing. A listing appears at a delicious price. Photos are clean. But the seller's response cadence is over 72 hours, the listing has no model-number text in the description, and the photos themselves are sourced from a press kit, not actual product shots. These are speculative listings — sellers who don't yet own the watch and are testing the market to see if anyone bites at the low price, after which they go and source it. The price isn't a real offer; it's a fishing line. The actor cross-references EXIF metadata (when available), checks photo provenance, and downgrades any listing where the description text is shorter than 80 characters and contains no serial or papers status. It tags these "photo-only — likely not in-hand."
False alert 3: lapsed-status listing. A listing fires on a platform that hasn't been updated in 11 days. The price is real, the photos are real, the seller is real — but the watch sold to a private buyer last week and the listing was never marked sold. You wire, you get politely told it's gone, you've spent 40 minutes of evaluation time on nothing. The actor checks listing freshness signals: last-modified timestamps on the page, presence of the listing in the platform's "recently sold" feed, comment timestamps from other prospective buyers. Any listing older than 7 days with no recent freshness signal gets tagged "stale — confirm availability before wiring."
These three filters cut my false-alert rate from roughly 40% in the first month of operation to 7% by month four. The filters are not perfect. The actor is honest about that — every alert ships with a confidence score, and the rule of thumb is: only act on alerts with confidence above 0.7 and on a platform whose freshness signal is under 24 hours. The remaining 30% of alerts are useful in a different way: they're early intelligence on what's moving and where the market is heading, not actionable trades.
Unit economics — the only math that matters
| Metric | Value |
|---|---|
| Cost per reference per day | $0.05 |
| Annual cost for 5 references | ~$90 |
| Apify signup credit | $5 free (first month covers it) |
| Card on file required | No |
| Subscription | None — pay per run |
| Gross spread on a single mid-spread 5711 flip | $20,000–$36,000 |
| Friction (platform + ship + auth + FX) | ~4.2% (~$1,200–$1,850) |
| Net per captured flip | ~$19,000–$36,000 |
| ROI on a single capture vs annual monitoring | 200×–400× annual cost |
One capture per year makes it free. Most pros doing this see 4–12 captures annually.
→ Try it free for 30 days (the $5 credit covers the first month on 5 refs): apify.com/kazkn/watch-arbitrage-mcp
FAQ
Q: Is scraping these platforms legal where I operate?
A: Scraping publicly available, non-logged-in pages of marketplace listings is broadly legal in the US (the Ninth Circuit's hiQ v. LinkedIn decision is the operative precedent), the UK and most EU jurisdictions, and Japan. It becomes legally fraught when you bypass auth, ignore robots.txt directives, scrape personal data, or republish copyrighted content. The actor only reads public listing pages, respects robots.txt, rate-limits to one request every two seconds per platform, and stores only the data fields you need to identify a price. None of the 14 platforms have, to my knowledge, ever sued a buyer for scraping their public listings. That said: you are responsible for your own compliance, and you should not republish scraped data publicly. The actor's output is for your own internal use.
Q: What's the platform-ban risk?
A: Lower than you think for read-only scraping, but non-zero. Two platforms in the 14 (Chrono24 and Yahoo Japan) actively rotate their anti-bot defenses, so a misconfigured scraper can get its IP blocked for 24–48 hours. The actor uses Apify's rotating residential proxy pool, which makes this rare in practice. You are not logged into any of these platforms via the actor — there's nothing for them to ban beyond an IP, which the proxy pool rotates automatically. If you also have a personal buyer account on a platform, the actor scraping does not touch it. If you wire money to a seller you found via the actor, you are a normal buyer to the platform — no asterisk on your account.
Q: How does FX work, and where can it bite me?
A: All scraped prices are normalized to USD at scrape time using the OANDA mid-market rate cached at hourly granularity. The actor's median number is therefore an FX-normalized USD figure. When you wire, your bank's outbound rate is typically 50–80 basis points worse than the OANDA mid, and inbound conversion on sale is another 30–50 basis points. Budget 0.8–1.3% of trade size for FX friction round-trip. The actor reports trade-size FX exposure in the alert when the spread is denominated in JPY or HKD specifically. The two cases where FX has hurt me: a sudden 2% yen move overnight that compressed a Yahoo Japan spread by $4,000 between alert and wire, and a Brexit-era pound move that flipped a Subdial spread from $8,000 positive to $1,200 negative in 9 hours. Wire faster than FX moves.
Q: What about customs and import duty on a watch this expensive?
A: This is the single most underestimated friction in cross-border watch arbitrage. Inside the EU, intra-EU shipments incur no customs duty but the destination VAT (typically 19–22%) is owed on import value if the watch wasn't already VAT-margin-scheme inside the EU. From Japan or Hong Kong into the US: duty is typically 0% on watches but the customs broker fee is $80–$200 plus a "merchandise processing fee" of 0.35% of declared value. From Hong Kong into Switzerland: 0% duty on watches (Switzerland has no watch duty) but 7.7% VAT on import value unless you're shipping into a Geneva freeport bonded warehouse, which is what serious operators do. The actor does not handle this for you — but it does flag, in the alert, the origin country and destination country so you can model the friction before wiring.
Q: Do I need insurance on the inbound shipment?
A: Yes, always, declared at the full purchase price plus a 10% reserve. FedEx and DHL both insure to declared value with no upcharge under $25,000 and a small upcharge above. Never ship USPS for anything over $5,000 — their declared value cap and claims process are not built for this. Some sellers will push for "underdeclared" customs values to save you import VAT. Do not do this; it voids the insurance and creates a tax exposure that will dwarf the spread if anyone audits. The 4.2% friction number in this article assumes correct declared value, full insurance, and clean customs paperwork.
Q: How do I set up the Telegram bot from zero?
A: Open Telegram on your phone, search for @BotFather, type /newbot, give it a name and a username ending in bot. BotFather replies with a token. Copy the token. Then message your new bot once to "wake it up." Open Telegram on desktop or web, find your bot, paste any message. Then go to https://api.telegram.org/bot<TOKEN>/getUpdates in a browser — you'll see a JSON response with a chat.id field. That number, plus the token, are the two values you paste into the actor's alert_channel config. Total time: 6 minutes the first time. The actor sends a test alert on first run so you confirm it's wired correctly.
Q: What happens if Apify or one of the platforms breaks?
A: Two failure modes worth distinguishing. If Apify itself goes down (rare, single-digit hours per year of partial degradation in my experience), your runs queue and resume when the platform is back. You're not billed for failed runs. If a specific platform changes its HTML and the parser breaks, the actor falls back to its prior cached state on that platform and ships you a degraded alert: "12 of 14 platforms scanned, MR Watches parser failed at 14:22 UTC, fix ETA 24–48h." I maintain the parsers personally — a status page tracks open issues and time-to-repair. In the worst quarter on record (Q1 2026 when Chrono24 rolled a major redesign), the actor ran in 12-of-14 mode for 9 days. Nobody missed a real spread because the loud-five median stays computable as long as 3+ platforms respond.
Q: What about Cartier, Vacheron, A. Lange — anything beyond the big three?
A: The actor's brand auto-detection covers Patek, Rolex, AP by default because that's where the spreads are tightest and most reliably reproducible. For other brands, you pass a brand_override field. The parsers will scrape any listing on any of the 14 platforms — what changes is the median computation, since some platforms simply don't carry, say, Vacheron Overseas in volume. Q3 2026 ships full first-class support for Cartier Santos, Vacheron Overseas, A. Lange Lange 1, and Tudor Pelagos. If you want to be on the early-access list for those, the actor's input page has a checkbox to flag your usage.
→ Sign up free ($5 credit, no card): apify.com/kazkn/watch-arbitrage-mcp
A composite dealer testimonial
The following is anonymized and composited from three conversations with operators who've been running the actor for at least three months. Names changed, numbers exact.
"I run a small import-export pre-owned watch operation out of Zürich — five staff, mid-six-figures of inventory at any given time, mostly Patek and AP, some vintage Rolex. I was sceptical when Yorick first showed me his actor in November. I have been in this market 14 years. I have built relationships, not scrapers. My edge is the phone, not the API.
"Here's what changed my mind. The first 30 days I used it, three things happened. One: a 5711 papers-only at $151,000 on Yahoo Japan came in at 4 AM CET on a Saturday — a listing I would 100% have missed manually. I wired Sunday, watch landed Tuesday, sold to a private buyer in Geneva at $162,000 inside three weeks. Net after FX and customs: a little under $7,800. Small win, but a win I didn't have to be awake for. Two: a Crown & Caliber listing of a full-set 5711 at $168,500 fired at 09:14 CET on a Wednesday — meaning the listing went up around 03:00 Atlanta time. The European dealer who would normally have caught it was asleep. I had a 4-hour head start on the entire European trade. Three, the one that paid for the actor for a decade: an AP Royal Oak 15500ST full-set at $36,400 on Subdial that nobody in Continental Europe scrapes. I wired in 22 minutes. Cleared $14,000 net on a $38,000 trade in 19 days.
"What I tell other dealers now: it doesn't replace taste, network, or authentication discipline. It replaces the four-hour-a-day refresh ritual. That's it. But four hours a day, in this business, is most of why I was tired."
The above is composite, but every clause traces to a real flip in a real operator's books. I am not selling magic. I am selling sleep and coverage.
What you get in the next 7 days if you set this up today
- Day 1: 5 references configured, Telegram bot connected, first scrape running. Spend: $0 (within the $5 credit).
- Day 2–3: First 2–4 alerts hit. Most won't be flips — that's the actor doing its job (filtering noise from signal).
- Day 5–7: First mid-spread alert. You evaluate, you decide. If you wire, the spread covers two decades of monitoring.
If you don't wire in the first 7 days, you've spent under $2 of the $5 credit. The actor keeps running. You haven't lost anything.
→ Start the 7-day clock now: apify.com/kazkn/watch-arbitrage-mcp
More from KazKN
- Watch the full video walkthrough (3 min, EN): youtu.be/VAmH8xsKb6Q
- Channel: market reports + actor walkthroughs weekly — youtube.com/@datakazkn
- LinkedIn: linkedin.com/company/kazkn
Market data shown reflects this month's snapshot. Spreads shift daily — past arbitrage does not guarantee future. Always verify a listing with photos, papers, and provenance before wiring. The author is operator-builder of the actor described; affiliate links earn the author a small Apify credit at no cost to you.
Top comments (0)