MarginPad
MarginPad Premium

The 4 AI indicators, explained

Four readouts built on data only MarginPad has. Each answers one question you should ask before opening a leveraged position, and each is explained here the way it should have been from the start: what it measures, what it draws, and what to do with the number.

Our dataOwn liquidation feed across nine exchanges, own leverage-ladder model, funding and crowd positioning
Self-gradedEvery arrow is scored 1.5x ATR target vs stop on the loaded chart; the hit-rate sits in the legend
Any symbol, any timeframeComputed in your browser from the loaded candles, so they work on every pair the chart can load
Premium$3.99 a month or Founder once, on /charts and the mobile charts

1 · FragilityCascade Radar

How fragile is the market right now?

A liquidation cascade is a chain: one forced close pushes price into the next stop, which forces the next close. Whether that chain can start depends on two things: how much leveraged position sits close to the current price, and how much volume the market has been absorbing. Cascade Radar measures both.

What it measures

At every close it opens a ladder of hypothetical positions at 5x, 10x, 25x, 50x and 100x, weighted by that bar's traded value, and places their liquidation prices above and below. Price trading through a level sweeps it. What is left is a live map of liquidation fuel; the radar sums the fuel within 5% of price (closer counts more), divides it by the average traded value of the last 20 bars, and ranks that against the last 200 bars. The result is a 0-100 score: 70 is fragile, 90 means one push can start the chain.

What you see

A 0-100 panel under the candles with a line at 90. The legend prints the score, a word for it (calm / fragile / EXTREME) and, since September 2026, which side the fuel sits on. A CASC arrow fires only when the score is 90 or more and the fuel is clearly one-sided: red above the bar when it sits below price, green below the bar when it sits above.

90 Cascade 93/100 fuel BELOW price (long liqs) CASC
Extreme and one-sided: the fuel is under price, so the arrow points down. A long here has a chain of liquidations beneath it.

In plain words, before you click

  • The line: under 50 = calm, nothing to do. Rising past 70 = fragile, one push can start a chain of forced closes. Past 90 = extreme.
  • Before a long: line above 70 and the legend says fuel below = you would be buying on top of a trapdoor. Wait for the flush, or take half size with a wider stop. Fuel above = a squeeze would run in your favour.
  • Before a short: the mirror. Fuel below is your target zone; fuel above and 70+ means you are the one standing on the trapdoor.
  • The arrow: red CASC over a candle = the chain is primed downwards; green CASC under a candle = primed upwards.
Example legend lineCascade 93/100 — EXTREME fragility · fuel BELOW price (long liqs): a dip can run
CASC 61% (11W/7L)

Limits. The ladder is a model of where positions would be liquidated, not a feed of resting orders. It says how fragile the market is, not when the push comes.

2 · TargetsLiquidation Magnet

Where is price being pulled next?

Price tends to visit big pools of liquidations before it reverses: the pool is where forced orders are waiting, and forced orders are what moves a market that nobody else wants to move. The Magnet finds the dominant pool on each side and says which one is pulling harder.

What it measures

The same leverage ladder as Cascade Radar, read as a list of pools: price level, size in dollars, and whether it is long liquidations (below price) or short liquidations (above). Each pool's pull is its size times its proximity; pools beyond 9% are ignored. The strongest pull above and the strongest below are the two magnets, and the larger of the two is the side price is likelier to visit first.

What you see

Two price lines on the candles: red for the short-liquidation pool above, green for the long-liquidation pool below, each labelled with its size. The bold line is the stronger magnet, the thin dashed one the weaker. The legend prints both pools with their distance from price.

SHORT $18.2M · +2.4% LONG $41.6M · −1.8% ← pulled DOWN price
The larger pool below is the bold magnet: the likelier next visit is 1.8% down. A long's stop belongs beyond it, not just above it.

In plain words, before you click

  • The lines: red above = where shorts get liquidated, green below = where longs get liquidated. The bold one is the stronger pull and the likelier next visit.
  • Before a long: bold line above you = your first target, take some profit there. Bold line below you = expect a dip to it first; put your stop under that line, not just above it, or yours is the order that gets hunted.
  • Before a short: the mirror. A bold green line below is the target; a bold red line above means the stop stays beyond it.
  • No lines within 9%: both sides were already flushed. Nothing is pulling; expect a slower market.
Example legend lineMagnet: pulled DOWN · above 76,410 (+2.4%, $18.2M short liqs) · below 73,220 (−1.8%, $41.6M long liqs)

Limits. A magnet is a probability, not a schedule. Big news moves price through both pools without stopping at either.

3 · AdaptiveMarket Brain

Which way does the evidence lean, and what is driving it?

Most "AI" indicators are one formula with a name. Market Brain is eight readings and a learning step: it measures which of the eight actually predicted the next six bars on this coin, on this timeframe, and weights them accordingly. A factor that does not work here weights itself to zero.

What it measures

The eight factors are trend (EMA 21 vs 50 in ATR units), momentum (MACD histogram and RSI), volume (this bar against the last 20, signed by direction), fragility (Cascade Radar's side and score), real liquidation flow (long vs short liquidations per bar from our collector, scaled by how unusual the bar was), funding (the current rate against its median, read contrarian), crowd positioning (long/short account ratio, read contrarian) and structure (where the close sits in the 55-bar range). Every 10 bars the model re-learns a weight per factor from the trailing 280 bars - the correlation between the factor and what price did over the next six bars - and never from bars it has not seen yet. The weighted blend is the score, -100 to +100.

What you see

A panel with the score and bands at +55 and -55. A BRAIN arrow fires the first time the score crosses a band and re-arms once it falls back inside 25. Arrows are held back within 2.5 hours of a high-impact macro event (FOMC, CPI) on purpose. The legend prints the score, a word for it, the two factors driving it right now, and the hit-rate chip.

+55 0 −55 BRAIN Brain +62 · strong bullish lean driven by: liq flow + funding
A crossing of +55, driven by forced short covering and funding paid by shorts: the crowd is on the wrong side and paying for it.

In plain words, before you click

  • The line: going up = the evidence leans bullish, going down = bearish. Inside the +55 / -55 bands the model has no strong opinion.
  • Before a long: you want the line above +55, or rising through it. Below -55 you are trading against everything the model learned here - do not.
  • Before a short: the mirror, below -55. Inside the bands: no edge, size down or trade the Magnet levels instead.
  • "Driven by": "funding + crowd" = the crowd is trapped and paying to stay in (a squeeze setup); "liq flow" = forced selling or covering is already happening; "trend + momentum" = plain continuation.
  • The chip: the percentage is how often its arrows worked on this exact chart. Under 50% = no edge on this coin and timeframe; ignore its arrows there, whatever the line does.
Example legend linesBrain +62 — strong bullish lean
BRAIN driven by: liq flow + funding
BRAIN 58% (14W/10L)

Limits. Learning from 280 bars means the weights lag a regime change by a few dozen bars. Funding and crowd factors need the exchange to publish them for the pair; where it does not, those two factors read zero and the model says so by not naming them.

4 · ForecastMarket Memory

When it looked like this before, what happened next?

Market Memory does not model the market. It looks it up. Using the same eight factors as Market Brain, it finds the past bars whose whole state most resembles right now and reports what price did after them. It is the market's own memory: a base rate, honestly labelled.

What it measures

Every bar is a point in eight-factor space. For the current bar it finds the 25 nearest past bars - only bars old enough to have a known six-bar outcome, so no answer is ever read from the future - and averages their forward move in ATR units. It also counts how many of the 25 went up. The forecast panel shows that average over time; the current one is projected onto the price axis.

What you see

A forecast panel, a dashed MEM line at the projected level (green above price, red below), and a legend line with the analog count, the up-rate and the projected move. A MEMO arrow fires only when the analogs agree: an expected move of at least 0.85 ATR with 62% or more of them on the same side.

MEM MEMO Memory: 25 analogs · 72% up proj +1.4% / 6 bars
Twenty-five look-alike moments, eighteen of which went up, averaging +1.4% over the next six bars. That is a target and a horizon, not a promise.

In plain words, before you click

  • The readout: "25 analogs · 72% up · proj +1.4%" = the last 25 times the market looked like this, 72% of them went up, by about 1.4% over the next six candles. A base rate, not a call.
  • Before a long: 65% up or more AND a projected move bigger than your fees and your stop = go, with the dashed MEM line as a realistic target and six candles as the horizon (six hours on a 1h chart).
  • Before a short: the mirror, 35% up or less.
  • 45-55%: history went both ways from here - a coin flip. Stay out, or trade the Magnet levels instead of a direction. 72% up with a +0.3% move is drift, not a breakout: read the size of the move, not only the percentage.
  • The arrow: MEMO fires only when the analogs agree; it is the strongest form of this readout.
Example legend lineMemory: 25 analogs · 72% up · proj +1.40% / 6 bars

Limits. The memory is only as long as the loaded history (a few hundred to a few thousand bars). On a coin that listed last month it has little to remember, and it needs 160 bars before it draws anything.

The hit-rate chip: every arrow grades itself

Any indicator can draw arrows on the past. MarginPad's arrow indicators grade every arrow they drew on the loaded chart: entry at the signal bar's close, target and stop both 1.5x the 14-bar ATR away, scanned up to 80 bars forward, and a bar that touches both counts as a loss. The chip appears once five arrows have resolved and is coloured for what it says:

CASC 61% (11W/7L) an edge on this chart   BRAIN 50% (6W/6L) a coin flip   MEMO 38% (5W/8L) no edge here - do not use it

The chip is per symbol and per timeframe, because that is where an edge lives or does not. An indicator that scores 61% on BTC 1h and 38% on a thin altcoin's 5m chart is telling you exactly where to use it.

Which one to open first

The four are meant to be read in order, each narrowing what the last one said.

1Cascade RadarIs the market fragile, and on which side? Decides your size and how wide the stop must be.
2Liquidation MagnetWhere are the pools? Sets the first target and tells you where a stop must not be.
3Market BrainWhich way does the learned evidence lean, and why? Decides the direction, or tells you there is none.
4Market MemoryWhat happened the last 25 times? Sets a realistic target and horizon, or warns that history went both ways.

Questions

Are these buy or sell signals?

No. Each one answers a question you should ask before a position. Arrows appear only at extremes, and every arrow indicator prints its own hit-rate on the loaded chart so you can see whether it has an edge on that symbol and timeframe.

What data are they built on?

MarginPad's own liquidation feed (nine exchanges, collected in-house), the leverage-ladder model behind the liquidation heatmap, funding history and crowd long/short positioning, plus the chart's own price and volume. None of it comes from an indicator vendor.

Why only four?

There were eight until September 2026. Four were generic technical analysis available on any chart (a MACD+RSI momentum cross, a Bollinger squeeze) or overlapped with the four that remain. The four that stay are the ones that use data no other chart has, and each now explains what its reading means for a position.

Do I need Premium?

Yes. The four are part of MarginPad Premium ($3.99 a month, or a one-time Founder payment) on the multi-chart workspace at /charts and on the mobile charts. Every other indicator on the chart is free.

Where do I turn them on?

On /charts, open a chart window's Indicators menu: the four sit at the top in lime. On a phone, the chart's indicator sheet lists the same four. The readouts appear in the legend at the top-left of the chart.

Try them on a live chart

Premium is $3.99 a month, or once as a Founder. Everything else on MarginPad stays free.