Daily and weekly loss limits

The morning that stops 

Two sessions, the same account, the same instrument, the same three losing positions in the same order before noon. In the first, the running total is compared against a figure written down before the market opened, the total is past it, and the terminal is closed until the following session. In the second, the question of whether to carry on is put for the first time at that moment, by someone who has just lost money three times in a row. The market treated the two accounts identically. What differed was when the stopping decision was made.

That is the mechanism, and it is worth stating before the arithmetic, because the arithmetic is trivial and the mechanism is not. A loss limit moves one decision out of the middle of a losing session and into a moment before that session exists. The number is the visible part. The transfer of the decision is the part that does the work.

What the rule has to state 

A stopping rule stated loosely is not a rule, because the moment it binds is exactly the moment its ambiguities get argued about, and the argument is held by someone with an obvious interest in the outcome. A rule that can be enforced settles four things in advance.

• The quantity. A money amount, or a fraction of the equity the window opened with. Either way it is fixed when the window opens and does not grow as the account grows inside it.
• What is counted. Realised results from closed positions only, or realised results plus the unrealised result of everything still open. On the same afternoon those are two different numbers.
• The window. A calendar day, a trading session, a calendar week, or a rolling count of sessions. A daily rule and a weekly rule are two rules, and both can be live at once.
• The action. Whether open positions are closed when the limit is reached or left to their own exit levels, and what stops: new positions, additions to existing ones, or all activity until the window resets.

The fourth is the one most often left blank, and it decides what the rule actually is. A limit that stops new positions while leaving open ones to run constrains activity. A limit that closes everything constrains exposure. Neither is put forward here as the correct reading. What matters is that a rule which has not made the choice will have it made, in the moment, by whoever is losing.

The middle row repays attention. The third position was opened with less room remaining than its result went on to consume, so the limit was not passed by a decision to keep trading after it was reached. It was passed by a position already open while the total was still inside the rule. A limit counting closed results is tested after each position rather than during it, so it is discovered rather than enforced unless the size of a late position is derived from the room remaining. The sizing methods earlier in this module know nothing about a daily total, and reconciling the two falls to whoever writes the rule.

Why it is written before the session 

The trader who writes a limit on a quiet evening and the trader who meets it on a bad Tuesday are the same person in two very different conditions. On the evening the question is abstract and the answer costs nothing. On the Tuesday the account is down and the options no longer look symmetrical: carrying on is framed as the route back to where the session started, stopping as accepting a loss rather than declining a larger one. That reframing is neither unusual nor discreditable, and naming it is not a claim about anyone's character. It is the reason the decision is conventionally taken out of that moment entirely.

The device has a name. A pre commitment is a decision taken at one time whose function is to remove an option at a later one, and its defining property is that it binds when it is least welcome. That is not a defect in it: a constraint that applies only when convenient is not a constraint. The cost is real, though. A rule written in advance will sometimes bind on a day when, with hindsight, a different course would have read better, and a rule that could be suspended on those days would be suspended on all of them.

A stopping rule of this kind is often called a circuit breaker, a name borrowed from exchanges, where trading in a market halts automatically once an index has moved past a stated threshold and nobody is consulted at the moment of the halt. The analogy carries the automatic character of the trigger and misses the enforcement, because an exchange halt is imposed by a venue on every participant at once while a personal limit is enforced by the same person it constrains. That difference is the whole practical difficulty of the subject.

A limit on loss and a limit on number of trades 

The two are usually discussed together and they constrain different quantities, so neither substitutes for the other. A limit on cumulative loss binds magnitude and says nothing about how many positions produced it: one position can pass it before a second exists. A limit on the number of positions binds frequency and says nothing about size: a window can reach the count with a negligible net result, or pass a money limit on its first position while the count is nowhere near its boundary. A third convention limits consecutive adverse results regardless of size, a count again, of a different thing.

Session C is the awkward one, and the awkwardness is the point. What a count constrains is not money at all. It is the rate at which decisions are made, on the reasoning that the number of positions opened in a window is a quantity a trader states while their results are not.

Enforcement without discretion 

A limit renegotiated at the moment it binds is not a limit. It is a preference with a number attached. The renegotiation has recognisable forms, recognisable because they sound reasonable rather than obviously self serving: excluding an open position from the total because it is expected to come back, treating the rule as applying from the following session, switching the window from the day to the week because the week is still positive, or opening one more position because it is unrelated to the ones that produced the total. None of those arguments was available before the session started, and all of them arrive together.

Which is why the conventional answer is to have the rule enforced by something other than the person it constrains. Whether a particular control exists is a question about a platform rather than about the rule, and the crude version works on the same principle as the sophisticated one: a terminal that is closed, a record written when the limit was reached rather than reconstructed afterwards. The rule itself belongs in the written document that already carries the sizing rules, because a rule held in memory is restated from memory at the moment it binds, and a restated rule is a renegotiated one.

What a limit is actually defending against 

It is not defending against the loss that triggered it. By the time a limit binds that money has gone, and no rule written in advance reaches back and makes it smaller. What the rule bounds is the number of further decisions taken in the state the loss produced. Those decisions are the subject of the final lesson of this module: size raised to recover a session in one position, a losing position held past its stated exit level because closing it would make the total final, a position taken for no reason except that the window is negative. A limit makes none of them less tempting. It ends the window in which they are available.

Where practitioners disagree 

The first argument is whether the calendar corresponds to anything. One tradition holds that an arbitrary boundary is still a boundary, and that a boundary drawn somewhere is what stops a bad session continuing into a worse one. Another points out that no market has an interest in a Tuesday, that a limit reached at noon relocates the same activity to the following morning rather than removing it, and that the window is a property of the trader rather than of the instrument. A rolling window across recent sessions answers the second objection and inherits the first, since the length of the roll is chosen by the same person.

The second is what the total counts. A rule measured on closed positions only is unambiguous and easy to check, and it can be deferred indefinitely by not closing anything, which turns a stopping rule into a reason to hold. A rule measured on equity including open positions closes that gap and opens another, because an excursion that later reverses can trigger it, and the window then ends on a total the account never recorded. Both objections are accurate, and the disagreement is about which failure is preferred.

The third concerns nesting and cost. A weekly limit set without reference to the daily one either binds before the daily rule can ever fire, which makes the daily rule decorative, or sits so far above it that no plausible week reaches it, which makes the weekly rule decorative instead. Beyond the arithmetic sits a broader objection: a rule that ends a window removes a trader from the market at a moment that may be informative. The answer conventionally given is that indifference to that argument is the function of a limit. That restates the commitment rather than refuting the objection.