Monolith Mason
Bring him the rough idea.
Mason is the stonecutter of the hub. You hand him a raw block, a hunch about the market put in plain English. He squares it, tests it against the record and hands back what is left: something you can rely on, or the honest news that there was nothing in the stone.
In development. A first version that explains what is on screen runs inside our own Workspace. The work described on this page, writing and running tests for you, is not built yet.
Meet Mason
A craftsman, not an oracle.
Mason does not have opinions about where the market is going. He has a bench, a set of tools and a record of what actually happened. Give him an idea and he does the slow part: finding the right data, cutting the test to size and checking the result from every side.
He is patient, he is exact, and he is not impressed by a good looking number until he has counted how many tries it took to get it.
- Works from your block. Your question, your trade log, your rules. He shapes what you bring; he does not hand you someone else's statue.
- Measures twice. Before any cut he checks which instruments and dates can answer the question, and says what cannot be answered.
- Shows every cut. You get the code, the table and the chart, so you can follow each step and change it.
- Tells you when the stone is cracked. Most ideas have a flaw. He would rather show you the crack than polish around it.
Mason · ES researchIllustration
Here is my ES trade log since September. What was dealer gamma doing right before each of my entries?
Reads the blocktrade_log.csv, entries matched to sessions
Checks the grainexposure snapshots available for every session in the log
Makes the cutnet gamma and distance to zero gamma at each entry
Hands back your entries grouped by gamma regime with the outcome per group, the chart and the notebook. He states how many groups he compared, because that number decides how much the best one means.
entries_by_regime.pyresult.parquetchart.png
Illustration of the planned workflow. No results are shown because none are claimed.
How he works a piece
01
You bring the block
A question about markets, or about your own trades if you upload a log. Rough is fine.
02
He reads the grain
Which instruments, dates and records can answer it, and where the record is too thin to say anything.
03
He makes the cut
A query and the test that goes with it, written in Python you can read, run in your workspace under the hub's rules.
04
He hands it over
The table, the chart and the script, with the sample size and the adjusted significance stated first.
What he will do
Fetch the stone
Bars, trades, quotes, open interest and the computed exposure tables, selected by instrument and date range, without you writing a query.
Do the carving
Event studies, conditional statistics and backtests in Python. The code stays in your workspace and is yours to change.
Set your trades in context
Upload a trade log and ask what positioning, volatility and flow looked like around your entries and exits.
Keep to the rules of the workshop
No access to forward returns as inputs, and every run he makes is counted as a trial. A hundred variations are a hundred tests, and the result says so.
What he will not do
Mason does not flatter, give investment advice, tell you what to trade or place orders. He can also be wrong: he writes code, and code can contain mistakes. That is why he always hands over the script and the sample size with the answer, so you can check both.
And he cannot make marble out of sandstone. If an idea does not hold up in the record, the most valuable thing he can give you is a clear no, early.