How TTMT Catches a Signal With a Decimal Error In It

BUY EURUSD 1.1850, SL 1.0850, TP 1.1920

Read it quickly and it looks like a trade. Read the stop again. It is a hundred pips away, if a hundred pips were a thousand. A digit slipped, and a stop that was meant to sit thirty pips below entry is now sitting a thousand below.

Place that as written and you have a position whose worst case is roughly thirty times what you agreed to. If you size by lots rather than by risk, you will not even notice until it matters.

Providers are people typing on phones. These messages happen. What a copier does with them is, I think, the most important thing about it, so here is what ours does.

Why simple bounds do not work

The obvious approach is a sanity range. Reject anything more than some percentage away from the current price.

It fails in both directions. Set the band tight and you reject legitimate swing signals with wide stops, or gold signals during a week when gold is moving a hundred dollars a day. Set it loose and 1.0850 passes, because on a percentage basis it is only nine per cent away from 1.1850, which is not obviously absurd for an instrument that could be anything.

Worse, a fixed band has no way to tell a wrong number from an unusual one. Those need different handling and the difference is contextual.

Phase one: collect references worth trusting

Rather than comparing against one thing, we gather everything we know about where this instrument's prices should be, and rate each source by how much it deserves to be believed.

Live market price is the strongest. It is a fact, fetched from the broker, at this moment. If EURUSD is trading at 1.1848, that is not an opinion.

Existing state on the account comes next. If you already hold positions or orders on this symbol, their prices are real prices that a real broker accepted.

Cross-validated values within the signal itself come after that. If the entry and two of the three targets all cluster around 1.185, they corroborate each other. Three numbers agreeing is weak evidence individually and meaningful collectively.

Your own configured defaults are weakest but not worthless. If your default stop distance is thirty pips, that tells us the rough scale of stop you work with.

The weighting matters. A strong reference should be able to overrule a weak one, and a stack of weak references should not outvote a live price.

Phase two: score every number

Each value in the signal, the entry, the stop, every target, is put through several independent checks.

Magnitude. Is this number within a sane multiple of the reference price? A value ten times off, or a tenth of where it should be, is a decimal error, not a trading opinion. This check carries the most weight because it catches the most dangerous class of mistake, and a number that fails it badly enough is rejected on that alone.

Distance from the anchors. How far is this from the prices we trust, measured in pips for this instrument? A stop four hundred pips from a spot price on a major pair is not impossible. It is unusual enough to count as evidence.

Peer consistency. Does this value agree with the other values in the same message? In our example, the entry and the target sit around 1.185 and the stop sits at 1.085. The stop is the outlier. Three numbers agreeing and one disagreeing is a strong signal about which one is wrong, and it works even when we have nothing else to go on.

Each check produces a score. One check failing conclusively is enough to reject the value. Several checks all scoring moderately badly also add up to a rejection, because a number that is a bit wrong on every measure is usually just wrong.

What happens after a number is rejected

It depends which number it was, and this is the part where the design has an opinion.

A bad stop is replaced from the best reference available, usually your configured default distance from the validated entry. A stop is the thing preventing an unbounded loss, and refusing to trade because the stop was mistyped means declining a setup that was probably fine.

A bad entry usually rejects the whole signal. Everything else is measured relative to the entry, so if we cannot trust it we cannot trust the rest.

A bad target is dropped and the remaining targets are used. Losing TP3 costs you some upside. It does not put your capital anywhere it should not be.

There is one thing no configuration can change. A market order on an alert with no stop loss is refused, on the server, every time.

Where the check runs in the pipeline

Before validation, not after, and the order is deliberate.

If you validate first and sanitize second, you validate numbers that are about to change, and every conclusion you drew is against the wrong values. Sanitize first and the downstream checks are working on the numbers that will actually be placed.

What it cannot do

It cannot catch a wrong number that looks right. If a provider means 1.1820 and types 1.1830, every check passes, because 1.1830 is a perfectly sensible price. We are catching implausibility, not reading minds.

It cannot catch a bad trade. A well-formed signal with sensible prices and no edge behind it sails through, as it should. This is a typo filter, not a judgement about whether the setup is any good.

And there is a tail risk in the other direction: an unusual but legitimate signal can be caught. A genuinely wide stop on a volatile day, or an entry level a long way from spot on a limit setup, can score badly enough to be treated as an error. We would rather be wrong in that direction. Skipping one valid trade is recoverable. Placing a position with a stop a thousand pips away is sometimes not.

Related: what happens in the two seconds after a signal posts, and what lot size actually means.

Start a free 7-day trial at telegramtometatrader.com.