Skip to content

What Marginatrix cannot know

The limits, first — so every number that follows means something.

Most of this site explains how markets work. This page is about one product — this one — and it exists because the honest version of “here is where positions get forced out” is short, and the list of things it cannot see is longer.

The one rule that makes the rest readable

Rounded numbers are estimates. Exact numbers are measured.

That is the whole grammar. When a message says “about $3M, centred near $64,300”, two significant figures are doing the work of a warning label: a model produced it. When it says “$310,000 force-closed, across 18 positions”, that is a count of events that happened, taken from an exchange feed.

A range like $63,600–$63,900 would carry four significant figures twice and read as precision. You will not see one used for an estimate here.

What is measured

  • The price levels. They are measured out from the most recent session high or low. Every post names that price above its ladder, so each level can be found on the chart. How the spacing is computed is not published.
  • Recently force-closed positions. Counted from Bybit’s liquidation feed: the price each one executed at, and its size. Every alert and every map states the window it covers; this page does not restate it, because a static page cannot follow a producer that is tuned.
  • Session gaps. The previous session’s close, the next session’s open, and whether price has traded back through the space between.
  • The funding rate, and the share of accounts on each side. Both published by Bybit.

What is estimated

One thing, and it is the map of where positions would be closed if price reached them. It is built from how open interest changed, bucket by bucket, priced at the mark price in each bucket, and projected forward using the exchange’s own liquidation rules.

What no outside observer can see

None of this is a gap in our data collection. It is not visible to anyone who is not the exchange:

  • Anyone’s actual leverage. The published map assumes 50x, which is the most fragile end. At lower leverage the same position sits much further from the price — often several percent further — and is not in that number.
  • Isolated or cross margin, and whether collateral was added. A position one dollar from liquidation stops being one the moment its owner tops it up, and nothing announces that.
  • Whether a position is hedged. An account short here and long the same size elsewhere is flat, and looks identical to a directional bet from outside.
  • Entry prices. Open interest says how much is open, not where anyone got in.
  • Which risk-limit tier an account sits in. Large positions need more margin to stay open, which moves their liquidation price.
  • Fees and funding paid so far, which erode margin quietly between settlements.
  • Every other exchange. The liquidation counts come from Bybit only. Not because other venues do not matter — because their feeds report different quantities in different units, and mixing them silently produced numbers that were wrong by two orders of magnitude. They are excluded and the exclusion is printed.

The two limits of the method itself

Open interest is symmetric. Every futures contract has exactly one buyer and one seller. A bucket where open interest rises by $10M created $10M of new long exposure and $10M of new short exposure, at the same price, at the same moment. So there is no “which side is crowded” to read out of it, and any product that claims to derive one from open interest is inventing it. The map places equal size on both sides of every bucket for exactly this reason.

A cluster is not a wall. This is the part almost everyone has backwards. When a leveraged position runs out of margin, the exchange closes it by trading against it — so closing a position that loses money when price rises means buying. A cluster of them above the current price is not resistance. It is fuel. That is why these messages talk about speed rather than direction, and why they never say which way to trade.

What the product will do with all that — not yet published

Everything in this section is built and running, and going to nobody but us. It is in a shadow period: the messages are composed on live data every cycle and delivered to the operator’s own chat, so that a month of them can be read against what the market actually did before anyone is asked to act on one. Nothing below arrives in a channel or a direct message today, and this page will say so until it does.

We are telling you this because the alternative is worse. A page written in the present tense would make our silence look like a measurement — “we looked, and there was nothing unusual” — when the real reason you are not hearing it is that we are not sending it yet. That is a fabricated reading, and this page exists to remove those.

What you do get today is on the pricing page, in the channels and in the free app: two scheduled posts a day per instrument, an alert when the high or low moves or the spacing of the levels changes, and the live chart with the ladder. Pro subscribers also get a direct message when a session gap closes; no channel posts that one.

When the shadow period ends, this is the shape of it — one of a small number of things, and nothing the rest of the time:

  • Nothing — when there is no trailing distribution to call anything unusual against.
  • A cluster above, or below — an estimated cluster in the heaviest 10% of the last thirty days’ worth, within 2% of the price.
  • Clusters both sides.
  • A cascade in progress — force-closed volume in the last fifteen minutes above the 99th percentile of the trailing month. This is the only one built on realized data, and it is the one place where backward-looking data is the right data: a cascade happening now is observed, not predicted.

None of these is a trade signal, and none of them carries a direction. They are statements about how fast price is likely to move if it gets somewhere, which is a different question from where it is going.