example report
What a leak report actually shows.
Below are the real numbers from one account of 1,153 cash hands. Not a mock up, not a rounded illustration, and not an average of anything. The most useful thing on this page is what the report refused to say.
the dataset
One account used during development. 1,153 cash hands, measured on 29 August 2026. This is not a customer, not a typical player, and not a promise. It is here because it is a real measurement, and a page about proof that used invented numbers would be arguing against itself.
results by position
Six seats, roughly 200 hands each. The third column is the number most tools stop at. The fourth is the one that decides whether the third means anything.
| Seat | Hands | bb/100 | Margin | 95 percent interval |
|---|---|---|---|---|
| BTN | 211 | -115.5 | 164.1 | -279.7 to +48.6 |
| BB | 205 | -78.6 | 198.5 | -277.1 to +119.9 |
| SB | 211 | -39.9 | 82.9 | -122.8 to +43.0 |
| HJ | 192 | -1.6 | 53.1 | -54.7 to +51.5 |
| UTG | 132 | +47.2 | 98.0 | -50.7 to +145.2 |
| CO | 202 | +51.3 | 159.4 | -108.1 to +210.7 |
The six seats add up to the full 1,153 hands, so nothing has been left out to make the table look better. The margin is 1.96 times the standard deviation of the per hand result, divided by the square root of the hand count, times a hundred. It is measured from the hands themselves rather than assumed. The intervals are computed from unrounded values, so two of the lower bounds differ by a tenth from what you get by subtracting the rounded columns.
Read the last column and the report writes itself: not one of the six positions excludes zero. The button looks like a disaster at minus 115.5 bb/100, and its interval still runs up to plus 48.6. Over roughly 200 hands a positional result cannot be told apart from break even. Whatever this account's positional game is, these hands have not measured it.
the finding that was withdrawn
before
Loses from BTN, minus 115.5 bb/100
Severity high. Price 54.08 euro per 100 hands. It sat at the top of the leak page, it steered the practice plan, and it picked the hand of the day. The gap against the reference was real and the arithmetic was correct. Only the margin had never been looked at.
after
Nothing
Since 29 August 2026 a position leak also has to clear its own margin: the favourable end of your interval must still sit below the reference. At this volume the button finding does not clear it, so it is gone. That is not lost content. It is a claim the data could never support being taken back.
The same idea had already changed a second detector three weeks earlier. Defending 51 percent of big blinds is wide by the usual standard, so the detector called it a leak. The money disagreed: across 94 defended hands the account lost 18.72 euro, and folding all 94 would have cost 32.00 euro, since folding the big blind costs exactly one big blind every time. Defending was 13.28 euro better than the alternative it was being compared to. Where the amount can be measured, the frequency is no longer enough on its own.
what a finding carries
When a detector does fire, this is what the entry contains. At most five open findings are shown at once, ordered by price, with the unpriced ones last.
Title
The pattern in one line. "Calling too often without raising", for example.
Description
The counts it was found on, written out in full. The limping entry reads: in this many of the hands you played voluntarily outside the blinds you only called instead of raising, which is this percentage. Both counts and the percentage they produce, so you can check the division yourself. Where a margin exists it is printed alongside, so a finding cannot read more certain than it is.
Severity
Low, medium or high, decided by how far past the threshold you are. It also feeds the leaks component of your Poker IQ, at 2, 5 and 10 points off a maximum of 30.
Price
Big blinds per 100 hands, converted to euros at your own weighted big blind. Blank where the loss cannot honestly be traced to individual hands, and blank means blank rather than zero.
Street
Preflop, flop, turn or river, so you know which decision to look at.
Recommended action
One sentence you can act on tonight. For limping: raise or fold, because a hand good enough to play is good enough to raise. For the big blind: defend around one hand in three and fold the rest before the flop.
Example hands
The hand identifiers behind the finding, worst result first, so the claim points back at the hands that produced it.
The price is worth one more sentence, because it is where a number can quietly become a guess. It is stored as a rate in big blinds per 100 hands and converted at your own big blind, weighted by the volume you played at each stake. On the snapshot verified in production on 25 July 2026 that was 172.2 bb per 100 hands times 33.67 cents, which is 57.98 euro per 100 hands. You can do that multiplication yourself, which is the point.
The thresholds every detector has to clear are listed on the leak finder page, and the score these findings feed into is explained on the Poker IQ page.
next
Make your own report.
Upload a session of your own hand history and the same detectors run on your hands, with your margins next to your numbers. If your sample is too small for a finding, the report will say so instead of filling the space. Getting your hand history out of GGPoker takes five steps, and the steps are here.
questions
Whose numbers are these?
One account used during development, over 1,153 cash hands, measured on 29 August 2026. It is not a customer, it is not an average, and it is not a claim about what you will see. It is on this page because it is a real measurement rather than a mock up.
Why does the report say nothing about position on this account?
Because at 200 hands per position nothing is measurable yet. Every one of the six positions has a 95 percent interval that contains zero, which means none of the results can be told apart from break even. The detector used to report the button as a high severity leak costing 54.08 euro per 100 hands. That finding was withdrawn, not because the button result improved, but because the margin was never checked.
Is an empty report a broken report?
No, and this is the part most tools get wrong. On a small sample the honest output is short. A report that always finds three leaks is a report that invents them, and acting on an invented leak is worse than acting on nothing.
How is a leak turned into an amount in euros?
The size of the leak is measured in big blinds per 100 hands and multiplied by your own big blind, weighted by the volume you played at each stake. On the snapshot verified in production on 25 July 2026 a position leak measured 172.2 bb per 100 hands at a weighted big blind of 33.67 cents, which gives 57.98 euro per 100 hands. Storing the rate in big blinds rather than cents means it stays valid when you change stakes.
How many hands do I need for a report that says something?
The style detectors start at 60 to 80 hands in the relevant spot, a VPIP finding needs 200 hands and a 3-bet finding 400. Positional results need far more than the 1,153 hands on this page, because the margin on a winrate shrinks with the square root of your volume, so cutting the margin in half takes four times the hands.