The unpunished violation

The first time I broke a dive rule and nothing happened, I learned the wrong thing.

It was a shallow dive in Phú Quốc, four metres maximum, on the back side of a small island, sixty minutes of slow patrol along a soft coral wall. The rule I broke was a small one. I had told myself before the dive that I would surface with no less than seventy bar of air. I surfaced with sixty-three. I had noticed the drift around minute fifty, watched it happen for ten more minutes, and then surfaced with seven bar less than I had committed to.

Nothing happened. The boat was right there. The visibility was clean. My buddy had been within arm's reach the whole dive. There was no scenario in which sixty-three bar versus seventy bar mattered, and I knew it, and that is exactly why I let it happen.

I want to write about what that day taught me, because it taught me the wrong thing first, and the right thing only after I had repeated the same drift twice more.

The wrong thing it taught me was that the rule had margin built in. The right thing it would have taught me, if I had paid attention, was that the rule did not have margin built in. The rule was the margin. The whole reason the rule existed was that the conditions in which sixty-three bar mattered are not the conditions in which I would be checking it. The rule was set for a dive I had not yet had.

This is the central problem of the unpunished violation. The dive on which you break the rule is almost never the dive on which the rule was written for.

The rule was written for the dive on which the current changes, or the visibility drops, or the regulator starts to free-flow at the worst possible time, or your buddy needs your octopus for two minutes longer than the script said. The rule is the margin for those dives. The rule is not the margin for the dive you happen to be on right now, which is calm.

Breaking the rule on a calm dive does not punish you. It cannot. The dive was not going to punish you anyway. The dive was a sample drawn from the easy tail of the distribution. The rule was written for the hard tail. The two tails do not talk to each other. So the calm dive teaches you nothing about the rule.

But you learn something anyway. You learn the wrong thing. You learn that the rule was conservative. You learn that the margin was generous. You learn that you can take seven bar back off the surface number and still be fine, because you were fine. The next dive, the surface number drifts to sixty. The dive after that, fifty-five. The dive after that, fifty. None of the dives punish you. The boat is always there. The buddy is always close. The conditions are always reasonable.

Then one dive, the conditions are not reasonable. The current has changed. The visibility has dropped to two metres. Your buddy is twelve metres away in the dark. You surface late. You surface stressed. You surface with thirty bar, on a day on which fifty would have felt like a safety net.

The dive that punishes you for the violation is not the dive on which you violated. It is the dive five months later, on which the conditions are different, and the rule you have been quietly relaxing across twenty dives was the thing that was supposed to hold for exactly these conditions, and it does not hold, because you have already moved the line.

There is a specific quality to this kind of feedback that makes it the most expensive kind. The market version of it is well known. You take a position size larger than your written rule, because the trade looks good. The position works. You make money. Nothing about the outcome teaches you that the position size was wrong. The outcome teaches you that the position size was right. The next trade, the position size drifts a little larger. The next trade, larger again. The trades work, because most trades work in a market that is going up. The position sizes drift upward across thirty trades. Then the market changes. The drawdown that arrives is sized for the new position sizes, not the old ones. The drawdown is twice what your rule was written to survive. You discover this on the day the drawdown arrives, which is also the only day on which the rule was supposed to be doing its job.

Nothing about any of the individual violations gave you information. Each violation, individually, was successful. The failure was the accumulation. The accumulation is invisible. There is no feedback signal for accumulation. There is only the feedback signal for any given violation, and the signal for any given violation, in benign conditions, is positive.

This is what I mean when I say the unpunished violation is the most expensive feedback you can receive. The cost is not the violation. The cost is the lesson the violation teaches, which is the wrong lesson, delivered in the most persuasive form, by an outcome you cannot argue with.

I have come to think of the unpunished violation as a small, free transaction in which I am the buyer and discipline is the seller. Every time I break a rule and survive, I am buying a small piece of permission from my future self. The permission is cheap. The permission is also nontransferable: I cannot give it back. Each piece of permission I buy makes the rule a little less mine. After enough purchases, the rule does not belong to me at all. It belongs to whoever I was a year ago, and the diver in the water today does not know that person any more.

I do not have a perfect cure for this. The cure I have been practising has three parts.

The first part is to record the violation when nothing happens. The dive on which I came up at sixty-three bar is on this site, in the log, because I wrote it down. If I had not written it down, it would already have left my head. The writing is the only friction the cheap transaction has. Writing the violation makes the violation cost something, in time and in a small private embarrassment that is the entire point. The friction is what makes the next violation a little harder. Not impossible. Just a little more expensive than the last one.

The second part is to refuse to interpret the lack of punishment as information about the rule. The lack of punishment is information about the conditions of the specific dive. The lack of punishment is not information about whether the rule was conservative. I do not let myself update the rule in the direction of the violation. I am allowed to update the rule. I am not allowed to update the rule on the evidence of a single calm dive in which I broke it.

The third part is the hardest. I am trying to be specific about which kind of violation I am committing when I commit one. Some violations are drift. They are small, repeated, and erode the rule across time. Others are emergencies. They are large, one-off, and reflect a real reason that overrode the rule in a specific situation. Drift violations are the dangerous ones, because they buy permission without ever announcing themselves. Emergency violations are not dangerous if I can name the reason out loud. The rule is allowed to bend for an emergency I can describe. The rule is not allowed to bend for an erosion I cannot describe.

The trade-off this misses is the obvious one: I am writing as if I can tell the difference, and the whole problem with drift is that you cannot. Drift is what slips by below the threshold of attention. I am not going to catch every drift event. I am going to miss some. The writing is the way I catch most of them. The friction is the way I make the ones I miss a little less frequent than they would otherwise be. I am not solving the problem. I am buying the problem time.

The diving version of this essay is the surface number drift. The market version is the position size drift. Both are visible across many small events. Neither is visible in any single event. The cure is the same. Record the violation. Refuse to interpret the lack of punishment as evidence about the rule. Name the kind of violation, drift or emergency, before you let yourself proceed.

The buddy check was the entry condition for everything else. The unpunished violation is the first thing the buddy check is supposed to catch.