A high VPIP gets you called loose and a low 3-bet gets you called soft, but ATS is a peculiar number: when it is low, nobody tells you.
Nobody at the table complains that you are not stealing. The two players whose blinds you let through just think they are running well tonight. You will not feel it either — the hands you folded were J7, K4, Q6, hands you did not want to play anyway, and every fold looked perfectly sensible.
Then a month goes by, your results have not moved, and you cannot work out where you went wrong. Because the thing that went wrong never happened. It merely failed to happen.
ATS is the statistic that turns the thing that never happened into a number. This post covers how it is calculated, what each of the three seats should be, what too low and too high each cost you, how to respond to an opponent’s ATS, and how large a sample it takes before it counts. Read the VPIP post first if you have not — the two ideas it sets up, range width and sample size, get used throughout here.
What ATS Is: It All Hangs on the Denominator
ATS stands for Attempt To Steal.
The numerator is simple: you raised. The denominator is where the number lives, and it is much narrower than most people assume:
Every hand where everyone in front of you has folded, the action is on you, and you are in the CO, the button or the small blind.
All three conditions must hold. Fail any one of them and the hand never enters the denominator at all:
| Spot | In the denominator? | Why |
|---|---|---|
| You are on the button, four players fold in front | Yes, and raising counts as a successful attempt | The standard steal spot |
| You are on the button, CO has already raised | No | The pot is open; that is a 3-bet, not a steal |
| You are on the button, CO merely called | No | Once someone is in the pot it is not a steal; that is an isolation raise |
| You are in the CO, everyone folds, you choose to call | In, but counted as no attempt | The chance was spent, the numerator did not move |
| You are in MP, everyone folds | No | MP is not a steal seat |
| You are in the BB | No | You are the one being stolen from |
Two things are worth stopping on.
First, calling — limping — is the most expensive cell in that table. Everyone folds, you are in the small blind, and you put in the other half-blind to see a flop: the hand is in the denominator with nothing in the numerator. Your ATS takes the hit, and what you bought is a pot with no initiative in which you speak first on every street. It costs you the number and the expected value in the same direction.
Second, ATS is naturally much larger than VPIP, and that is not an anomaly. VPIP’s denominator is every hand you are dealt; ATS’s is only the subset where nobody is in front of you and you are in late position. Someone with a VPIP of 22% and a button ATS of 45% has two healthy numbers, because they are counting entirely different groups of hands. The VPIP post mentions that “playing 40% on the button is reasonable” refers to the opening range rather than VPIP — that opening range, measured, is exactly this.
(Short-handed tables shrink the steal seats along with everything else: three-handed only the button and small blind count, heads-up only the button. The fewer players there are, the less the idea of “late position” means anything.)
Why It Matters: The Only Money on the Table That Moves by Itself
Nearly all the money in poker requires somebody to put it in voluntarily. There is exactly one exception: the blinds appear on the table every orbit, no matter what anybody is dealt.
At six-max you post one small blind and one big blind per orbit, 1.5bb in total. A hundred hands is about 16.7 orbits, so:
Every 100 hands, roughly 25bb is taken from you by force.
That number deserves to be held next to your win rate. A 5bb/100 winner earns, over 100 hands, less than a quarter of what he posts in blinds over the same stretch. Blinds are not a rounding error; they are the most reliable, least justified expense you have at the table.
And one part of the table exists specifically to pick that money back up: late position. When everyone folds and only the two players forced to put money in remain behind you, there is 1.5bb sitting in the pot and nobody has shown anything at all. Raise, both fold, and you take the 1.5bb without showing a hand.
Look at how generous the arithmetic is. Say you raise to 2.5bb on the button:
- Worst case (assume that any time you get called you lose the 2.5bb outright): you need both to fold 62.5% of the time to break even, because 2.5 ÷ (2.5 + 1.5) = 62.5%
- But “called means lost” is false. You have position, you sometimes actually hit, and you can bet again after the flop to make him fold
So the real threshold is far below 62.5%, which is why theoretical button steal ranges run past 40% of hands — it does not need to win, it only needs to not have to play. (The full version of this “how often does it have to work” arithmetic is in how to work out pot odds.)
Put the other way round, what a player with a low ATS is doing is this: posting 1.5bb every orbit, then handing the chance to pick it back up to the next player whenever he reaches the best seat at the table. That is not conservative. That is a one-way subsidy.
Three Positions, Three Numbers
There are three steal seats, and their reasonable frequencies differ a great deal. RushCraft’s three built-in ATS diagnostic cards set their target bands like this:
| Position | Target ATS | Who is left behind you | Character |
|---|---|---|---|
| BTN (button) | 43% – 48% | Only SB and BB | The widest. Last to act on every street after the flop |
| SB (small blind) | 40% – 48% | Only BB | Also wide, but first to speak on every street after the flop |
| CO (cutoff) | 27% – 30% | BTN, SB, BB | Clearly narrower. The button behind you has position on you |
The specific hands behind those bands are the trainer’s default opening ranges — the preflop range charts draw all five positions out square by square. Measured, that set of defaults comes to 46.9% on the button, 46.9% in the small blind and 29.1% in the cutoff, all three inside their own bands.
Three numbers, three different reasons:
The button is widest because positional advantage compounds. You are not merely last to act this round; you are last on the flop, the turn and the river. An ordinary hand is fine, because the information advantage will make up the difference over the next three streets.
The small blind’s number is close to the button’s for the opposite reason. It is not wide because the position is good — it is the worst position at the table, and you speak first after the flop forever. It is wide because you only have to get through one player, and because you have already been forced to put half a big blind in. The right conclusion in the small blind is not “play more”, it is raise or fold: raise if it is worth playing, fold if it is not, and the call in the middle is the most expensive option there is (see the first section).
The cutoff is barely half the button because someone with better position is sitting behind you. Open from the CO and the button can 3-bet you with a whole range of medium-strength hands, knowing that he has position for the rest of the hand. Stealing from the CO is not pressure on the two blinds; it is getting past the button first.
And something that matters more in practice: those three numbers are a starting point, not a destination. The correct ATS depends on how the two players to your left defend. With someone who never defends his big blind on your left, your button should push toward 48% and beyond; with someone who 3-bets you every time, 43% is already too much. For how position systematically changes your choices, see the full guide to position.
Too Low and Too High: Two Completely Different Illnesses
ATS is one of the few statistics where both directions cause trouble, and the trouble is nothing alike.
Too low: giving up free money and going transparent while you are at it
The expected value you forgo is only the first layer, and it is the less painful one. The real problem is the second: when a player whose button ATS is 25% finally opens, both blinds know it is a real hand.
A narrow steal range means your opening range is your value range. Your opponents do not have to guess, do not have to balance, and never pay tuition — they simply fold what should be folded and fight back where they should, straight off your frequencies. What you saved was a handful of small pots. What you paid was readability in every big pot from then on.
Too high: big pots your range cannot support
The punishment for a high ATS arrives later and hits harder. Open with a pile of marginal hands and, as soon as the two blinds start 3-betting wide, you face the same question over and over: swallow the open and fold, or take K7s into a big pot you want no part of.
The key here is that the hands you open need a next move against a 3-bet. Able to 4-bet, able to call, or with a plan for after the flop — at least one of the three. A hand that is purely “nobody is in, let’s have a go” turns negative the moment opponents start hitting back.
One thing worth knowing: ATS is one of the two statistics in RushCraft’s player classification that can label somebody a maniac on its own (the other is 3-bet). The thresholds are 50% at micro stakes and 60% elsewhere, measured on the combined ATS across all three steal seats — cross it and, however normal the rest of your numbers are, that is the classification.
That line is further away than it looks. The three target bands above top out at 48%, so even hugging the ceiling on the button, in the small blind and in the cutoff, the weighted total does not reach 50; and the cutoff, which offers the most opportunities of the three (it only takes UTG and MP folding), drags the total further down with its 27% to 30%. A combined ATS touching 50 does not mean “I steal a bit more often”; it means at least one seat is opening far wider than it can support.
Compare VPIP and you can see what the design is guarding against. However high VPIP goes it never convicts on its own, because it needs PFR beside it to tell aggressive from loose — a high VPIP with a low PFR is loose-passive, which is a different type. ATS has no such ambiguity: by the time it applies there is nobody in front of you, and the only thing you can do is raise, so a high number has exactly one interpretation.
So the ceiling on ATS is not derived theoretically; it is set for you by the two players on your left. That is also why RushCraft’s diagnostic cards do not only look at your current number but compare your most recent 5,000 hands against your overall figure: if the deviation from the target band only became severe recently, that is usually not a strategy problem but a state-of-mind one (the card says so directly: watch for tilt).
Reading Your Opponent: His ATS Against Your BB Fold Rate
Everything so far has been about being the one stealing. But every steal has two sides, and a third of the time you are sitting on the other one.
Translate an opponent’s ATS into a single question: “how wide is his opening range from that seat?”
- The button on your right has an ATS of 30% → his opens have some substance, and your big blind defence should be selective
- The button on your right has an ATS of 55% → he raises with more or less anything, and you are facing a range of very low average strength
The number with which you answer him is Fold BB to Steal. RushCraft’s line for it:
Target 55% or below; above 65% it is flagged as a serious leak.
Why that order of magnitude? Because the 1bb in the big blind is already yours. He raises to 2.5bb, you add 1.5bb, and you get to see a flop in a pot that already holds 4bb — close to 2.7 to 1. You do not need much of a hand for that to be playable. Someone whose fold rate has climbed to 75% is announcing to the table that raising his blind is free money.
The two numbers should be read together:
| His ATS | Your BB fold rate | What is actually happening |
|---|---|---|
| 45% | 50% | Normal attack and defence; nobody is getting anything free |
| 55% | 75% | He is printing money: the widest range meets the thinnest defence |
| 28% | 50% | He is selective; you may be over-defending |
| 28% | 70% | He should raise his ATS, and you are both leaking |
Read the other way, that table is also the basis for adjusting your own ATS: you are not playing against a theoretical frequency, you are playing against the two people on your left. If they fold a lot, push toward the top of the band; if they fight back every time, pull it in.
(RushCraft’s diagnostic cards have a matching drill for both numbers: the ATS card opens the open/fold trainer for that seat, and the BB fold-rate card opens the three-way fold / call / 3-bet big blind defence trainer.)
Sample Size: 30 Opportunities Only Answers “Are You Stealing At All?”
The conclusion first: ATS accumulates a sample far more slowly than VPIP does, slowly enough to change what you are allowed to conclude from it.
VPIP gains a data point on every hand dealt. ATS does not — it needs you to be in one of three particular seats and everyone in front of you to fold. A rough estimate (assuming each player in front opens about 22% of the time):
| Position | Opportunities per 100 hands | Hands to reach 30 opportunities |
|---|---|---|
| CO | 10 | ~300 |
| BTN | 8 | ~380 |
| SB | 6 | ~490 |
RushCraft’s threshold is at least 30 steal opportunities per position; below that it answers “not enough data” and gives no conclusion. So it takes somewhere around four or five hundred hands before all three cards will speak.
But what can 30 tell you? Work out the interval and it is obvious. Assume a true button ATS of 45%:
| Opportunities | Observed number lands around | What that interval is good for |
|---|---|---|
| 30 | 27% – 63% | Only whether he is stealing at all |
| 100 | 35% – 55% | Separates conservative, normal and too wide |
| 250 | 39% – 51% | You can start having a conversation with the 43-48 band |
| 1,000 | 42% – 48% | The first time it sits inside the width of the target band |
(Normal approximation of the 95% interval; the orders of magnitude are right.)
Look at the last row: 1,000 button opportunities takes about twelve thousand hands. Which means something very practical — “is my button ATS 44 or 49?” is a question a few thousand hands cannot answer. Inside that range, everything you are looking at is noise.
So the right way to use ATS is bracketing, not tuning:
- 30 to 100 opportunities: confirm whether you are stealing at all. A CO at 15% or a button at 25% is a conclusion already; you do not need more data
- A few hundred: put yourself in the conservative / normal / too wide bracket and change your behaviour per the section above
- Beyond ten thousand hands: only now does “top of the band or bottom?” become a real question
Incidentally, in the AI coach’s style summary, ATS has to differ between recent and overall by 6 percentage points or more before it will flag “recently aggressive” or “recently conservative” — below that it says nothing. The threshold sits there not out of laziness but because anything said below it is an interpretation of noise.
If You Play Rush & Cash
The section on reading opponents is, once again, of no use on fast-fold tables — you fold and you move, you never play 100 hands against the same person, and the sample to read him never accumulates. That much is identical to VPIP.
But ATS has an advantage VPIP does not: of all the statistics you use to review your own game, it depends least on knowing anything about your opponents.
The reason is that the structure of the steal spot is completely unchanged by fast-fold. You still reach the CO, the button and the small blind; everyone in front still folds sometimes; the same 1.5bb still sits in the pot. The value of position holds without knowing who anybody is. Compare that with judgements like “does this fish bet at everything”, which reset to zero the moment the table changes.
Add that hands accumulate quickly on fast-fold tables, and the few hundred opportunities the sample-size section asks for takes an evening or two. The way to do it is to export your hand history from PokerCraft first (six steps, see downloading your Natural8 hand history); import it into RushCraft and the ATS for each of the three positions is computed separately, along with the distance between each one and its target band.
ATS comes down to one sentence: when nobody is in front of you, how often you actually act from late position.
- The denominator: only hands where everyone folded and you are in the CO, the button or the small blind. Calling spends the chance without counting as an attempt, and it is the most expensive cell
- The reason: at six-max about 25bb is taken from you every 100 hands, and late position is the only place you systematically pick it back up
- The positions: BTN 43-48%, SB 40-48%, CO 27-30%. Three different reasons, not three notches on the same scale
- Both directions: too low is a one-way subsidy plus transparency; too high is big pots your range cannot support. The ceiling is set by the two players on your left
- The sample: 30 opportunities only answers “am I stealing at all”; telling 44 from 49 takes tens of thousands of hands. Bracket, do not tune
Next time everyone folds around to you on the button and your hand looks like nothing much — remember that the 1.5bb in the pot is real, and that nobody has shown anything.