The most you should bid on an eBay auction is what the item actually resells for, minus eBay’s cut, minus what it costs you to get it delivered, divided by the margin you need — and eBay auction sniping does nothing at all to change that number. Sniping solves the timing problem. It’s real, it works, and it is not the expensive part of buying to resell. The expensive part is deciding the number you’d walk away at and then actually walking away.
This page is about that number: how to work it out, the two mistakes that wreck it, and what a real scan of live eBay auctions looked like when we pointed our own tool at the category we sell in.
Sniping is a bid-timing trick, not a price
eBay runs proxy bidding. You enter the most you’re willing to pay, and eBay bids on your behalf in the minimum increments, never going above your maximum and never paying more than one increment over the runner-up. That mechanic is the whole reason sniping exists — and the whole reason it can’t save you.
What a snipe actually buys you is reaction time. Bid three days out and you’ve announced that the item is worth something to a stranger, and given them seventy-two hours to raise their own maximum. Bid with eight seconds left and nobody gets to re-evaluate. That’s the mechanism. It’s a good one.
What it doesn’t do: lower the closing price, or stop you overpaying. The highest maximum wins whenever it was entered. If your maximum is wrong, a sniping tool just makes you overpay quietly and on schedule.
Worth knowing before you plan a snipe at all: across 73,038 sold eBay auctions in our own data, 26.5% closed on a single bid. A quarter of the time nobody turns up, and the item goes for whatever the seller set as the opening price.
How to work out a maximum bid you can defend
Four steps, in this order:
- Start from what the item actually sells for. Not the keyword-matched average — the cleaned one. A search for a model number returns the parts, the accessories and the lots alongside the thing itself, and averaging that pile gives you a number for a product nobody sells. We took our own 864,000-row sold database apart to show exactly how badly this goes in our breakdown of sold-comp research tools.
- Take off the cost of selling it. eBay’s final value fee runs around 13% plus a fixed per-order charge in most categories (check your own category), plus whatever you’ll spend shipping it out. What’s left is your net.
- Divide by the margin you need. Divide, don’t subtract. If you want 30% on cost, that’s net ÷ 1.30. A $173.10 net becomes $133.1538 — and it gets truncated to $133.15, never rounded up to $133.16. Rounding up gives away the last cent of the exact margin the ceiling exists to protect.
- Take the inbound shipping off the bid. This is the one people get wrong, and it has its own section below.
Shipping comes out of the bid, not out of the profit
Winning an auction costs you the hammer price plus whatever the seller charges to ship it. So a listing with $40 shipping has a maximum bid $40 lower than the identical machine with free shipping. It is not a $40 haircut on your profit at the end; it’s $40 you cannot bid. The same subtraction runs on a live auction feed, except that there the buyer’s premium has to be divided out of the bid rather than taken off it.
On a live scan against our own connected account, three listings of the same miner came back with ceilings of $97.03, $57.03 and $31.85. Same machine, same resale value, same margin requirement. The only variable was what each seller charged to ship it. A fourth row’s ceiling came back at $0 — the shipping alone exceeded what the machine resells for, so there is no bid that makes money, including $1.
Related trap: unstated shipping is not free shipping. eBay’s API omits the shipping block entirely on some listings, and a tool that reads a missing value as zero will hand you a ceiling that’s too high by however much the seller eventually charges. Treat unknown as a number you have to guess, and take the guess off the ceiling before you bid.
A current bid three days out is not a price
This is the defect that makes most “underpriced auction finder” lists worthless. A $12 auction with three days left is not a $12 item. It’s an auction that hasn’t happened yet, and the discount you’re looking at is an artefact of nobody having bid.
Our rule is a hard one: anything more than 24 hours out is marked “too early to price”, contributes nothing to any total on the board, and carries its ceiling forward as the number to come back with when the clock is real. Fixed-price listings are never too early — their price doesn’t move.
The same reasoning inverts a rule that’s right everywhere else. A price floor — nothing this cheap is the real thing — is sound for a Buy It Now and exactly backwards for an auction, which legitimately opens at $0.99. Auctions get judged on identity alone, and a suspiciously cheap one gets flagged to go and read the description rather than silently dropped.
What a real scan looked like
Sixty-nine live listings across two searches. Sixty-one dropped as not the same item. Zero worth bidding on.
The four rows that survived the filter were an S21+ asking $1,950 with thirteen hours left, an S21 at $800 with one bid on it, and two S19 Pros at $219 and $300. Every one sat above what the comps supported, so every one came back Pass, with the amount it was past the ceiling in red: $1,822, $672, $203.
Now the part that’s uncomfortable to publish. Every one of those rows also carries an “Insufficient Evidence” tag. The 20 sold comps behind “sells for $200” on an S21 are themselves contaminated — hashboards, control boards and multi-unit lots matched by the same keyword. So the honest reading of that board is not “S21s sell for $200.” It’s “nothing here clears the bar, and the price basis isn’t strong enough to argue with anyway.” The app says so on the row rather than dressing a weak number up as a verdict.
Across three real scans over six search terms, the board has never yet produced a single bid-worthy row: every live miner was priced above comp value. That’s the feature working. It also means the bid-worthy path is verified against 69 analyser tests and a synthetic browser payload rather than against a real underpriced auction, and we’d rather say that than imply otherwise.
Half of what you searched for isn’t what you searched for
A keyword search for a technical product returns a landslide of things that are not that product but name it in the title. Three that were live on eBay when we ran the scan:
- “Cooling Fan 4 pin fit Bitmain Antminer S19” — $15. Names the brand, names the model, and got priced against a $148 machine until we started matching “fit” as a word rather than a substring. The reason it slipped through is that “benefit ” also ends in “fit “.
- “ASIC Miner Hosting Europe — Antminer S21” — $1.00. A hosting contract. Exactly the right model name, and not a thing you can resell.
- “Overclock Antminer S21, adds 10–20%” — firmware. A service listing, priced like a machine by anything matching on words alone.
Fans, fan controllers, fan simulators, rubber standoffs, risers, shrouds, PSU harnesses and hashboards all cluster around $15 and all name the model. After we added them to the rejection list, one live scan went from rejecting 48 of 71 listings to rejecting 58. That ratio is normal for technical categories, and it’s why an auction board that shows you everything cheap is worse than no board at all.
When a dedicated eBay auction sniping tool is the better answer
Plainly: if your problem is timing, use a sniping service, not us. Gixen has been running a free last-second bidding service for years and it does that job well. It places the bid. We don’t — ING Listing Engine has never placed a bid on eBay and doesn’t intend to. The method exists in the codebase and is called from nowhere, because software that spends your money while you’re asleep is a different product with different risks.
We’re the other half of the problem: which listings are worth looking at, what the ceiling is, and whether the flip pays after fees. Run both. Get the number here, enter it as your maximum, and let a sniper place it at the buzzer if that’s how you like to bid.
One more honest boundary: if you sell mainstream catalogued goods — phones, consoles, cameras, sneakers — Terapeak’s three years of first-party sold data is a better price basis than our scrape, and it’s free. Our comp database earns its keep in the awkward categories: ASIC miners, servers, enterprise networking, industrial gear.
What the Auction Sniper does today
Shipped as of 26 July 2026, and described as it currently behaves rather than as a roadmap:
- Pulls live auctions and Buy It Nows through eBay’s search API. Read-only — nothing lists, relists, reprices, messages or bids.
- Leave the search box empty and it builds the watch list from your own completed sales, grouped by product — “you’ve sold 18 of these.” Those are the only items whose demand, shipping cost and condition grading you already know first-hand.
- Prices every surviving row against cleaned sold comps and returns one number: the most you can bid, to the exact cent.
- Sorts by what closes first, not by margin. The biggest margin on the board is worth nothing if it closes while you’re reading a different row.
- Lists risks instead of scoring them — thin comp evidence, unstated shipping, a seller with three feedback — because a score hides the exact thing you need to go and look at.
- Hands a row to the Deal Pipeline with one button, freezing the ceiling and its basis so you can check later what you actually paid against what you said you’d pay.
- Bounded per scan: 8 search terms, 50 listings per term per format, one comp lookup per term.
It’s part of the same buy-side toolkit as the Opportunity Finder, which does this for supplier files and local pickup rather than auctions, and it shares its arithmetic with the repricing side that works on stock you already own. The listing half — turning the thing you won into a live listing — is the AI listing engine.
ING Listing Engine is a Windows desktop app and it’s free while it’s in open beta — no card, no seat limit. If you want to check the arithmetic on this page against your own category, that’s the fastest way to do it.
Common questions
Does eBay allow auction sniping?
Yes. Bidding in the closing seconds is bidding, and eBay’s proxy system treats a maximum entered with eight seconds left exactly like one entered three days earlier. What eBay prohibits is bid manipulation — shill bidding, retracting bids to manipulate a price, or coordinating with other bidders.
What’s the most I should bid on an eBay auction?
Take what the item actually resells for, subtract eBay’s fees and your outbound shipping to get a net, divide that net by one plus the margin you need, then subtract the shipping the seller will charge you. Truncate to the cent rather than rounding up. That number is your maximum, and you enter it once.
Does sniping actually lower what you pay?
Not directly. It denies other bidders the time to react and raise their own maximums, which can keep a price from climbing. It cannot beat someone whose maximum is higher than yours, whenever they entered it.
Why do most underpriced-auction finders show junk?
Two reasons. They treat a current bid as a closing price, so an auction three days out looks like a 90% discount when it’s simply an auction that hasn’t happened. And they match on keywords, so a $15 cooling fan gets priced against the $148 machine named in its title.
Does ING Listing Engine place bids for you?
No. It reads live listings and tells you the most you can bid. It never places a bid, never sets a max on eBay, and never spends money on your behalf. If you want automated last-second bidding, use a dedicated sniping service alongside it.
Does this work outside electronics?
The maximum-bid arithmetic works anywhere. The comp data behind it is strongest in the categories our sold database has depth in — ASIC miners, servers, networking and industrial equipment. Consumer categories price thinly or not at all, and the board says “insufficient evidence” rather than guessing.
Filed under Pricing & Profit · Tagged Fees & margin, Sold comps, ASIC miners, Dirty data · Browse all guides by topic