Eight short lessons. Every chart on this page was mined from real Coinbase candles for the 20 coins the tracker follows (hourly, Apr–Oct 2026; 15-minute, Oct 2026) — the coin and dates are printed under each one, and the overlays are drawn on the actual prices. Where a pattern has no clean real example, the lesson says so instead of faking one. Nothing here is a promise: the system's own backtests and live record are quoted as they are, including the parts that don't work.
A perpetual contract (a "perp") is a bet with a clock on it. You pick a crypto, you pick a line — a price — and you pick a cycle: how long until the bet settles (5 minutes, 30 minutes, an hour…). When the cycle closes, only one question matters:
Did the price close OVER your line, or UNDER it?
On real exchanges, perps usually track the price itself (you profit from the size of the move, not just the side) — Lesson 8 shows that version with real numbers. This tracker works in the over/under world: it watches price against your line and estimates the chance of each side, live, as the cycle runs down.
Every minute, the tracker sorts each coin into one of four states, using the same maths on every chart below: how far price travelled in the last day compared with how much that coin usually moves (its volatility, σ), whether the move is easing, and how often price crossed its own average. The live tracker runs this on 15-minute, 1-hour and 4-hour windows; the examples here use the 24-hour window with the identical formulas, so the shapes are big enough to see. All four charts are real.
Price travelled +10.8% in 24 hours — about 2.0× the move this coin typically makes in a day (a ≥0.5σ standardised slope, still accelerating, is what the tracker calls jumping). What to look for: higher highs and higher lows, pullbacks that get bought quickly, most hours closing green.
-4.8% in 24 hours (2.0× the typical daily move, still accelerating down). Lower highs, lower lows, bounces that get sold. When a coin is dropping, OVER needs the fall to stop and reverse before your cycle closes — the trend is not your friend.
Price crossed its own 24-hour average 14 times in a day and finished just +0.3% from where it started. On the tracker this shows as RIDING THE AVERAGE — or RIDING THE LINE when price is hugging your line instead. Riding is the hardest state to bet on: tiny wiggles decide the close, so treat any call here as low-confidence unless the cycle is very short.
The whole day's range was 0.53% — a fifth of the 2.7% range this coin usually covers in a day. No slope, no energy. Lesson 6 is devoted to this state's cousin, the plateau, because it quietly eats more beginners' money than any crash.
Trends rarely move in straight lines. They zig-zag: a push one way (a leg), a pullback, then the next push. String legs together and you get a wave. The cyan line below connects the real turning points (pivots) of a real ten days — a leg counts when price reverses by at least 1.2% (the threshold used for this chart; the tracker draws its live swing stats with a finer one).
Across this wave, legs averaged ±4.4% over about 15 hours each. The last full leg was -4.8%, so the pattern's call at the end of the window was: next leg up, roughly 4.4% over ~ 15 hours. And this time the wave broke: over the next 48 hours price moved -6.0% — the opposite of the pattern's call. That is the break signal in real life: instead of turning at the pivot, price kept pushing through it. When a leg travels much further than the typical leg size (~4.4% here) without turning, the wave count is dead — do not keep betting on the turn.
Most of the time, sudden moves are luck. But one pattern in our own backtests genuinely beat the base rate, and it is the only one the tracker flags: the squeeze → shock. First, volatility collapses — hours of unusually quiet bars (the squeeze). Under the quiet, one-sided pressure builds: price keeps leaning one way and volume creeps up (the spring loading). Then the coil releases in one fast move (the shock).
The fitted rules (pooled across all 20 coins, hourly bars): squeeze = the last 6 hours' volatility in the bottom 20% of that coin's own history; pressure = the 6-hour move at least 1.0σ one way; shock = a bar beyond 2.5× its typical move on a volume spike. Judged over the following 4 hours, flagged setups hit their big move 23.1% of the time versus a 14.4% base rate — a 1.61× lift. Honest footnote: the same flags did not beat the base rate at the 1-hour horizon or on 15-minute bars. A snipe flag means "elevated chance of a sudden move", never "a move is coming".
What to look for before it fires: bars getting visibly smaller while price refuses to reverse; volume rising into the quiet; a price that keeps closing near the same edge of its little range. What kills the setup: the quiet ends with a drift back to neutral — pressure gone, no trade.
A spike (or its evil twin, a drop) is one violent bar: here, a 15-minute move of at least 4× the coin's typical bar, on at least double the usual volume. Two real ones, chosen because they teach opposite lessons about warning signs.
That bar moved +2.3% in fifteen minutes — a 7.1σ event on 8.7σ volume. Before it fired, the 2-hour volatility was already running at the 78th percentile of its recent history and volume across the approach was trending 3.1× upward. Translation: the market was winding up in plain sight. Spikes like this are what the squeeze lesson (4) hunts in advance.
-1.7% in fifteen minutes, a 7.6σ bar. And before it? Volatility at the 2nd percentile — nearly the quietest stretch in its recent history — and volume trending 0.8×, i.e. flat. No squeeze, no build-up, no warning in the price or volume data this system tracks. Some drops are news, liquidations, or one huge seller, and no chart pattern sees them coming. Anyone who tells you otherwise is selling something.
The tradable difference: you cannot predict the second kind. You can only size for it — never stake so much on one cycle that a 5–8σ bar against you ends your day.
A plateau is a stretch where price goes nowhere, slowly. No trend pays you, no wave completes, and every over/under close is decided by noise — while fees, spread and your own impatience take a cut of every attempt. It does not feel dangerous. That is exactly why it is.
Three full days of BCH: total range 2.0%, net move 0.01%. A normal 72 hours for this coin covers about 8.5% — this stretch managed a quarter of that. Every line you could have set inside that band would have been crossed dozens of times by pure jitter.
Everything in these lessons is wired into the tracker itself. What you are looking at on a coin card:
Getting the direction right is half the trade. The other half is what you do while the position is open. On an exchange, a perpetual position is not a lottery ticket you hold to the end — you can add to it, trim it, close it early, or reverse it, and each choice reprices your risk instantly. This worked example uses a real move and real prices throughout; stakes are a flat $1,000 per step so the arithmetic is easy to audit. Fees and funding are ignored (they would shave every number slightly). The tracker advises; it does not place trades — these buttons exist on your exchange, not here.
| Step | PEPE price | What happened |
|---|---|---|
| 1 · ENTER 2026-07-02 15:00:00Z | 0.00000244 | First leg of the wave turns up — enter LONG (over) with $1,000. |
| 2 · ADD 2026-07-03 22:00:00Z | 0.00000275 | Price pulled back and held above the entry — the trend confirmed. Add a second $1,000 LONG at the pullback low. |
| 3 · CASH OUT (half) 2026-07-04 04:00:00Z | 0.00000284 | Peak of the second leg. Sell HALF the position, bank the profit, let the rest ride for free. |
| 4 · FLIP 2026-07-04 06:00:00Z | 0.00000273 | Price closes back below the pullback low — the setup is broken. Close the remaining long and open a $1,000 SHORT (under). |
| end 2026-07-04 10:00:00Z | 0.00000268 | The reversal leg ends. Cover the short. Trade over. |
$1,000 long at 0.00000244. The first leg ran +15.6% before the pullback — at the add point the position was already up +$127.05.
The dip was -2.1% and held — so the second $1,000 went in at 0.00000275. Average cost 0.00000259. Adding after confirmation, not before, is the whole trick.
At the peak, 0.00000284, selling half banked +$98.33 of locked profit. Selling everything there was worth +$196.66 in total — more, but only hindsight knows it was the peak. Half is the compromise: paid, and still in the game.
The close below the pullback low killed the long story. Closing the rest banked +$55.79 more; the $1,000 short then earned +$18.32 as price slid to 0.00000268. Doing nothing at the flip and holding the long to the end would have finished at +$72.91 instead of +$172.44.
| Strategy on the same real prices | Result |
|---|---|
| The worked plan (enter → add → half out → flip) | +$172.44 |
| Buy $2,000 at the entry and just hold | +$72.91 |
| Sell everything at the exact peak (hindsight — nobody can do this live) | +$196.66 |
Read the gaps honestly: the plan beat holding because of the flip, not the entry — the entry alone was worth little once the trend broke. And the hindsight row beats the plan, which is the standing tax on every real decision: you act with the information at the time, not the information in the chart afterwards. Fees, funding and slippage would narrow every gap above; a faster reversal would have punished the flip. Position craft is risk management, not a money printer.