What does it mean to test a betting method?

2026-09-24

What does it mean to test a betting method?

"I won all week with this method" is a common line. But how do you tell whether that week was the method or luck? Testing a betting method comes down to answering that question. Here are four things we learned while collecting shoes.

1. How many hands do you need?

Banker share in real consecutive hands. Each bar shows the range 95% of samples fall in.

Banker share in real consecutive hands. Each bar shows the range 95% of samples fall in.

Overall, banker comes up about 50.8% of the time. Cut out just 100 hands, though, and it swings between 42% and 60%. Even over 1,000 hands it moves between 48% and 54%. It takes 10,000 hands to settle near 50%, and 100,000 to narrow to 50.5% to 51.1%.

In other words, spotting a difference of about 1% per hand needs tens of thousands of hands. A day or two of play cannot tell you whether a method is good or bad.

2. Every day is different

Banker share by day. Even tens of thousands of hands a day move this much.

Banker share by day. Even tens of thousands of hands a day move this much.

Collect tens of thousands of hands of the same game every day and the banker share is still a little different each day. A method that worked one day and failed the next is, most of the time, not the method but this movement.

3. The best of many is hard to trust

This is the trap to watch most. Run hundreds of methods and a few will look excellent by chance alone. We ran a real test: we took thousands of betting orders for the start of a shoe, ran them on the first half of our data, and picked the 218 that never broke. Then we ran those 218 unchanged on the second half. 208 of them broke.

They had looked good not because of the method, but because their turn had simply not come yet. That is why we do not trust "the best one so far" as it stands.

4. Split the data in two

So we split. Choose a method on the earlier period, then check it on the later period that played no part in choosing. Only if both periods point the same way is it worth trusting even once. And we only use shoes we actually collected. We never test on shuffled or invented data.

Summary

100 hands         banker share swings 42 to 60%
1,000 hands       48 to 54%
10,000 hands      50 to 52%
100,000 hands     50.5 to 51.1%
The best so far   208 of 218 broke on the second half
Rule              keep choosing data and checking data apart · real shoes only

Hiddenpick is a Windows program that runs Evolution live baccarat on Stake for you, following the rules you set. How we collect shoes and how many we have so far is published in the R&D centre.

HIDDENPICK
Built on data we collect ourselves, run by a program that never tires
  • 25,000+ real shoes collected and analysed since September 2026
  • 90+ baccarat tables from the lobby on one screen · refreshed every 0.35s · seated 2 to 4 seconds after a click
  • Stops by itself at your take-profit and stop-loss · tie several PCs into one team and stop on the total
  • No browser extension · your money stays in your own Stake account · 12 languages
Go to www.hiddenpick.my →
← Bacaan