Reports: we measure the claims about money that are going around

We take a claim, measure it with public data and publish the result. Free, no sign-up, and we don't recommend anything.

Past data and simulations: they do not guarantee future results.

Each report states how much data it uses, from what period, and what would overturn the conclusion. We include our mistakes; if you spot another one, write to us.

Start with this one

5 August 2026

«It triples the losses»

A claim from hundreds of videos about leveraged products. In July, chip stocks fell 21.20% and the “3X” product fell 56.99%. Not the 45% going around. And what hurts it most isn't the crashes: it's the boring months.

2,661 sessions · January 2016 to August 2026 · includes two mistakes of ours and the prediction we wrote before looking at the data   Read →

Method

8 August 2026

When you look decides more than when you enter

We measured 129 ways of starting to buy bitcoin little by little. Just by changing the day you take the snapshot, the best entry point becomes the worst. In nine out of ten windows the bottom zone comes out among the best; in one, the worst of all — and that one was the one we had chosen.

Corrects an earlier report of ours · we published what would knock us down, we checked it, and it knocked us down   Read →
14 August 2026

A high win rate is not good news

It is the first number people show and the most reassuring one. It is also the easiest to raise without improving anything. We tightened the filter of our own system to check it, and the opposite of what we expected happened: the result went down and so did the hit rate.

Test on past data · 11,988 simulated trades · 148 stocks · 16.6 years   Read →
14 August 2026

How much history do you need to trust a strategy?

The question comes up again and again and almost nobody answers it. It matters less how many years you have than how many pieces you split them into: here is a case of ours where splitting them in half changed which version went into the system.

148 stocks · 16.6 years · both halves published, including the one that makes us look worse   Read →
14 August 2026

The same test, twenty-five different results

If your account has a limit on open positions, there are days when more marks arrive than fit, and what decides which ones get in is usually chance. We measured how much that chance weighs: it weighs more than almost everything we were comparing.

Test on past data · 11,988 simulated trades · 148 stocks · 16.6 years   Read →

Bitcoin

3 August 2026

Bitcoin: the cycle clock

Where a bitcoin bear market ends, measured along five different paths over the three cycles that exist. With the control group, which is almost never published: what buying on a random date returns. And with the conditions, written in advance, that would break each path.

4,962 days of on-chain data · 2013 to August 2026 · three complete cycles, which is a small sample, and we say so   Read →
8 August 2026

Does “buy when the average holder is at a loss” work?

It's the most repeated rule in Bitcoin. We measured it with two reasonable comparisons: one says it doesn't work; the other, that it's right 18 times out of 18. What decides isn't in the market: it's what you compare it against.

4,963 days · 2013 to August 2026 · both tables published, including the one that makes us look worse   Read →
8 August 2026

The average price everyone bought at

The average price people bought at looks like an objective figure. Calculated in two defensible ways, the market is either 20% above it or 27% below it. Same day, same data. What changes is who you include in the average.

52,760 $ versus 87,404 $ · the most common trick and the hardest to spot   Read →
8 August 2026

Four Bitcoin bottoms: are they getting milder, or is it chance?

There's a way to measure how much bitcoin holders suffer. In the three completed winters it rises cycle after cycle: 0.564 → 0.690 → 0.754. Four observations don't make a rule. We measured it and we say so.

Four bottoms, three completed · four observations do not make a rule, and we say so   Read →
8 August 2026

When miners switch off: which indicator measures it and how many times it got it right

When mining stops paying, some miners switch off and the network loses power. One indicator claims to read that, and we measured it. Our first count was wrong: it came out at 35 alerts in thirteen years. Counted correctly, there are 19.

4,962 days · January 2013 to August 2026 · includes our whole mistake and the test nobody has done yet   Read →
5 August 2026

What the halving really does, and what it doesn't

The halving gets credited with the whole cycle. What it does is concrete and verifiable: it cuts new issuance in half. What it doesn't do is set a date, because it doesn't happen on a given day: it happens at a block height.

Dates and rewards verified against block height reconstructed from the circulating supply   Read →
5 August 2026

The clock that can't be adjusted

The time distances between the milestones of the three complete cycles. Bottoms are given with three different definitions —the lowest day, the lowest week and the lowest month— because with only one the precision is misleading. It is an estimate of when, not of how much.

4,963 days of daily closes · up to 2 August 2026 · three cycles, which is a small sample, and we say so   Read →
5 August 2026

Bitcoin's clock doesn't count months, it counts blocks

The chain does not know what day it is. We reconstructed the block height day by day and something came out that we did not expect: between bottoms there is a difference of thousands of blocks, but of only six calendar days. It is a rejected strategy candidate, and it is published anyway.

Heights reconstructed from the circulating supply · actual average pace of 9.58 minutes per block   Read →

And how do you react to a loss?

Take the test: 5 minutes, free →

Training material. It is not financial advice or a recommendation to buy or sell anything.