Every 5 minutes this site re-scans the live OSRS price feed, ranks the top 90 items by the profit you could realistically capture inside one 4-hour buy-limit window, and subtracts the real 2% GE tax before it calls anything a margin. No account, no paywall. This page is the whole method — including, just as importantly, what it can't do.
Start flipping free — live board ↗ Read the guides
Selling on the GE costs a 2% tax (since 29 May 2025), but with two rules people forget: it's capped at 5,000,000 gp per item, and items selling for 49 gp or less are exempt, as are a handful of items like bonds. Every margin here is computed after that real tax — so cheap items and bonds aren't wrongly hidden, and ultra-expensive items aren't over-taxed.
Concretely, the tax the site applies is min(price × 0.02, 5,000,000), zero when the price
is under 50 gp or the item is on the exempt
list — 17 names, mostly tools (hammer, saw, chisel, spade, secateurs, pestle and mortar…) plus the
Old School bond. Two consequences worth knowing: the 5M cap only starts biting at a sell price of
250,000,000 gp, so below that you are always paying a flat 2%; and because the tax is floored to a
whole gp, sub-50 gp items are genuinely tax-free rather than merely rounding to zero. The same tax math
runs on every board here — decanting, recipes and the
position tracker all net it out. Full tax
guide →
Instead of a single average, the buy/sell targets are percentiles of recent prices, weighted toward the last few hours (older data counts for less). That keeps the targets tracking the market you're actually trading in right now, not last week's.
The defaults are the 12th percentile of recent lows to buy at and the 93rd percentile of
recent highs to sell at, computed over the 5-minute price series. Each observation is weighted by
exp(−hours_ago / 8), so a print from 8 hours ago counts for about a third of one from the
last few minutes, and a day-old print counts for roughly 5%. Those percentiles are deliberately less
aggressive than the ones flipping guides usually quote (10th/97th): a 97th-percentile ask looks better on
paper but fills far less often, and a target that never fills is worth nothing. An item needs at least
6 weighted highs and 6 weighted lows in the window to get a band at all; below that it falls back
to live prices and is scored at half weight so it can't top the board on thin data.
Those percentiles start as sensible defaults. A backtest then replays each item's real price history and asks: which buy/sell percentiles would have actually made the most money? Crucially it uses no lookahead — at every step the bands are built only from past data, never the future (that's the #1 way a backtest lies to you). A flip only counts when a buy would have filled and then a sell would have filled, profited at real prices after tax. Items whose learned bands hold up (enough real, profitable flips) get a green ★ badge and use those bands; everything else falls back to safe defaults. Treat the learned numbers as a strong starting point, not a promise.
The mechanics, precisely: for each item the backtest sweeps 36 combinations — six buy percentiles (5/8/12/16/20/25) against six sell percentiles (75/80/85/90/93/97) — over its hourly price history, using a 48-hour trailing window. At every hour the bands are rebuilt from the previous 48 hours only and then tested against the next unseen hour, which is what "no lookahead" means in code rather than in marketing. A buy fills when the hour's average low touches the buy band, the sell fills when a later hour's average high touches the sell band, and the flip is only banked if it cleared a profit after the real tax. An item needs at least 60 hours of history to be testable, and it only keeps its learned bands — and the ★ — if that simulation completed 3 or more profitable flips with positive profit per day. Everything else stays on the 12/93 defaults. The whole sweep re-runs every 6 hours, so the bands drift with the market instead of being frozen at deploy time. Per-item backtest heatmaps are on each item detail page and in the ▣ terminal.
Rows are not sorted by margin, and not by ROI. Sorting by either one puts a dead item with one lucky print at the top every time. The board sorts by a score: the profit the item could realistically return over a 4-hour window, multiplied by a quality factor built from how evenly the opportunity recurs and how volatile the price is. The quality factor runs from 1.0 for an item whose buy and sell windows appear steadily around the clock at a calm price, down towards 0.5 and below for one that only worked during a single spike. In practice a 10% price volatility divides an item's score by about 1.3, and an opportunity that recurs in only one hour of the day loses roughly half its score outright. The effect is that a steady, boring, liquid item will outrank a flashier one with double the headline margin — which is the ranking you actually want. Why volume beats margin →
Profit is margin × how many you could realistically buy and sell in a 4-hour buy-limit window (from real traded volume) — not margin × a buy limit you'll never fill. RECUR shows how evenly an opportunity recurs across the day (steadier = more trustworthy), and VOL% shows recent price volatility (shorter bar = calmer).
The realistic quantity is deliberately pessimistic. The site takes the smaller of the item's last hour of volume scaled to four hours and its 24-hour volume, then assumes you can capture just 18% of it — because you are not the only person trading that item — and finally caps the result at the item's actual GE buy limit. Items trading fewer than 12 units an hour, or priced under 30 gp, are dropped before scoring at all. RECUR is measured properly rather than eyeballed: the day is split into 24 hourly buckets, and the site measures how evenly the buy-window and sell-window opportunities spread across them, so an item that only becomes flippable during one busy evening hour scores low even if that hour was very profitable. How buy limits work →
A fat margin can still be a bad flip if it ties up your whole bank or you can't sell out. So every row shows CAPITAL (the gp you actually deploy at a realistic size), EXIT (roughly how many hours it'd take to offload that position at the item's average volume — lower is safer), and a RISK rating that blends exit-time with price volatility. Use the filter boxes up top as your own cutoff: cap the capital you're willing to risk, set a minimum ROI and minimum volume, or hide anything that'd take too long to exit. The header also sums the capital to work the whole board and the theoretical ceiling (margin × full buy limit across every item) — handy, but remember the ceiling assumes you fill every limit, which crowding and volume rarely allow.
The RISK rating is not a vibe, it's two thresholds. An item is low risk only if it would take 2 hours or less to sell out and its price volatility is under 5%. It's high risk if either the exit takes 8 hours or more or volatility reaches 12%. Everything between is medium. Exit time itself is just the position size divided by the item's hourly volume, so it answers the only question that matters when a flip goes wrong: how long am I stuck holding this? The TREND arrow compares the most recent third of the price series against the oldest third and marks the item up or down past ±1.5%, with a ⚠ crash flag once a fall reaches 3% — a description of what already happened, never a prediction of what comes next.
Being clear about the gaps is more useful than pretending there aren't any. It does not predict prices — every number describes the market that already traded. It does not model you: the fill simulation assumes your offer sits at the band and gets hit, when in reality you are queuing behind other players at the same price, which is the crowding problem above. It does not know about updates, so a fresh Jagex announcement will move an item long before the trailing window notices. It has no account, no trade history and no way to place an order, by design. And it does not chase the thinnest possible spread — the percentiles, the 18% volume capture and the volume floors all trade headline profit for the likelihood that a flip actually completes. If a row looks impossibly good, the usual answer is that the item is illiquid and the margin is an artefact of two stale prints. Is flipping actually worth it? →
Live prices come from the excellent free OSRS
Wiki real-time prices API. To be a good citizen, this site polls it on a schedule from a single
server and serves everyone the cached result — so a thousand visitors don't become a thousand hits on a
free community API. There's a raw /api/flips JSON feed if you want to build your own thing.
The polling is on a 5-minute cycle, matching the cadence of the upstream 5-minute feed — asking
more often would return the same numbers. Requests identify this site by name in their user agent, retry
three times, and back off when the API asks them to, so a bad night here never becomes someone else's
outage. The item-name mapping is cached to disk and refreshed at most once a day; if the upstream API is
unreachable the site serves its last good snapshot rather than an empty board. Everything the scanner
produces is public: /api/flips for the board, /export.csv for the same data as
a spreadsheet, and per-board endpoints behind alch, movers and
recipes. No key, no rate limit, no sign-up.
This tool only shows you the data; you place every trade by hand. Automating the Grand Exchange — bots, auto-clickers, scripted buying/selling — breaks Jagex's rules and gets accounts permanently banned. "Providing liquidity with a bot farm" sounds clever but it's just botting, and it's not worth your account. The honest version of that idea is what a human flipper already does: buy the bid, sell the ask, and the volume/risk signals here help you do it sensibly.
If you've read this far you want the reasoning, not just the board. These go deeper on the parts people get wrong most often:
Yes — no account, no paywall and no download. The flip board, the F2P board, the
CSV export and the raw /api/flips JSON feed are all open to everyone. Ads cover the hosting.
The scanner re-runs against the OSRS Wiki real-time feed every 5 minutes, and board pages refresh themselves every 120 seconds. The backtest that tunes each item's price bands re-runs every 6 hours.
That item's buy/sell percentiles were learned from its own price history instead of the site defaults. An item only earns the badge once the backtest completes at least 3 profitable flips on it with positive profit per day; otherwise it falls back to the 12th/93rd percentile defaults.
After. Every margin, ROI and profit figure subtracts the 2% sale tax, including its 5,000,000 gp per-item cap and the exemption for items selling at 49 gp or less.
Usually less. The figures assume both your buy and your sell fill at the shown bands, and they ignore competition — popular flips crowd, which bids your buy up and pushes your sell down. Treat a row as a ranked starting point, not a forecast.
Flipping by hand is ordinary gameplay. Automating it is not: bots and auto-clickers on the Grand Exchange break Jagex's Code of Conduct and get accounts permanently banned. This site only displays data; you place every trade yourself.
No. It's an unofficial fan project. Prices come from the community-run OSRS Wiki real-time prices API.
Start flipping free — live board ↗ Read the guides