Learn

Every calculation on this site, explained plainly

Twelve short lessons, in reading order. No statistics background needed. Each lesson gives the plain answer first, one worked example, the actual method for those who want depth, and what the number does NOT mean.

The curriculum

01
Why profit is not proof
Over a few dozen bets, profit mostly measures luck. The site exists to separate the part of a record you can read from the part you cannot.
02
Entry price and realized edge
The core measure: the average of resolved outcome minus entry price. A real 74% win rate with a negative edge shows why the price paid is half the story.
03
Confidence intervals
Every estimate carries a 95% interval in the same sentence. The bootstrap in plain words, and why a wide interval means thin evidence.
04
The four-state verdict
Skilled, not separable from luck, thin sample, plus a reserved integrity hold. The gates in order, the two sufficiency floors, and how style labels travel with a verdict without changing it.
05
Win rate vs edge
Across 3,871 tested wallets, entry price and win rate move together at Spearman 0.951. Win rate describes taste; the edge nets the price out.
06
Calibration and the Brier score
Entry prices are implicit forecasts. Calibration checks them band by band against what happened, and the Brier score compresses it into one number, lower is better, split per category.
07
Concentration
One event at 60% or more of the net result attaches the concentrated style label: the per-position edge can be clean while the dollar outcome rode on a few events. The ratio can exceed 1.
08
Edge Score
The descriptive composite (posture, conviction, discipline) fit on 8,656 wallets, including the forward test it failed and why it is never sold as a per-wallet predictor.
09
Many wallets at once
Test 3,871 records and some look skilled by chance. False-discovery control in plain words, and the registry result stated exactly: 178 clear at q = 0.10, in sample.
10
Negative controls and evidence grades
500 size-matched placebo cohorts through the same pipeline keep us from fooling ourselves, and the A/B/C/D legend compresses the evidence with frozen gates.
11
Markets, not just wallets
Market Trust (research preview), the market integrity read, and the coherence engine, each with its evidence boundary printed on the surface itself.
12
What we never say
Not advice, not forecasts, in-sample labeled as in-sample, forward tests filed with verdicts pending, and failed tests published. The restraint is the product.

Reference glossary

One canonical page per term, for when you want the full definition rather than the lesson.

Every number quoted in these lessons matches its published receipt: the methodology papers under /research, the frozen registry artifacts, and the live analyzer. When a lesson and a receipt disagree, the receipt wins and the lesson gets fixed.