Is Stop Hunting Real? What Happens at Round Numbers

Stop hunting, price running to a level where many stop-loss orders rest so that they trigger, is real as a crowd effect and, on exchange-traded futures, mostly a myth as a personal one. Stop-loss orders (orders that close a position once price reaches a set level) cluster just beyond round numbers and, once triggered, push price further, but on CME futures a resting stop stays out of the visible order book, so no market maker or other trader can see yours.

Key takeaway

Round numbers pull stop-loss orders to just beyond them, so a trade through a round number can set off a chain of stops that overshoots the level. On CME Globex, individual stops are invisible to other traders until triggered; only the firms that hold or route a stop (the exchange, the broker, the platform, a CFD dealer or a prop firm's simulator) can see it, and the documented cases of deliberate stop triggering involve dealers holding their clients' orders. Stops placed where the trade idea fails, rather than one tick past the obvious level, sit outside most of the crowd.

Where do stop-loss orders cluster around round numbers?

Stop-loss orders cluster just beyond round numbers, while take-profit orders (orders that bank a gain at a target) cluster exactly on them, according to Carol Osler's studies of the first available data on individual currency stop-loss and take-profit orders. The abstract of Osler (2003) in the Journal of Finance states: "Take-profit orders cluster particularly strongly at round numbers, which could explain the first prediction. Stop-loss orders cluster strongly just beyond round numbers, which could explain the second prediction." The two predictions come from technical analysis: trends reverse at support and resistance, and speed up once a level is crossed.

Osler's 2002 New York Fed staff report on price cascades describes the data: the complete order book of the Royal Bank of Scotland from August 1999 to April 2000, 9,655 orders with a face value above $55 billion in dollar-yen, dollar-pound and euro-dollar. Weighted by value, 14.3 percent of executed stop-loss buy orders sat at rates ending 01 to 10, against 6.9 percent at endings 90 to 99, where an even spread across the 100 two-digit endings would put 10 percent in each band. The zone just above a round number therefore held 2.07 times as much buy-stop value as the zone just below it; sell stops mirror the pattern below the number. Take-profit orders sat on the number itself: 9.9 percent of executed take-profit orders by value ended in exactly 00, against 3.8 percent of stop-loss orders and 1 percent under an even spread.

Take-profits sit on the round number and stops sit just past it. That asymmetry is why price often stalls at a round number and then accelerates once it breaks through.

Osler's order-level evidence comes from one bank's currency orders in 1999 and 2000. This article found no study measuring round-number stop clustering in CME futures, so applying the pattern to E-mini S&P 500 (ES) or Nasdaq-100 (NQ) futures is an inference from the same mechanism (people pick the same obvious prices), not a measured result.

Can market makers see your stop loss on CME futures?

No: on CME Globex, CME's electronic trading platform, a stop order waits off the order book until a trade reaches its trigger price, so it does not appear in the market data that market makers and other traders receive until it fires. CME's Globex order-type documentation says a stop "does not immediately go on the book, but must be 'triggered' by a trade in the market", and adds: "There are two types of Stop order: the Stop-Limit, which goes on the book as a Limit order when activated, and the Stop with Protection, which goes on the book as a Market order." The depth-of-market ladder shows resting limit orders, never untriggered stops.

Fett and McPhail (2017), research economists in the CFTC's Office of the Chief Economist, reached the same conclusion: "Existence of stop orders is generally unknown to market participants as stop orders are not visible in the orderbook but must be triggered by a trade in the market at the corresponding price." Using CME's complete transaction audit trail for 2014 to 2016, they counted 22,789,175 stop-order contracts out of 2,483,400,538 ES contracts traded, about 0.9 percent of volume.

Only CME, the firms that carry your order to it (your platform or routing vendor and your broker, the futures commission merchant, or FCM, that clears it) and regulators know about a CME-native stop. Everyone else can only guess, and the guess is easy: just beyond round numbers and recent highs and lows. A trader pushing price toward a level is betting on where a crowd put its stops, not reading yours.

Who can see your stop: broker, platform, CFD dealer or prop firm?

Whoever holds the stop can see it, and where it is held depends on the order type, the route and the account. According to TradeStation's help pages, a stop-market order on its CME route is either held on the broker's Stop Server and sent to Globex as a market order once triggered, or routed natively to Globex, depending on the trigger setting. NinjaTrader's Simulated Stop, by contrast, is a locally held order that waits on your own computer until triggered.

Stop type (where it lives)	Other traders see it?	Broker sees it?	Exchange sees it?	Works with your PC off?	Fills against
CME-native stop-limit or stop with protection (CME matching engine, off the book)	No	Yes	Yes	Yes	Real exchange liquidity
Broker-held synthetic stop (broker's server)	No	Yes	Not until triggered	Yes	Exchange liquidity, after an extra routing hop
Platform-simulated stop (your computer)	No	Not until triggered	Not until triggered	No	Exchange liquidity, after trigger and routing
CFD dealer stop (dealer's system)	No	The dealer holds it and is your counterparty	No exchange involved	Yes	The dealer's own price
Prop firm simulated account, such as Topstep's XFA (firm or platform simulator)	No	The firm or platform does	Never	Yes	A simulated fill against market data

Only whoever holds the order can see it. On exchange futures that is the exchange or your broker, never another trader; the cases of deliberate stop triggering on record involve dealers holding their own clients' orders.

How does a stop cascade push price through a round number?

A stop cascade starts when one ordinary trade prints at a cluster's trigger price: the triggered stops become sell orders that eat through the bids, and their fills print lower prices that trigger the next cluster. Osler's price-cascade study (a 2002 New York Fed staff report, published in 2005 in the Journal of International Money and Finance) found currency trends unusually rapid when rates reached levels where stops clustered, and concluded: "Together, these results indicate that stop-loss orders propagate trends and are sometimes triggered in waves, contributing to price cascades." Fett and McPhail (2017) found the same pressure in CME futures, where stop orders executed more often on volatile days and supplied significant liquidity in the direction of the move.

The cascade table walks one sequence through an illustrative thin ES bid book. Inputs, all illustrative: last trade 5000.00; bids of 8 contracts at 5000.00, 32 at 4999.75, 20 at 4999.50, 15 at 4999.25, 25 at 4999.00, 20 at 4998.75, 40 at 4998.50 and 60 at 4998.25; hidden sell stops of 100 contracts at 4999.75, one tick under the round number, and 40 at 4999.00. Each triggered stop sells into the best remaining bid. Slippage in ticks = (trigger price minus fill price) / 0.25, and one ES tick is worth $12.50 ($50 per index point times 0.25).

Step	What happens	Fills (contracts @ price)	Slippage (ticks)	Cost on ES	Last trade
Trigger	A 10-lot market sell	8 @ 5000.00, 2 @ 4999.75	Not a stop	n/a	4999.75
Wave 1	100 sell stops at 4999.75 trigger	30 @ 4999.75, 20 @ 4999.50, 15 @ 4999.25, 25 @ 4999.00, 10 @ 4998.75	165 (average 1.65, average fill 4999.3375)	$2,062.50	4998.75
Wave 2	The 4999.00 print triggers 40 more	10 @ 4998.75, 30 @ 4998.50	70 (average 1.75, average fill 4998.5625)	$875.00	4998.50
Total	150 contracts sold, 140 of them stops	Price 5000.00 to 4998.50	235	$2,937.50	6 ticks lower

Nobody hunted anything here: a 10-lot sell reached a level where 100 stops waited, and their own market orders did the rest. Thin books turn crowded stops into self-fulfilling moves.

In the illustrative cascade, a 10-lot sell released 140 contracts of stop selling, 150 contracts in all or 15 times the initiating order, and moved price 6 ticks through the round number. On Micro E-mini (MES) contracts, at $1.25 per tick, the same 235 ticks cost $293.75, and each tick is a better price for the buyer on the other side (see whether futures trading is zero-sum). Real sequences are messier, because CME triggers each stop as its price prints (wave 2 would fire in the middle of wave 1) and new orders keep arriving during the sweep; futures slippage explained shows how the cost appears in your own fills.

A stop cascade needs no one to see your stop, only a guessable level and a thin book beneath it.

Stop-limit or stop-market: which protects you better in a cascade?

Neither is strictly better: a stop-limit order caps how far below the trigger you can be filled but may leave part of the position open, while a stop-market order (on Globex, a stop with protection) fills at whatever the book offers down to the exchange's protection limit, where any remainder rests as a limit order (futures order types explained covers both). Fett and McPhail (2017) found 98 percent of executed ES stop orders were stop-limit (99 percent in WTI crude oil, 95 percent in 10-Year Treasury Note futures), citing the risk that a triggered stop-market order fills far below its trigger. The paper reported the ES protection range as 3 index points in 2014 to 2016, or 12 ticks and $150 per contract; the current range may differ.

The same 100-contract wave 1 was rerun two ways: as a stop-limit with its limit at 4999.25 (two ticks below the trigger), and as a stop-market into a thick book with 300 contracts bid at 4999.75.

Wave 1 variant (100 contracts, trigger 4999.75)	Filled	Left open	Slippage (ticks)	Cost on ES	Last trade
Stop-market, thin book	100	0	165	$2,062.50	4998.75
Stop-limit, limit 4999.25, thin book	65	35	50	$625.00	4999.25
Stop-market, thick book (300 bid at 4999.75)	100	0	0	$0.00	4999.75

The stop-limit cut the wave's cost from 165 ticks to 50 and capped the worst fill at 2 ticks, but 35 contracts stayed long behind a resting sell limit at 4999.25, and if price keeps falling they can lose more than the slippage saved. In this simplified path nobody sells at 4999.00, so wave 2 never fires. The thick book isolates the variable that matters most: with 300 contracts bid at 4999.75, the identical wave fills at its trigger with zero slippage.

A stop-limit caps the price you accept and pays for it with the risk of staying in the trade. In a cascade that trade-off is real, which is why CME's native stop with protection exists at all.

Why does price hit my stop and then reverse?

When price spikes through a round number and snaps back, the order data offer an explanation: stops beyond the level add a short burst of one-way orders, and once the burst is spent, the take-profit orders and buyers waiting at the round number remain. In an illustrative continuation of the cascade book, a sell stop at 4999.75 fills during the cascade, price bottoms at 4998.50 and then trades back to 5000.25, a rebound of 1.75 points, or 7 ticks, after the stop has sold.

This is the shape traders call a hunt. It is what a stop cluster plus a thin book produces with or without anyone intending it, and a stop placed where the idea fails, rather than one tick past the obvious line, usually sits outside it.

ICT and smart money traders call a spike through a level and back a liquidity sweep; the evidence on ICT and smart money concepts covers whether the label predicts anything.

Sometimes a stop is hit because it sat inside ordinary price noise, not because anyone hunted it. Fett and McPhail (2017) measured a daily average resting time of about 20 minutes for executed ES stop-loss orders, with 11.3 percent resting less than a minute, and read short rests as a sign of stops placed close to the top of the book or refreshed often. A stop one tick past 5000.00 sits $12.50 per ES contract from the most obvious level on the chart, where both the crowd and routine fluctuation arrive first.

Do prop firms or CFD brokers hunt your stops?

A stop in a prop firm's simulated account cannot be hunted in the market, because the simulator fills it against live market data instead of sending it to the exchange; a CFD dealer is a different case, because the dealer holds your stop and is your counterparty. Topstep's help center, for one, describes its Express Funded Account as the simulated funded-level account a trader earns after passing the Trading Combine (page updated August 2026). The market can still trade through your level, and each simulator models stop fills its own way, one reason sim results don't match live. In a live-funded account, the exchange rules above apply.

CFDs (contracts for difference, bilateral contracts that track a market's price) sit outside the exchange. Germany's financial regulator BaFin (2017) set out the structural problem: "In all CFD transactions, a CFD provider acts as counterparty to the client in the chain of execution, which means that the client's losses correspond to the CFD provider's profits. This constitutes a major conflict of interest for the CFD provider in question." BaFin saw a possible exception only where a provider hedges every client order, and noted that providers have extensive discretion over prices once markets turn turbulent.

Deliberate triggering of client stops is documented where a dealer holds the orders. When the UK Financial Conduct Authority fined five banks £1.1 billion in 2014 over G10 spot FX failings between 2008 and 2013, it said traders at different banks had shared information about client activity and "attempted to manipulate fix rates and trigger client 'stop loss' orders". Those were bank FX desks holding client orders, not futures brokers or CFD providers; the case shows what holding the orders makes possible, which is why who holds your stop matters more than who can guess its level.

Where should you put a stop loss so it does not get hunted?

Put the stop where the trade idea is proven wrong, then check that it is not sitting one tick past the round number or swing point everyone else is watching.

Size the stop from volatility and structure, not the nearest round number (ATR-based stop placement gives a method). If the invalidation point is a round number, give it room beyond the likely cluster and cut position size so the wider stop risks the same dollars.
Choose in advance which failure you prefer, a stop-limit that may leave you in or a stop-market that may slip, and know whether your stop lives at the exchange or on a PC that can disconnect.
Avoid thin windows, when the DOM shows small size at each level (often around releases on the economic calendar), because the same burst of stops travels further.

Thor, this blog's own product, copies a master account's trades to follower accounts, so a stop one tick past 5000.00 on the master becomes the exit on every follower; across live accounts those copies add size to the same exit at nearly the same moment, while in simulated prop accounts they have no market effect. A trade copier changes neither who sees a stop nor where it belongs: it multiplies the placement decision, so fix the stop on the master first.

Go deeper

Frequently asked questions

Is stop hunting illegal?

Trading through a price where other people's stops happen to rest is legal; trading to create an artificial price is manipulation, which the US Commodity Exchange Act prohibits. Placing orders you intend to cancel before execution, known as spoofing, became a specific violation of that Act under the Dodd-Frank Act of 2010.

Is a liquidity sweep the same thing as a stop hunt?

They describe the same price event with different assumptions: a sweep says price ran through a level where orders rested, while a hunt says someone aimed it there. A chart cannot tell the two apart, because a sweep driven by one large trader and one driven by a crowd of small stops print identically.

Should I use a mental stop so nobody can hunt it?

Not for secrecy on CME futures, because an exchange-held stop is already invisible to other traders. A mental stop (an exit you plan to place by hand) adds the risk that you hesitate, lose connection or are away, and if you exit by hand during a cascade you sell into the same thin book later than the stops did.

Why do my ES stops get hit overnight?

Overnight hours usually carry less depth at each price level, so a smaller burst of orders moves price further. The CFTC's 2017 study also found that right after the close the share of new orders that are stops spikes, which the authors attribute to traders protecting positions they hold overnight.

Does a trailing stop avoid stop hunts?

No. A trailing stop is a stop whose trigger follows price at a set distance, so at any moment it can sit one tick past a round number or swing point like any other stop. It changes where the trigger is over time, not who can see it.

What percentage of stop losses get hunted?

No study has measured it, and the percentages that circulate online have no traceable source. The closest measured data describe how stops are used (share of volume, resting time, order type), not how often anyone targets them.

Do DOM or order flow tools show where stops are?

No. Any indicator that labels stop levels on a CME chart is estimating them from price history, because CME market data carries no untriggered stop orders. Heavy one-sided trading just after a level breaks can show that stops fired, but only after the fact.

Does price usually reverse after a stop cascade?

Not reliably. Osler's price-cascade study (2002, published 2005) found the price response to stop-loss orders was larger and lasted longer than the response to take-profit orders, with most results statistically significant over hours though not over days, so a cascade can start a trend rather than end one.

Why can't I send a plain stop-market order to CME Globex?

Globex accepts only protected versions: its futures order types are Limit, Market Order with Protection, Market-Limit, Stop-Limit and Stop Order with Protection. A platform's stop-market button therefore becomes a stop with protection at CME, or is held by the broker or platform and sent as a protected market order when triggered.