Picture three practice counters working the same shoe, each keeping a perfect Hi-Lo count. The running count is -1, the discard tray says two decks remain, and each of them holds 16 against a dealer 10. All three know the index for that hand: stand at a true count of 0 or higher.
One of them hits. The other two stand. Nobody miscounted, nobody misread the tray, and nobody forgot the index. The difference is a single arithmetic habit: what each one does with the -0.5 that falls out of the division.
That habit is the true count rounding convention. There are three common ones, they agree most of the time, and the places where they disagree are predictable enough to learn in one sitting.
Why the fraction has to go somewhere
The true count is the running count divided by the estimated decks remaining, and divisions rarely come out clean. A running count of +7 with two decks left is +3.5.
Index numbers, the triggers behind every index play, are whole numbers. The published Illustrious 18 says take insurance at +3 and stand 15 against a 10 at +4. For plays like these, below the index you play basic strategy; at or above it, you deviate. Comparing a fraction with a whole-number trigger needs a rule, and that rule quietly shifts every trigger by up to one point.
Usually the rule is invisible: a true count of +1.2 becomes +1 whichever way the fraction is handled. It only bites when the raw count lands just around an index, and some important indices sit exactly there.
Three ways to make a whole number
Simulation software documentation, such as the help pages for CV Blackjack, describes three conversion methods. They differ only in which direction the fraction goes.
- Floor. Take the next lower integer. +5.8 becomes +5, and -5.2 becomes -6.
- Truncate. Drop the fraction, which moves toward zero. +5.8 becomes +5, and -5.2 becomes -5. On positive counts this matches floor; on negative counts it moves up.
- Round. Go to the nearest integer, with exact halves rounding up. +5.8 becomes +6, -5.2 becomes -5, and +3.5 becomes +4.
| Raw true count | Floor | Truncate | Round (halves up) |
|---|---|---|---|
| +5.8 | +5 | +5 | +6 |
| +3.5 | +3 | +3 | +4 |
| +0.4 | 0 | 0 | 0 |
| -0.5 | -1 | 0 | 0 |
| -1.5 | -2 | -1 | -1 |
| -5.2 | -6 | -5 | -5 |
Notice what happens near zero. Truncating sends every raw count between -1 and +1 to zero, a band two points wide. Flooring sends only 0 up to just under +1 there, and rounding sends -0.5 up to just under +0.5. Truncating produces the most zeros of the three, and that matters for one hand in particular.
Where the methods disagree: 16 against a 10
Standing on 16 against a 10 at 0 or higher is the most used hand deviation in the Illustrious 18, and its index sits right where the methods diverge. Back to the opening: running count -1, two decks left, raw true count -0.5.
- Floor gives -1. That is below the 0 index, so you play basic strategy and hit.
- Truncate gives 0. The index is met, so you stand.
- Round, with halves rounding up, gives 0. You stand.
Now try a running count of -3 with two decks left, a raw count of -1.5. Floor says -2 and the other two say -1, but all three sit below the index, so all three hit. For this hand the action only changes between methods when the raw count falls between -1 and 0.
That makes zero-index plays the pressure point: negative fractions are where the conventions disagree most. For why this hand is such a close call to begin with, see our breakdown of 16 against a 10.
21 Trainer ProA true count drill plus index deviation training, so the conversion and the decision get practiced together.
Get the app
Positive fractions: insurance and 15 against a 10
Above zero, floor and truncate are identical, because dropping the fraction and stepping down are the same move. Only rounding can disagree, and it does whenever the fraction is a half or more.
Take a running count of +7 with two decks remaining. The raw true count is +3.5.
- Floor and truncate both give +3. Round gives +4.
- Insurance, indexed at +3, is met under all three methods.
- Stand 15 against a 10, indexed at +4, is met only under rounding. The rounding counter stands; the floor and truncate counters hit.
This also sharpens advice common in counting guides, ours included: round conservatively, toward zero, when the division is messy. For positive counts, toward zero means downward, which keeps a counter from overstating a good shoe. For negative counts, though, toward zero is upward, which makes a poor shoe look slightly better than it is. The only method that never puts the count above its raw value, whatever its sign, is floor.
Choosing a convention and sticking to it
Index numbers are generated by simulation, and the simulation has to assume a conversion method when it finds the count where a play flips. The ideal is to use the method your index set was generated with. If your source documents it, follow it. If not, choose one and treat that choice as part of the index set.
The choice matters less than the arguments about it suggest. Simulation software documentation reports that flooring and rounding perform similarly, and describes truncating as slightly inferior because it produces more true counts of zero, which is exactly the double-width band in the table above.
Consistency matters more. A counter who floors on one hand and rounds on the next shifts every index by a varying amount, and the resulting decisions match no index set at all. Pick one method, practice with it until it is automatic, and never renegotiate it mid-shoe. Memorized index numbers are only as reliable as the count they are compared against.
The denominator is usually the bigger error
Rounding moves a true count by less than one point. The deck estimate can move it that much on its own.
Take a running count of +6 late in a shoe. With exactly two decks remaining, the true count is +3. Read the tray as 2.5 decks and it becomes +2.4; read it as 1.5 decks and it becomes +4. One half-deck miss, at the precision many counters work to, moves the count by up to a full point before any rounding happens.
Early in a shoe the same miss barely registers: +6 over five decks is +1.2, and over 4.5 decks about +1.33. That is why careful deck estimation late in the shoe often repays more practice than the rounding debate does.
One system sidesteps the question. The KO count is unbalanced by design so players can skip the true count conversion and compare the running count with fixed values. With no division, there is mostly no fraction to round.
Drilling the conversion until it disappears
A convention only helps if it runs automatically: glance at the tray, divide, apply the rule, compare with the index, act. Rounding is the step people rarely drill, because it feels too simple.
A useful routine: write out divisions that land on awkward fractions, such as +7 over 2, -1 over 2 and -3 over 2.5, and say the whole number under your chosen method until it takes no thought. Then attach an index to each and say the action.
Keep the larger frame in view. Counting with your head is legal in the United States, but casinos are private businesses that may refuse service to suspected counters, and no technique, a well-chosen rounding convention included, removes risk from the game. For anyone learning Hi-Lo as a skill, accuracy is the whole point of the drill.
Practice the whole loopCounting drills for Hi-Lo, KO, Omega II and Zen in 21 Trainer Pro. Virtual chips, no account, works offline.
Download
Frequently asked questions
Should you round the true count up or down?
It depends on the convention your index numbers assume. Flooring always moves down, truncating drops the fraction toward zero, and rounding goes to the nearest whole number with halves rounding up. Simulation software documentation reports that flooring and rounding perform similarly and that truncating is slightly weaker. The most important rule is to pick one method, ideally the one your index set was generated with, and apply it the same way on every hand.
What is the difference between flooring and truncating a true count?
On positive counts there is no difference: +5.8 becomes +5 under both. They split on negative counts. Flooring takes the next lower integer, so -5.2 becomes -6, while truncating drops the fraction toward zero, so -5.2 becomes -5. The gap matters most for plays indexed at zero, such as standing on 16 against a 10, where a raw count of -0.5 means hit under flooring and stand under truncating.
How do you round a negative true count in blackjack?
Apply the same convention you use for positive counts. Under flooring, -0.5 becomes -1 and -1.5 becomes -2. Under truncating, both move toward zero, giving 0 and -1. Under rounding with halves up, they also become 0 and -1. Negative fractions are where the three methods disagree most, so a counter who has never thought about the sign of the count may be applying two different rules without realizing it.
Does true count rounding matter more than estimating decks remaining?
Usually not. Rounding moves the true count by less than one point. A half-deck error in the deck estimate can move it by as much as a full point late in a shoe: a running count of +6 is +3 with two decks left, +2.4 if you read 2.5 decks, and +4 if you read 1.5. Settle the rounding method once, then spend practice time on reading the discard tray.
Do you need to round the true count when using the KO count?
Mostly no. KO is an unbalanced count designed so players can skip the true count conversion and compare the running count with fixed values instead. Without a division there is usually no fraction to round. Balanced counts such as Hi-Lo are where rounding conventions come up, because their betting and playing decisions key off a true count that rarely divides evenly.