10,000 Trials Independent software studio

An edge you cannot measure is not an edge you can bet on.

We build small, exact tools for people who play games with an edge — at a table, or in a market. The name is the method. Anything can look like an advantage over ten hands or ten trades. The question is what survives ten thousand.

13 runs × 10,000 hands 1
100 handstrue edge −0.38%10,000

Real output, not an illustration. Thirteen independent ten-thousand-hand runs of the simulator inside Blackjack Ledger, against a house edge of −0.376% — the casino's average cut per hand, measured over a million of them. All thirteen finished inside the band. Six of the thirteen still finished up.

Why ten thousand

Because that is where the guesswork gets small enough to act on. Repeat a test and the answer steadies — but slowly, and at a known rate: to cut your uncertainty in half you need four times as many runs. Since √10,000 = 100, ten thousand is the smallest round number at which the answer is a hundred times steadier than a single trial. The name is not aspiration. It is the point on the curve above where the band has closed far enough to read.

± 1 point

Run something ten thousand times and the percentage you measure lands within about one point of the truth — 0.98, to be exact — nineteen times out of twenty. That is close enough to settle a decision: which of two plays is better, or what a rule change actually costs you.

100× per digit

It is also the last cheap number. Getting within one point costs about 9,600 trials. Within half a point costs 38,000. Within a tenth of a point costs 960,000. Past ten thousand, each additional decimal place costs a hundred times what the one before it did.

edges want 5×

Ten thousand settles which play is better. Proving you are actually ahead is harder, because single hands swing far more than any edge does: a spread of 1.15 units per hand against an advantage of one. After ten thousand hands a one-percent edge still scores t ≈ 0.9 — well inside what luck alone produces — and needs roughly 53,000 before anyone should say it out loud.

So ten thousand is a threshold, not a ceiling — enough to settle what to do, and a down payment on what it is worth. Below it you are telling stories. Above it you are buying decimal places at a hundredfold each, and you had better want them.

Measured, not assumed

Both products will tell you when a result is inside the noise. A number without an error bar is a guess wearing a decimal point.

Your machine

Drill history stays in your browser. The trading side runs on your hardware against your broker, with your keys. Neither is a service that holds anything of yours.

The losing ones too

Results are published whichever way they went. A tool that only shows you its wins is marketing, not measurement. See what we closed →