Big orders in the book as a reference point

The idea: a big order in the book is a level price will bounce off, or a level it will aim for. When we tested it, price reached such a wall no more often than it reached a randomly picked level at the same distance.

Why it sounded right

A wall in the order book is the most tangible object on the whole market. It is not a line drawn through two touches, and not an indicator with settings. It is real money, placed by a specific participant, at a specific price. You can see it. You can measure it. It seems obvious that price cannot get through a big order without eating all of it, so there will either be a pause or a reversal.

Two ways to trade grow out of that, and both look sound. First: trade off the wall. A wall above means sell into it, because price has nothing left to move on. Second: trade the break. If the wall did get eaten, the buyer is serious, and after that clear-out the road is open. Both rest on the same thought: a big order is an obstacle, and an obstacle changes the move.

It gets extra credibility from the fact that in the moment it often looks true. Price walks up to the wall, slows, bounces — and there is your story. On top of that, a wall gives you something no chart level gives: an exact size. Not "there is resistance here" but "there is this much money here". Once a level has a size, it feels like you can weigh it against the strength of the move and decide whether it gets pushed through.

How we tested it

We collected a long feed of alerts about big orders appearing, over several weeks: ticker, wall price, its size, which side it sat on. For each wall we pulled the minute-by-minute price history and looked at three things: does price reach the level, does it break through, and what happens after the touch.

The main control is a random level. Take the same coin, the same moment, the same distance from the current price, but pick the level arbitrarily — no wall there. If the wall is an obstacle, price should reach it less often and break it with more difficulty. The second control is "what if we take a different coin": same moment in time, same move, but the ticker next door. The third one turned out to be decisive: "same coin, same move over the previous hour, but no wall". It answers the question of what we are even measuring — the wall, or the move that brought price to it.

How it ended

As a level, the wall does not work at all. Price reaches it almost exactly as often as it reaches a random level at the same distance, and breaks it just as often too. There is no difference. Worse, after touching a real wall price goes further past the level than past a random one — it gets pushed through more willingly. So trading a bounce off the wall is worse than trading a bounce off an arbitrary price. Trading the break is empty too: collapse repeat signals on the same coin and the average result is around zero, with all the profit resting on a couple of very lucky cases.

One weak spot did survive: entering off the wall for a short hold gives a small plus, and it holds up both when you trim the extreme values and under both swap tests — different coin, different moment. But the decisive control explained that too. The alert about a wall arrives after price has already been driven toward it. The whole effect sits in the subset where price was moving toward the wall, and disappears completely where price was moving away from it. And if you take the same coin with the same move but no wall at all, what is left is a tiny bit extra — less than the cost of getting in and out.

The conclusion is simple: a wall is not a level, it is a trace of the recent move. A snapshot of the book describes what has already happened, which is why it is no good as a guide to what comes next.

What this teaches

• Being visible is not the same as being predictive. You can see the order; that does not mean it holds price.
• A snapshot of the book depends on what price just did. Any conclusion drawn from it has to be checked against the control "same move, object absent".
• The test "what about a random level at the same distance" is cheap, and it kills almost every level-based idea.
• If a rule only survives in the subset where price was moving toward the object, you are measuring the move, not the object.
• A small plus that is smaller than the cost of trading is not a small finding. It is a zero.