How a Trading Educator Should Score Their Own Calls
A call only counts if it was logged before the outcome was known, and it is only scored against its own stated reference — the direction, the timeframe, and the invalidation given at the moment it was made. That is the whole model. Everything else on this page — the grade scale, the conservative rule, the evidence chain — exists to protect those two rules from the person most motivated to break them: the caller.
I score my own public record this way, and I graded 139 of my own calls with it — the full result, misses included, is published in I graded four years of my own public calls. This page is the model itself, in full, so you can apply it to your own calls or use it to interrogate anyone who claims a track record.
What makes a call scorable in the first place?
Most statements traders make are not calls. They are commentary. A call is scorable only if it carries five things, all captured at the time it is made:
- A date — logged the same day it was made, before the outcome exists. A call written down after the move is not a call; it is a memoir.
- An instrument — a specific, named thing whose price can be checked independently.
- A direction — up, down, or explicitly neutral. "Interesting chart" is not a direction.
- A window — a timeframe in which the call lives and then expires. A call that can never expire can never be wrong, which makes it worthless as evidence.
- A reference — the level, target, or invalidation that defines what "wrong" looks like, in the caller's own words.
Anything missing one of these gets excluded from the denominator — and the exclusions get disclosed, not buried. In my own database, 139 calls are graded and 132 of them reach the published record — 6 are held because I cannot prove they were public, and 1 is a standing process rule with no instrument to score. Of those 132, another 20 fail the tests above: conditional setups whose trigger never fired, and calls too recent to resolve. That leaves 112, and 112 is the denominator: 33 clean wins (33 of 112, about 29%), 73 of 112 at least directionally right (about 65%), and 26 outright losses (26 of 112, about 23%). The denominator is where most published track records lie, which is why it has to be printed.
What should the grade scale look like?
The trap is binary win/loss. Real outcomes are messier than that, and a scale that cannot express "right, but not the way I said it" will quietly inflate everything into a win. The scale I use:
| Grade | What it means | Counts as |
|---|---|---|
| Clean win | Direction and stated outcome achieved inside the window, without a serious adverse move first | Hit |
| Direction-correct | Right direction, but the target missed, the timing was off, or price moved against the call first | Soft hit |
| Theme-correct | The broad read held; the specific level or trade did not | Soft hit |
| Partial | Mixed — half-worked, or the stated stop would likely have hit before the move | Neither |
| Loss | Price moved against the call inside its window, or the stop hit before the target | Miss |
| Unclear / too early | Not gradeable, or the window has not elapsed | Excluded |
The separation between "clean win" and "direction-correct" is the one that matters. A call that was eventually right after a double-digit adverse move first did not help anyone who acted on it when it was made. Folding those into the win column is the most common way an honest-looking record overstates itself. Keep them in their own row and say so.
The long version of this is in the book — Become a Cyclitecnical Trader: the cycle ladder, the FLD, and the eight interactions, written out end to end. It's free. Send me a copy. We email it to you. No card, and you can unsubscribe any time.
Why grade ambiguous calls against yourself?
Because conservative grading is the only thing that makes the favorable numbers believable. Two rules, both taken from my own ledger:
Ambiguity resolves to the less favorable read. One 2026 call framed a retailer both as a bearish breakdown and, in the same breath, as a "search for the bottom." Genuinely ambiguous — one framing would have won, the other lost. The stock bounced about 3% over the following week, so the call was scored against the dominant bearish framing and logged as a miss. Painful, and correct: if the caller gets to pick the framing after the fact, every ambiguous call becomes a win.
Self-reported results get downgraded to what independent data can verify. A 2026-07-09 same-day note on crude oil described the session as a drop of about 3.5%. The independent close-to-close check came in nearer 2%, so the record carries the smaller, verified figure rather than the one written in the moment. The grade got harder in the same pass: the bearish lean that number was attached to was contradicted over the following two weeks, so the entry is logged an outright loss, not a same-day observation that happened to be right about the day. The record keeps the verified number and says why.
How do you keep yourself from editing history?
The scoring model is worthless if the past can be quietly rewritten. Four links make the record tamper-evident:
- Third-party timestamps. My live calls are spoken on recorded video sessions, so the recording platform owns the timestamp and the exact words. The caller cannot move either one.
- Same-day logging. The ledger entry is written within hours, before the outcome exists. You cannot write tomorrow's call with today's hindsight.
- Append-only public history. The ledger lives in version control with a public change history. Turning a past miss into a hit is not a silent edit — it leaves visible tracks. Honest limit, said out loud: version history can technically be rewritten by its owner, so this link alone is tamper-evident rather than truly immutable. That is enough, because of the fourth link.
- Publishing. Once a scored result is public, it is out in the world. A published miss cannot quietly become a hit — people saw it. Publishing is the lock.
Here is the model doing its job on my own record, on the terms each call set for itself:
2026-07-09 — S&P futures, bullish into a peak. Miss. The call named its own invalidation on air: "The invalidation is clean: today's session low." Six sessions later the index closed below that low — the call's own tripwire — and it was scored a miss on the terms it set, not on anyone's interpretation. One more honesty note carried in the same entry: the exact futures session low was not captured at logging time, so the score used the cash index as a proxy — and the ledger says so rather than pretending to a precision it does not have.
2026-07-09 — AMD, long. Miss. Everything agreed at the moment it was made: the sector read, the broad market read and the cycle read all pointed the same way, and the call was stated with that agreement attached. It worked for exactly one session, then rolled over and sat about 9.3% below the call level a week later. Logged a miss — a loss made memorable by how much confidence was behind it, which is exactly why it has to stay in the record.
Which numbers should you publish — and which should you refuse to?
- Percentages, never dollar amounts. A graded call is a scored public statement, not an account statement. Dollar figures imply a portfolio that this record does not represent.
- Hits and misses in the same table, at the same prominence. A scoreboard that hides misses is the thing skeptics and regulators both distrust — for the same reason.
- Never blend real and simulated. If simulated results exist anywhere, they are labeled and kept in their own ledger, permanently. Two ledgers, never merged.
- Print the denominator. State how many calls were excluded and why. Excluding un-gradeable calls is legitimate; doing it silently is not.
- Never cherry-pick a source. If the record is shown, it is shown whole. My 2022 slice graded far better than my 2026 live slice; leading with the 2022 number alone would be textbook selective presentation. The number I lead with is the whole database.
- Attach no promises. The record describes a process. It does not forecast outcomes, and no performance claim of any kind hangs off it.
What is a scoreboard actually for?
Not what most people think. An honest scoreboard is character proof, not skill proof.
Mine currently says: stronger on the short side and the broad regime read than on the long side. Of the 68 gradeable calls made to the bearish side, 49 were at least directionally right — 49 of 68, about 72%. Of the 39 made to the bullish side, 20 were — 20 of 39. The most recent full on-air session on the published record, 2026-07-09, produced 9 gradeable calls: no clean wins, 5 of them outright losses, 4 of them directionally right. Read plainly, parts of my own record argue against me, and I publish them anyway, because that is the point. Every other record you will encounter in this niche is unverifiable marketing. A dated, public, conservatively graded record — losses included — is the one thing a competitor will not copy, because they can copy a method but they will not publish their misses.
What it buys the reader is weighting: which half of an operator's read to trust. What it buys the caller is discipline — once every statement is scored, "nothing here today" becomes a legitimate entry instead of a failure, and that restraint has its own page: the null trade. If you are auditing someone else's claimed record before trusting it, pair this scoring model with how to verify a trading guru's credentials.
The scoring model itself is method-agnostic — it grades dated statements against price no matter what produced them. The cycle work that produces my own calls is a separate subject, documented at the methodology hub.
Questions traders ask
Can I score my own calls in a spreadsheet?
Yes, and you should. Five columns capture a call — date, instrument, direction, window, reference — plus a grade column filled in only after the window closes. The tooling matters far less than the two disciplines: log before the outcome, and append rather than edit. If you find yourself rewording an old entry, you have stopped keeping a record and started writing a story.
What if a call has no timeframe?
Then it is not scorable, and it goes in the excluded pile — disclosed, not deleted. An open-ended call can always claim vindication eventually, which means it can never be falsified, which means it is not evidence of anything. When you hear a prediction with no expiry, you are hearing content, not a call.
Isn't grading your own calls a conflict of interest?
Completely — which is why the model is built to lean against the grader. Independent price data supplies the outcomes, ambiguity resolves to the less favorable read, self-reported magnitudes get downgraded to what the data verifies, and every entry is dated and public so anyone can re-grade the set. Self-graded and private means "trust me." Self-graded, conservative, dated, and public means "check me." Only the second one is worth anything.
How many calls does it take before the record means something?
More than feels natural. My most recent week graded 3 hits and 6 misses — on its own, noise. Against the 112 published, graded calls in the public record it becomes a data point inside an established pattern: the long side is my weak half. Judge any record — yours or anyone's — on its length and its worst slices, never its best week. And even a long record only describes the past; it entitles nobody to a prediction about the next call.
Tools our own desk actually runs. No vendor pays to be listed, and no link here earns us a commission.
Keep reading
Get the Cycle Pass — from $297