You're pledging to donate if the project hits its minimum goal and gets approved. If not, your funds will be returned.
$4,500 left to reach the funding goal.
I have a method for reading data, and I do not trust my own judgement of it. Nobody should. So I am going to put it where judgement cannot protect it: in public, in advance, with the scoring fixed before anything happens.
The method is correlational. I call the discipline Correlational Studies, and its rule is simple — never read data in isolation. A data point on its own tells you almost nothing. Siloed across separate domains, each one kept apart from the others, it tells you less than nothing, because it looks like knowledge. Read together, connected, the same points start to form a totality you can actually decide on. That is the claim. Claims are cheap. This ledger is how the claim gets tested: forecasts written down and dated before the outcomes exist, the data and the correlations behind each one published with it, and every forecast scored afterwards against what really happened. The misses stay in. Especially the misses.
Three things. A ledger of forecasts, each one dated, each one carrying the data and the reasoning that produced it, published before the fact. A score for every single forecast once the outcome is known — a Brier score, not a vibe. And at the end, a report that puts the method next to the dumbest baselines I can construct and says, in plain language, whether it beat them.
If it did not beat them, the report says that too. That is the whole point of doing it this way.
Worked does not mean the method won. Worked means the record is complete enough that a stranger could re-run the scoring and reach the same verdict I reach. Every forecast present, every score computed the way the rules said it would be, the comparison against baselines done honestly, and a written verdict at the end that follows from the numbers rather than from what I wanted them to say. A negative verdict, fully evidenced, is a success of this project. An unevidenced positive one would be a failure.
The total is USD 4,500, and none of it pays me. Data licences and data access, about USD 2,400 — the method eats data from several domains at once, and the good sources are not free. Hosting and tooling, about USD 900, so the ledger lives somewhere public and permanent rather than on my laptop. Documentation, about USD 600, because a method nobody can reconstruct is just an opinion. Reporting and contingency, about USD 600, for the quarterly score reports and the final write-up. Every expense gets tracked against those four lines and reported in the open, next to the scores.
There is no team. It is me — Alexander Garside — working alone, unpaid, in the open. I have spent years building the correlational method this project tests; it is the same discipline I apply in my commercial work, where nothing counts as decision-grade until it has been connected and checked against the whole. I have never run it as a scored public experiment before. That gap is precisely what this project exists to close, and I would rather close it in front of everyone than keep asserting the method works because I believe it does.
The likeliest failure is the embarrassing one: the method scores no better than the simple baselines. Second: the data arrives late or incomplete, and the forecasts thin out. Third: too few forecasts in the window to judge anything fairly. If any of those happen, I publish them — the negative result, the data, the scoring, all of it. A failed method, documented properly, still leaves the field better off than an untested one making claims. And it leaves me knowing something true, which I did not know before.
None. Nothing, from nobody. This project starts at zero.
There are no bids on this project.