New to the terms? The Crypto Glossary covers leverage, longs, shorts and liquidations. Want to see how the calls actually score? Every call is on the public record, misses included.
Order flow tells you who was behind a move, not just which way price went. By pairing net buying and selling with the change in open interest, you can tell a week of fresh short positions, which builds fuel for a squeeze, from a week of leveraged longs being flushed out, which does not. On a price chart those two weeks can look identical.
It Started in a Spreadsheet
Every serious market read starts the same way. You are staring at a chart and you have a feeling. The feeling might be right. It might just be the last three green candles talking. Having a view was never the hard part. The hard part was having one I could trust enough to act on, week after week, without my mood doing the deciding.
A couple of years ago there was no engine and no dashboard. There was a Coin Metrics spreadsheet and a lot of manual work. Every week I logged the same handful of things about Bitcoin by hand: how the weekly candle closed, whether buyers or sellers had been more aggressive, how heavy the volume was, where price sat in its recent range, what the mood of the crowd looked like.
None of those numbers tells you much alone. A big red candle can be the start of a crash or the shakeout before a rally. Heavy volume can mean panic or conviction. Quiet volume can mean nobody cares, or that something is about to give. The reason for lining them up was to stop treating any one as the answer and start asking a better question: when do several of these agree at once?
Why One Reading Is Never Enough
Every market signal works brilliantly right up until it doesn't. Sentiment is a superb contrarian tool at genuine extremes and useless in the messy middle. Volume confirms a real move and lies during a quiet chop. Candle patterns catch reversals on the weekly and generate endless noise below it.
The tools are not wrong. The problem is that no single one of them knows what kind of week it is looking at. So instead of betting on one, the engine scores several independent readings and looks at them together. When they disagree, that tells me something too. It tells me to wait.
The Four States: Price Against Open Interest
For a long time the engine could tell me that sellers had been aggressive in a given week. What it could not tell me was who they were. That turns out to be the question that matters most.
So I added an open interest read. Open interest is the total number of leveraged positions still open. On its own it says very little, because every long has a short on the other side. What makes it useful is pairing the change in open interest with the direction of price. That gives four distinct states.
The two on the left are both red weeks. On a price chart they look the same. Underneath they are opposites, and the next two sections show what happened when I had one of each within a month.
Every call I make is recorded with the date it was made, the condition that would prove it wrong, and what happened. Wins and misses both. Free to read, no account needed.
See every call, scoredAugust 2026: The Fuel
On 14 August I posted that the week was shaping up to be one of the quietest in months, and that it was pairing that quiet volume with heavily negative delta. I thought a big move was coming, and soon.
The week closed at $62,843 on a big red candle, with $356 million of net selling on 17.6 billion of volume, the quietest week in two months. At the time I read that as exhausted sellers, weak hands running out of conviction.
The open interest read shows it was something more specific. It labels that week new shorts. The selling was not people giving up. It was people opening fresh positions betting on lower prices, quietly, on thin volume.
The following week Bitcoin closed at $77,702, up 23.6%, with $1,984 million of net buying, and the read for that week is short squeeze. Price rose while open interest fell, which is what forced covering looks like. The shorts that had piled in the week before were the fuel, and they got lit.
September 2026: Same Shape, Opposite Plumbing
A month later the engine served up what looked like a rerun. The week to 13 September closed at $76,816 on another big red candle, with $856 million of net selling, the heaviest since the breakout, on volume that had been falling for three weeks. Same kind of candle, same kind of volume story. The obvious read was that August was repeating.
It was not. The open interest read labels that week long flush. Nobody was building a short position. Leveraged longs were being closed out. And the two weeks before it tell the same story: both read new shorts, but selling fell from $194 million to $67 million as volume dropped from 33.0 billion to 28.9 billion. Fresh shorts were still arriving, with less size and less conviction each week.
August's lead-in looks different when you put them side by side. Volume fell into it too, but the selling bounced around rather than fading, and the week right before the squeeze was a fresh round of new shorts rather than a flush.
Which meant the week after could not be a squeeze. There was no trapped short position to run. Any move up had to come from buyers choosing to open positions, and that is what happened: the next week closed at $81,158, up 5.7%, and the read was new longs.
Look at the difference in size. August's recovery week came with nearly $2 billion of buying because people were being forced to buy. September's came with $178 million because nobody was forced. A smaller number, and a sounder one. Price rose on thin supply rather than heavy demand, which cuts both ways, and the engine says so rather than dressing it up.
From Spreadsheet to a System That Runs Itself
The spreadsheet worked, but it depended on me. If I was travelling, busy, or simply could not face another Sunday of data entry, the read did not happen. And human hands make human mistakes: a mistyped number, a row in the wrong place, a week quietly skipped.
So it is automated now. The weekly candle is pulled in once it closes. Buying and selling pressure is calculated from raw trade data. The open interest change is captured alongside it. The scoring that I once did cell by cell runs the same way every week, with no room for me to fudge it or let a good mood tip the scales.
A couple of inputs still come in by hand, deliberately. Some data cannot be pulled automatically in a way I trust, so once a week I feed it in myself. That keeps me close to the data.
What it does for me is slow me down in the right way. If I feel bullish and the read is flat, that gap is information. It also stops me chasing. Its most common output is a quiet nothing, and seeing that every week is a real antidote to overtrading.
What It Is, and What It Is Not
The Delta Engine is a structured weekly read of Bitcoin's market conditions. It is not a price prediction. It does not hand out targets. It is not a buy or sell button.
What it gives me is consistency: the same evidence, looked at the same way, whatever my mood. It started as a spreadsheet and a suspicion that a feeling was not good enough. It is now the thing I check before I let myself have an opinion.
Discussion