The yellow line in a tracker, all-in EV, promises to show the "real" result with luck removed. That is why people love it: if it is above the green line, you play better than the balance shows. That is partly true, but a much smaller part than it seems. The yellow line is the most popular graph in poker for one reason: for almost everyone it sits above the green one.
What all-in EV counts
The tracker looks only at hands where the money went in before the river with the cards face up. For each such hand it computes your equity at the moment of the all-in and records the expected win instead of the actual one. The difference between the sum of actual results and the sum of expected ones is EV diff.
If you went all-in on the flop with 70% equity into a 100 bb pot and lost, the green line records −50 bb, the yellow one +20 bb. The 70 bb difference is your "bad luck" in that hand. The tracker offered its sympathy but did not refund the money.
What all-in EV does not see
Here is the main error in reading it. All-in EV accounts for luck in only one type of hand: where there was an all-in before the river. Everything else it counts as is.
- You bet the turn with a flush draw, the opponent folded, you took the pot. Luck? No, by all-in EV that is just a result.
- You called the river with a bluffcatcher against a bluff. By all-in EV that is pure skill, although next time in the same situation the opponent would have had it.
- You were dealt AA three times less often than average, or coolers went your way. That is luck too, and the yellow line does not see it.
All-ins before the river are a few percent of hands. The spread of results in the remaining hands the yellow line smooths exactly as much as the green one, that is not at all. It does not measure luck. It measures one variety of it, the kind that is convenient to count.
When EV diff is useful
Three situations where it is worth looking at.
A short stretch with a large EV diff. If over 20,000 hands you are at −30 bb/100 in all-ins, that explains a noticeable part of the downswing and helps you not fix what works. This is the most honest use: not "I play well" but "this specific part of the minus is random".
Spots with frequent all-ins. Defense against 4bets, stack-off play in 3bet pots on the flop, shoves with a short stack. In them all-in EV shows whether your decisions to go for the stack pay off, regardless of how the cards fell.
Comparison with the sample. On 200,000 hands an EV diff of ±5 bb/100 in all-ins is normal, ±15 is already rare. If your difference is outside the usual range on a large sample, check for a systematic cause: for example, you often go all-in with draws that the opponent calls with a better hand, and that is recorded as "bad luck" although the line itself is losing. Bad luck that repeats every month has a different name.
What you cannot do with it
- Add EV diff to the winrate and call the result "real". The real result is unknown to both lines equally.
- Explain a minus by bad luck if EV diff is small. Bad luck hides outside all-ins too.
- Judge the quality of play by EV diff. It describes only how the cards fell in all-ins, and nothing about whether you should have been in them. Losing with 70% is bad luck. Regularly finding yourself with 30% is not.
What to count instead
The quality of play is better measured by the price of decisions, not by lines. In one and the same spot, say a call against a river bet in a single-raised pot, you made the decision thousands of times. On such a sample the result relative to the alternative converges much faster than the overall winrate, and it does not depend on how the river fell in a specific hand. The yellow line answers the question "was I lucky in all-ins". The question "where does the money go" is answered by spots.