“No,” I say.

“Once is accident, twice is coincidence, and third time is enemy action. So if something that only might be aimed at you happens three times, then it’s time to consider someone’s after you.”

I puzzle over this a moment. “But… if it is enemy action, then it was enemy action the first time, too, wasn’t it? Not an accident at all?”

He looks surprised, eyebrows up and mouth rounded. “Actually — yeah — you’re right, but the thing is you don’t know about that first one until the others happen and then you can put it in the same category.”

“If three real accidents happen, you could think they were enemy action and still be wrong,” I say.

He stares at me, shakes his head, and says, “How many ways are there to be wrong and how few to be right?”

The calculations run through my head in an instant, patterning the decision carpet with the colors of accident (orange), coincidence (green), and enemy action (red). Three incidents, each of which can have one of three values, three theories of truth, each of which is either true or false by the values assigned each action. And there must be some filter on the choice of incidents, rejecting for inclusion those that cannot be manipulated by the person who may be the enemy of the one whose incidents are used as a test.

It is just such problems I deal with daily, only in far greater complexity.

“There are twenty-seven possibilities,” I say. “Only one is correct if you define correctness by all parts of the statement being true — that the first incident is in fact accident, the second is in fact coincidence, and the third is in fact enemy action. Only one — but a different one — is true if you define correctness as all three incidents being in fact enemy action. If you define correctness as the third incident being enemy action in all cases, regardless of the reality of the first two cases, then the statement will correctly alert you to enemy action in nine cases. If, however, the first two cases are not enemy action, but the third is, then the choice of related incidents becomes even more critical.”

He is staring at me now with his mouth a little open. “You… calculated that? In your head?”

“It is not hard,” I say. “It is simply a permutation problem, and the formula for permutations is taught in high school.”

“So there’s only one chance in twenty-seven that it is actually true?” he asks. “That’s nuts. It wouldn’t be an old saw if it wasn’t truer than that… that’s what? About four percent? Something’s wrong.”

The flaw in his mathematical knowledge and his logic is painfully clear. “Actually true depends on what your underlying purpose is,” I say. “There is only one chance in twenty-seven that all parts of the statement are true: that the first incident is an accident, the second incident is a coincidence, and the third incident is enemy action. That is three-point-seven percent, giving an error rate of ninety-six-point-three for the truth value of the entire statement. But there are nine cases — one-third of the total — in which the last case is enemy action, which drops the error rate with respect to the final incident to sixty-seven percent. And there are nineteen cases in which enemy action can occur — as first, second, or last incident or a combination of them. Nineteen out of twenty-seven is seventy-point-thirty-seven percent: that is the probability that enemy action occurred in at least one of the three incidents. Your presumption of enemy action will still be wrong twenty-nine-point-sixty-three percent of the time, but that’s less than a third of the time. Thus if it is important to be alert to enemy action — if it is worth more to you to detect enemy action than to avoid suspecting it when it does not exist — it will be profitable to guess that enemy action has occurred when you observe three reasonably related incidents.”

“Good God,” he says. “You’re serious.” He shakes his head abruptly. “Sorry. I hadn’t — I didn’t know you were a math genius.”

“I am not a math genius,” I say. I start to say again that these calculations are simple, within the ability of schoolchildren, but that might be inappropriate. If he cannot do them, it could make him feel bad.

“But… what you’re saying is… following that saying means I’m going to be wrong a lot of the time anyway?”

“Mathematically, the saying cannot be right more often than that. It is just a saying, not a mathematical formula, and only formulae get it right in mathematics. In real life, it will depend on your choice of incidents to connect.” I try to think how to explain. “Suppose on the way to work on the train, I put my hand on something that has just been painted. I did not see the WET PAINT sign, or it got knocked off by accident. If I connect the accident of paint on my hand to the accident of dropping an egg on the floor and then to tripping on a crack in the sidewalk and call that enemy action—”

“When it’s your own carelessness. I see. Tell me, does the percentage of error go down as the number of related incidents goes up?”

“Of course, if you pick the right incidents.”

He shakes his head again. “Let’s get back to you and make sure we pick the right incidents. Someone slashed the tires on your car sometime Wednesday night two weeks ago. Now on Wednesdays, you go over to a friend’s house for… fencing practice? Is that sword fighting or something?”

“They’re not real swords,” I say. “Just sport blades.”

“Okay. Do you keep them in your car?”

“No,” I say. “I store my things at Tom’s house. Several people do.”

“So the motive couldn’t have been theft, in the first place. And the following week, your windshield was broken while you were at fencing, a drive-by. Again, the attack’s on your car, and this time the location of your car makes it clear the attacker knew where you went on Wednesdays. And this third attack was accomplished on Wednesday night, between the time you got home from your fencing group and when you got up in the morning. The timing suggests to me that this is connected to your fencing group.”

“Unless it is someone who has only Wednesday night to do things,” I say.

He looks at me a long moment. “It sounds like you don’t want to face the possibility that someone in your fencing group — or someone who was in your fencing group — has a grudge against you.”

He is right. I do not want to think that people I have been meeting every week for years do not like me. That even one of them does not like me. I felt safe there. They are my friends. I can see the pattern Mr. Stacy wants me to see — it is obvious, a simple temporal association, and I have already seen it — but it is impossible. Friends are people who want good things for you and not bad ones.

“I do not…” My throat closes. I feel the pressure in my head that means I will not be able to talk easily for a while. “It is not… right… to… to… say… what… you are… not sure… is true.” I wish I had not said anything about Don before. I feel wrong about it.

“You don’t want to make a false accusation,” he says.

I nod, mute.

He sighs. “Mr. Arrendale, everyone has people that don’t like them. You don’t have to be a bad person to have people not like you. And it doesn’t make you a bad person to take reasonable precautions not to let other people hurt you. If there’s someone in that group who has a grudge against you, fair or unfair, that person still may not be the one who did this. I know that. I’m not going to throw someone into criminal rehab just because they don’t like you. But I don’t want you to get killed because we didn’t take this seriously.”

I still cannot imagine someone — Don — trying to kill me. I have not done anyone harm that I know of. People do not kill for trivial reasons.

“My point is,” Mr. Stacy says, “that people kill for all sorts of stupid reasons. Trivial reasons.”

“No,” I mutter. Normal people have reasons for what they do, big reasons for big things and little reasons for little things.