Search "trade thesis example" and you'll find chart screenshots. Some have a green box where the trade goes on, a red box where the stop sits, an arrow pointing where the trader thinks price is heading. There's a caption: long here, target X, stop Y. Sometimes a sentence about the setup. Sometimes a one-word mood — bullish.
Those aren't theses. They're charts with notation. A trader at a fund whose desk requires written rationale before risk goes on would have it back in their inbox in twenty minutes asking for what is the actual claim, and why.
A real thesis has four parts. The fourth one is the part everyone skips, and it's the one that turns the document from a chart annotation into something you can grade later.
1. The claim
What you think will happen, in a sentence, with enough specificity that someone else could check whether it came true.
"NVDA will gap down at least 3% on the print and not recover within five trading days." That's a claim. "NVDA looks heavy." Is not a claim.
The test isn't whether the language is fancy. It's whether a reasonably attentive friend, looking at the world six weeks later, could say yes that happened or no it didn't. If they can't, the claim isn't doing the work a claim is supposed to do — it's a feeling dressed in finance vocabulary.
The cleanest claims have a price level (or a range), a horizon, and a direction. Brent holds 70 through end of June. XLE underperforms SPX over the next quarter. The next CPI print comes in below consensus. Specific enough to be wrong.
2. The mechanism
Why the claim should be true. The thing about the world that makes the move likely.
This is the load-bearing piece for self-knowledge, because the mechanism is what you're actually testing. The claim is a prediction; the mechanism is the theory the prediction is testing. When you're right, the question worth asking is did I get the right answer for the right reason — and that question is unanswerable without the mechanism written down.
Worked example: NVDA will gap down at least 3% on the print could be supported by half a dozen mechanisms, and they imply different things if it pays off:
- Hyperscaler capex commentary will turn cautious and the multiple compresses on guidance, not on the print itself.
- Inventory builds in the channel reports already suggest a slowing data-center cycle.
- Positioning is too long going in; the bar is set too high regardless of the actual numbers.
If you wrote the first mechanism and the print gapped down on the second, you got the trade and lost the thesis. That distinction is what compounds. A year of trades graded against P&L only will tell you you're up; a year of theses graded against mechanism will tell you which kinds of reads you're actually any good at.
3. The horizon
When the claim has to resolve.
Markets reward being right at the right time and punish being early indistinguishably from being wrong. A claim without a horizon — eventually NVDA derates — is not falsifiable in any operational sense, because eventually is doing the work of an excuse you haven't needed yet.
A horizon is a date, a window, or an event. By the next earnings. Within three months. Before the Fed's December meeting. Once the horizon passes, the claim has resolved one way or the other. If it hasn't, the fault is in the claim, not in the time.
This is also where conviction starts to mean something. I think this happens within a quarter, with 60% confidence is a probabilistic forecast you can grade. I'm bullish but it might take a while is performance.
4. The falsifier
What would make you abandon the thesis — not the trade, the thesis.
This is the part that almost nobody writes down, and it's the part that does the most work. A thesis without a falsifier is a position you defend; a thesis with one is a position you can update.
The falsifier is not the stop-loss. A stop-loss is a risk-management decision about price; it can fire while the thesis is still alive. ("Got stopped, still bullish" is a real sentence on a real desk, and it's coherent if the thesis is intact.) The falsifier is the state of the world that would make the thesis itself wrong.
Worked example, continuing the NVDA setup. The trade was: short going into the print, gap down expected of 3%+. The falsifier might be: if the print prints in line and the call adds buybacks or guides up the data-center segment by more than 5%, the multiple compression thesis is wrong; close out. That sentence does work that a stop level can't, because it tells you what information should make you change your mind, separate from what price should make you exit.
When you grade the trade, the falsifier is the cleanest signal of whether the thesis was sound. A trade can lose money and the thesis can still have been correct — the falsifier didn't fire, the call paid off on a longer horizon, you just sized too aggressively. A trade can also make money and the thesis was wrong — the falsifier fired and you got bailed out by a different mechanism. Both are real. Both are invisible without the fourth piece.
What the four together produce
Put them in sequence and a thesis reads like this:
Claim: NVDA gaps down at least 3% on the next print and doesn't recover within five trading days.
Mechanism: Hyperscaler capex commentary on the call turns cautious for 2027, multiple compresses on the forward.
Horizon: Resolves on the day after earnings; recovery window through the following Friday.
Falsifier: If hyperscaler commitments on the call are flat or rising vs. last quarter, the mechanism is wrong; the trade is also wrong, regardless of whether price gets there for unrelated reasons.
That's a hundred and twenty words. It can be written in five minutes. It is also what every serious sell-side note, every long-form Twitter thread that earns its likes, and every fund-manager pitch is doing under the prose. Strip out the formatting and the brand voice, and the four pieces are what's left.
Why this is the version that compounds
The point isn't that you'll get every trade right if you write the four down. You won't. The point is that the version of you who has been doing this for a year will be able to answer questions that the chart-screenshot version of you can't:
- Which mechanism types do I have an edge on? (Earnings catalysts? Macro repositioning? Single-name de-rates?)
- What horizons am I calibrated on? (One-week catalysts? One-quarter theses? Multi-year structural calls?)
- When my claims are wrong, are they wrong in the same way each time? (Right direction, wrong horizon? Right mechanism, wrong instrument?)
A year of charts can't answer those. A year of four-part theses can.
You don't need software to do this. A notes app and the discipline to write the four lines before the trade goes on is enough. You do need something, because the version that lives only in your head is going to consolidate with the outcome by next week and tell you a flattering story about what you actually thought.
That's all HunchLog is for, in the end. Four fields, captured before the answer is in. The discipline is in the design.